mobile big data

48
MOBILE BIG DATA THE RIGHT INTELLIGENCE, RIGHT NOW ! Intelligence: The knowledge of an event, circumstance - received or imparted Brian Blackmarr Fusion Labs, Inc. April 2014

Upload: lisa-nlly

Post on 07-Mar-2016

223 views

Category:

Documents


1 download

DESCRIPTION

 

TRANSCRIPT

Page 1: Mobile Big Data

MOBILE BIG DATA THE RIGHT INTELLIGENCE, RIGHT NOW !

Intelligence: The knowledge of an event, circumstance

- received or imparted

Brian Blackmarr

Fusion Labs, Inc.

April 2014

Page 2: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 2

CONTENTS

I. EXECUTIVE SUMMARY………………………………………………………………………....05

II. INTRODUCTION………………………………………………………...................................06

III. “WHY” IS MOBILE BIG DATA A TRUE BUSINESS ADVANTAGE?................................... 09

A. Empowering Better Decisions……………………………………………….….. 10

High Knowledge Intensity : The Three Vs of Data……………………….….. 11

Liquid Data, Free Range Data and Data Combo……………………….… 12

Customer Facing: Living at the Edge…………………………………………..13

B. Minimizing Decision Latency, Speeding Time to Value…………………… 14

The Need for Speed and the Race to Zero…………………………………. 14

Real Time is Money: Halting Fraud and Enabling Mobile Marketing…... 16

Democratizing Intelligence…………………………………………………...….16

C. Leveraging Market Trends and User Preferences………………………….. 17

Petabyte is the New Terrabyte and Gigaflop Blow-Out………………... 18

Crowdsourced Intel: Likes, Tweets and Twerks………………………….…. 20

BYOE Phenomenon………………………………………………………………. 21

D. “Why” Mobile Big Data Takeaways…………………………………………… 21

IV. “WHAT” ARE THE CORE CAPABILITIES OF MOBILE BIG DATA ?...............................23

A. Industrial Strength Processing of Big Data……………………………………..23

MPP, Cloud Option and HAL Jr……………………………………………….. 24

Speed Dating Data: Hadoop/MapReduce………………….……………. 25

Dark Data, Active Archive and High Density Processing………………... 27

Fabric-Like InfiniBand Topology……………………………………………….. 27

B. Mobile Based Interfaces, Access and Applications………………………. 28

User Mobility with Preferred Platform…………………………………………. 28

Provisioning Actionable Intelligence…………………………………………. 29

Mobile Data Capture and Platform Transparency……………………….....30

C. Avanced Analytics Derrived Intelligence……………………………………...31

Strategic Analytics and Analytics in Motion…………………………………..31

Page 3: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 3

Complex Event Processing……………………………………………………… 32

Soft Science: Market Listening and Data Visualization…………………… 33

Self-Serve Analytics………………………………………………………………. 33

D. “What” Is Mobile Big Data Takeaways…………………………………………35

V. “WHERE” DOES MOBILE BIG DATA FIT BEST? .............................................................36

A. ROI Rich Environments Profile…………………………………………………….37

Big Data Extension……………………………………………………………….. 37

Financially Valued Actions……………………………………………………… 38

Complex Event Response………………………………………………………. 38

Customer Facing Edge………………………………………………………….. 39

B. Real World Examples ……………………………………………………………. 39

Minimizing Disruption in Complex Transportation Network……………….. 39

Edge Worker Communication and Citizen Mobilization………………….. 40

Improved Customer Retention and Cross Selling…………………………...40

Accelerating Product Claims Response……………………………………... 41

VI. RECOMMENDED MOBILE BIG DATA BEST PRACTICES……………………………….42

A. Organizational Recommendations…………………………………………… 42

Management: Stand and Deliver…………………………………………….. 42

Empiricists vs. Analysts: MIA Skills……………………………………………. 43

User and Tech Staff Preparation………………………………………………. 44

B. Implementation Best practice…………………………………………………. 44

Make vs. Buy vs. Both…………………………………………………………….. 45

Interative Piloting with Metrics…………………………………………………. 45

Step-Wise Extension……………………………………………………………… 45

GLOSSARY OF TERMS AND ABBREVIATIONS………………………………………………… 47

©2014 by Fusion Labs, Inc. All rights reserved. Reproductions in whole or in part are prohibited except by written

permission. Email requests or feedback to [email protected] Product and company names mentioned

herein may be trademarks and/or registered trademarks of their respective companies.

Page 4: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 4

ABOUT THE AUTHOR

Brian Blackmarr is the Chairman and a co-founder of Fusion Laboratories Inc. Brian

is a former IBM senior engineer and has authored more than 200 technical articles

and publications. He is a Registered Engineer in the State of Texas and has more

than 20 years experience serving as a NASDAQ company Director and a Trustee of

the A+M Research Foundation. He’s personally designed and developed

numerous statistically based decision models addressing complex situations in a

variety of industries (banking, electronics manufacturing, airlines, aerospace, oil

and gas, federal and state agencies, U.S. military, etc.), has authored technical

journal reports and taught senior IT management seminars internationally. Brian has

an MS in Operations Research from the Mechanical Engineering School of the

University of Texas at Austin and has received awards and nominations including

High Tech Exporter of the Year, E+Y Entrepreneur of the Year, etc.

ABOUT FUSION LABORATORIES INC.

Fusion Laboratories Inc, Fusion Labs, is headquartered in Dallas, Texas with offices in

Charleston, SC. and Houston, Tx. Fusion Labs is focused on the development and

support of a variety of specialized application software. Fusion Labs provides a

suite of proprietary software to the large non-profit foundation market and

previously provided proprietary advanced supply chain software through its former

subsidiary, RFID Systems Inc. Fusion Labs and its predecessor BRBA ( became

Brightstar Technologies, a NASDAQ company ) have consulted internationally

regarding complex large scale information systems design, development and

operation and have provided ongoing hosting and managed services support for

major international applications.

Fusion Labs has developed numerous mobile platform commercial applications for

retail, medical, insurance and financial sectors and recently celebrated

exceeding 100,000 downloads of its ChromeRDP utility, a Mobile application cross-

platform ulitily developed by Fusion Labs in partnership with Google Inc.

Additionally, Fusion Labs has several major Mobile platform development efforts in

process directed at medical and health care specific environments. Fusion Labs

also has channel partner relationships with major Big Data related suppliers

including SGI., Silver Peak, Solid Fire, Centerity, etc. and is regularly a joint sponsor

of Big Data related events and conferences. In the U.S. Fusion Labs also supplies

and supports Fuzed, a high security oriented social media type system for

enterprise employee interconnection and communication. The website

fusionlabs.net provides an overview of the Fusion Labs proprietary software and

service offerings

Page 5: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 5

I. EXECUTIVE SUMMARY

Mobile Big Data is above all a practical business tool having a unique capacity to

provision easy to use and up to the minute business intelligence to decision makers

at all levels and all locations within an enterprise. Although Mobile Big Data

continues to rapidly evolve and improve, it’s based on the consolidation of solidly

mature technologies and its functional capabilities have been deployed to

address a number of genuine real world, and fully ROI driven, business needs.

The clear objectives of Mobile Big Data are to (1) enable better quality decisions,

especially in dynamic and complex environments and (2) reduce the decision

latency and thus reduce the “Time to Value “ of business decisions and optimize

outcomes with time critical opportunities. Mobile Big Data meets these objectives

by effectively discerning and delivering multi-type multi-source derived business

intelligence from internal files, IoT liquid data and social media. These strong

Mobile Big Data capabilities represent a convergence of the massive data

handling capabilities of Big Data, the user preferred intuitive interfaces and

delivery of pervasive Mobile platforms and the sophisticated data analysis of

Advanced Analytics tools. The functional confluence of these established

technologies provides the proven capabilities of Mobile Big Data. The business

advantage of using those capabilities will be a 20% improvement in all financial

metrics per the Gartner Group.

The sweet spot for Mobile Big Data deployment starts with enterprises having Big

Data in place, those with customer facing edge workers (knowledge workers at

the organization’s physical and logical perimeter) and complex event based

environments with ever changing variables. ROI rich deployment opportunities

may also include financial transaction based organizations with high value

activities, fashion oriented retail, data driven medical, complex airline/rail

networks, public utilities, multi-level manufacturing and major public sector

organizations. In each case the ability to provision timely and fully actionable

business intelligence directly to the enterprise knowledge workers is the business

advantage of Mobile Big Data.

Certainly “ the winds of change are sweeping across the land “ is a gross over

statement, but just as certainly, a variety of practical needs and technology

factors now make Mobile Big Data, and the business intelligence it can cost

effectively provision, a valuable business tool and potentially as a true competitive

edge.

Page 6: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 6

II. INTRODUCTION

The mission statement of this white paper is to introduce and outline the currently

available and ever evolving capabilities of Mobile Big Data. Taken separately, the

three primary component elements of Mobile Big Data are significant in

themselves. Equally impressive are (1) Big Data’s ability to efficiently handle

today’s immense and disparate data files, (2) Mobile platform’s ease of use for

directly delivering user information and (3) the ability of Advanced Analytics to

efficiently discern valuable business intelligence. Taken together, these distinct

technologies combine to provision seamless and timely business intelligence in

support of the targeted user’s decision making. Facilitating a more effective user

decision making process is especially important with the complex and time critical

decisions common to rapidly changing enterprise environments. The combined

Mobile Big Data technical capabilities can also assure that users receive cross

referenced information derived from multiple disparate input data sources,

including unstructured, and providing a solid contextual point of reference. The

net Mobile Big Data objective is better quality decision making with significantly

reduced decision latency.

This white paper will introduce the Mobile Big Data features and functions

including a general overview of the business case rationale and potential ROI

available through enterprise level implementation. The white paper’s enterprise

level orientation is primarily due Mobile Big Data always needing to show a valid

ROI, so implementation costs today will likely require a degree of scaling. To better

illustrate the practicality of Mobile Big Data concepts a set of enterprise level

examples are included. Mobile Big Data is continuing to evolve rapidly, both from

a technical capabilities standpoint and from a business case/ROI perspective.

Because Mobile Big Data continues to evolve, the contents of this white paper

should not be considered as fully definitive. A comprehensive tutorial for Mobile Big

Data implementation would be premature. This white paper serves as an

introduction to Mobile Big Data concepts and as a general guideline for its

potential enterprise level deployment.

Largely for practical reasons, the truly accurate white paper title of “Mobile Big

Data - Provisioning Vital Decision Support Information to the Appropriate Enterprise

Level Persons in a Timely and Actionable Manner “ has been greatly shortened.”

Nonetheless, the focus of this white paper is the ability of Mobile Big Data to

improve and speed the decision processes typical of an enterprise and to provide

a significant ROI as a result. In this context a serious effort has been made to

minimize the hype seemingly attendant with any such emerging technology. This

white paper is primarily directed at addressing the following key questions;

(1) Why is the deployment of Mobile Big Data a true business advantage?

Page 7: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 7

(2) What are the principle components and capabilities of Mobile Big Data?

(3) Where does Mobile Big Data likely fit best and provide a solid ROI?

(4) What Best Practice recommendations apply to Mobile Big Data?

The white paper’s ROI based focus derives from the fact that the rationale for

significant technology investment should be driven by its ability to effectively

achieve specific and well defined business requirements (with measurable results

metrics and thus calculable ROI) rather than being an implementation based on

technical capabilities. Mobile Big Data is also best considered as being highly

individualized and customized, definitely not a generic “plug and run” type

technical offering. It should be noted that successful to-date Mobile Big Data

examples (including described herein) have been specific business needs driven

and have avoided the generic “ hammer in search of a nail “ type approach.

In moving ahead with a Mobile Big Data implementation the white paper’s

recommended methodology outlines a fairly cautious sequential stepwise effort

rather than being a “Big Bang Theory “approach. The typical implementation

gating factors for Mobile Big Data are not unique to the technology and clearly

include (a) senior management buy-in and visible support (b) end-user

management buy-in and visible participation (c) significant technical staff skills

ramp up and (d) solid user start-up support with full ROI based feedback. Properly

addressing the organizational related issues of a Mobile Big Data deployment is at

least as important as properly addressing the technical concerns. Technically

compounding Mobile Big Data implementation, however, is likelihood for the

inclusion of unstructured data (sensors, etc.) and the potential usage of somewhat

irregular external data (social media, etc.). A variety of Mobile Big Data

implementation alternatives exists (e.g. buy vs. build, etc.) to mitigate possible

technical issues and these are also included.

A general overview of Mobile Big Data concepts best begins with a visualization of

the confluence of the three primary component elements as shown by Figure 2.1.

Page 8: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 8

Figure 2.1: Component elements of Mobile Big Data

Figure 2.1 visualizes the confluence of the primary Mobile Big Data components of

Big Data, Mobile platforms and Advanced Analytics to enable the seamless

provisioning of fully actionable business intelligence to the decision making

knowledge workers at all levels and locations of the organization.

The primary sourcing of the information contained in this white paper includes

direct experience, a variety of technical publications and the specific capabilities

of current product offerings.

Big Data

Advanced Analytics

Mobile Platform

MOBILE BIG

DATA

Page 9: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 9

III. “WHY” IS MOBILE BIG DATA A TRUE BUSINESS ADVANTAGE?

The clear focus of Mobile Big Data is intelligence provisioning for decision making

employees, especially those in complex environments. The business value of

providing effective decision support intelligence is always dependent on the

situation but can be quite considerable. To emphasize the need to evaluate the

deployment of Mobile Big Data as a business advantage for intelligence

provisioning, versus just continuing with a “ business as usual “ approach, consider

these findings from several recent studies;

(1) Knowledge workers, persons who regularly make some form of business

decisions, currently represent approximately 80% of all enterprise level

employees and less than half believe they’re appropriately trained and

properly supported.

(2) Approximately 80% of knowledge workers regularly access social media

and many use the information obtained in their business related decision

making.

(3) About 70% of knowledge workers utilize informal data analysis tools

(spreadsheets, ad hoc reports, etc.) to define their own intelligence, which

is then used in decision making.

(4) Over 50% of knowledge workers consider the decision support intelligence

formally supplied to them by their IT function not to be complete, timely or

actionable.

Clearly, knowledge workers aren’t that satisfied with the business intelligence they

now receive formally so have moved on to informally obtain and analyze their

own decision support information. Under this scenario complex business decisions,

some having critical customer facing and bottom line implications, are regularly

being influenced by, and perhaps even based on, disparate information typically

taken out of context and of indeterminate accuracy.

The business advantage “Why“ of Mobile Big Data derives directly from this serious

situation and its ability to provide timely, easy to use and fully actionable business

intelligence to decision makers at all levels and locations of the organization. These

Mobile Big Data capabilities are directed at (1) empowering higher QUALITY

decision making (especially in complex environments) and (2) reducing the

LATENCY of the decision making process to take better advantage of time critical

opportunities and thus speed the “ Time to Value”.

Page 10: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 10

To quote David Boyle, SVP of Insight at EMI Music, regarding provisioning business

intelligence to their knowledge worker employee base;

“I want quick, but I want complex.”

Delivering such business intelligence is exactly the focus of Mobile Big Data. The

fairly extensive range of likely enterprise benefits that could result from improved

business intelligence provisioning, in the opinion of the actual decision makers, is

displayed in Figure 3.1.

Figure 3.1: Potential Benefits of Better Business Intelligence

Source: The Economist

It’s key to note from Figure 3.1 that better decision quality and faster decision

making are the primary drivers for most of the majority of these important potential

benefits.

A. EMPOWERING BETTER DECISIONS

The interest in improving decision quality frequently relates directly to the poor

business intelligence now being supplied to decision makers. Today’s business

intelligence may be (1) restricted by being based on limited data types and/or

sourced only from the internal data of the Enterprise Data Warehouse (EDW), (2)

based on stale information that’s become more historical than actionable or (3)

delivered in a confusing and hard to use format. The foundational capabilities

intrinsic to Mobile Big Data focus totally on the provisioning of timely intelligence

derived from multi-source multi-type data and delivered in an intuitive and

actionable format.

Page 11: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 11

High Knowledge Intensity: the Three Vs of Data

Business intelligence must be knowledge intense, have solid content, to be

valuable and achieving that goal begins with the raw input data being analyzed.

In producing effective business intelligence it’s key to utilize the best possible input

data for analysis. To do this it’s essential to understand and effectively deal with all

“three Vs “of source data.

Volume - How much source data is available and appropriate for analysis?

Velocity – How fast is the source data flowing and what is its useful shelf

life?

Variety – How many distinctly different types of source data are

appropriate?

The ability to successfully account for and use these three Vs of source data can

have a huge impact on usability of the business intelligence in support of critical

decisions. Clearly there will always be compromises and tradeoffs to the general

rule that the more distinctly different data types/sources, and the larger the

sample size, the better will be the resultant intel. The judicious use of front end data

filtering, compression and sampling techniques may also seriously lessen the

processing burden by reducing storage and analysis volumes as much as possible.

Producing knowledge intense intelligence for an affordable cost is at the heart of

Mobile Big Data. Figure 3.2 displays the various data types that intelligence

analysts consider should be used by them in defining business intelligence.

Figure 3.2: Preferred Business Intelligence Input Data Types

Source: Several Recent Surveys

Page 12: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 12

As displayed, decision support intelligence should be derived from a wide variety

of distinct data types and data sources.

Liquid Data, Free Range Data and Data Combo

The term “liquid data“ refers to the streaming data available as a result of the

emerging “Internet of Things” (IoT). Recall that today the internet connects far

more “ things“, devices of all types, than people and this IoT trend is accelerating.

The available streaming data is often derived from sensors, video feeds, etc. and

typically consists of largely unstructured bit streams. The processing implications for

effectively analyzing the IoT streaming data, with the huge data volumes involved,

have recently driven many Big Data initiatives. Liquid data is an ever flowing

current of semi-structured or unstructured bits as opposed to the neat stacks of

discrete data elements (often relational in nature) typically found in an enterprise’s

Master Data Base (MDB). For process control, etc. this is important data, frequently

with a short shelf life, and the use of periodic sampling and range limit monitoring

can be appropriate. Using these methodologies to analyze steaming data to

derive business intelligence, in support of and real time decision support, is

practical, however, the set up may be a non-trivial effort. The detection of a liquid

data analomy typically requires an algorithmic based analysis to determine if the

inconsistency was likely random or the beginning indicator of an important

pattern. This type analysis of liquid data may require considerable real time (and

in-memory) processing capability using fairly specialized analytical tools. With such

provisions Mobile Big Data can fully support liquid data as a data type.

The term “free range data“ describes most of what may be obtained directly from

social media sources. The good news is that, unlike liquid data, social media data

typically has some degree of structure; the bad news is that this structure is likely to

be in the form of textual files bordering on gibberish. Along with understanding the

anticipated variances and inflections of text data (similar to voice recognition)

there’s also an embedded amalgam of one-off abbreviations, emoticons, etc.

Free range data can appear to wonder all over the language landscape and

serious efforts to categorize it, boil it down or filter it run the risk of reducing or losing

the spontaneity and raw meaning of the data - often the main value proposition

for using a social media data. However, the advanced social media analytics

available with Mobile Big Data do a reasonable job of decoding a garbled stream

of free range type data to identify nuggets (consumer sentiment, competitor

negatives comparisons, etc.) of true market intelligence. Social media data is

another distinct data type that can be included in business intelligence through

using specialized Mobile Big Data Advanced Analytics tools.

To improve decision quality often requires providing users with cross referenced

“combo data“ intelligence derived from multiple data types and independent

Page 13: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 13

sources. For example, complete and actionable provisioned intelligence could be

derived from customer demographic files, competitor pricing and market trending,

transaction history, sensor driven data (temperature, etc.) and public sourced

information (interest rate fluctuations, etc.). Decision makers often require the

convergence, or at least general agreement, of data from multiple sources in

order to feel comfortable making complex decisions. Mobile Big Data is oriented

to processing combined and cross referenced data types from multiple and

sources to generate intelligence a fully contextual manner (inclusive of associated

time references, etc.).

Customer Facing: Living at the Edge

To further accelerate business processes of all types, enterprises have increasingly “

pushed “ key decision making out to the edges of their organization. “Edge

workers “now include many types of remote location employees, field sales

persons, work from home staff, off-site call centers, etc. Many edge workers are

customer facing, and thus required to make on the spot decisions that may be

vital to generating or retaining customers, preserving enterprise assets, etc.

Customer facing service concerns and challenges are important for every

enterprise; the difference comes in how this activity is handled. A competitive

business advantage may come by empowering the customer facing employee to

be able to say “ here’s specifically what we will do right now to resolve this matter

“ as opposed to “ let me check on this matter and get back with you about what

we may be able to do ”. The typically HQ based marketing, product

development, etc. staff also require decision support intelligence but historically

are more likely to have situation targeted support than the edge workers.

Customer facing edge workers obviously need intelligence provided in a rapid

but, just as important, in an easy to understand and actionable manner. Edge

workers are already familiar with mobile platforms (ubiquitous smart phones and

tablets) so these devices are an excellent means to deliver intelligence. Figure 3.3

displays where today’s knowledge workers are typically located.

Page 14: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 14

Figure 3.3: Typical Knowledge Worker Location

Source: Several Recent Surveys

As can be noted in Figure 3.3 more than half of today’s knowledge workers are

regularly making their business decisions outside an organization facility, employee

mobility is a fact. With regard to the Mobile Big Data associated cost of providing

edge worker support, most enterprises should note that major elements of the

required infrastructure are very likely already in place. The needed mobile Wi-Fi

connectivity is fairly common and in many cases an available Big Data capacity

of sufficient size to get started exists. The required additional elements likely consist

mostly of interfaces and activity specific applications that may be sufficiently

inexpensive to get started with an initial Proof of Concept (POC) or pilot phase.

The POC phase should then provide feedback sufficient to better define and

evaluate the ROI for moving to Mobile Big Data based edge worker support.

B. MINIMIZING DECISION LATENCY, SPEEDING TIME TO VALUE

In dynamic and complex in environments the ability to immediately make a

decision (real time or near-real time) may be highly important. In many cases a

very accurate decision, but one made after the opportunity has expired, can

have little to no value. There are numerous examples where a late decision is in

effect, no decision at all. In addition to avoiding a lost opportunity, accelerating

the decision cycle can reduce the “time to value “(often expressed as when it can

be booked), typically the sooner the better. The blindingly fast intelligence delivery

of Mobile Big Data (close to instantaneous) can be critical to its real world value

proposition.

Page 15: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 15

The Need for Speed and the Race to Zero

Largely as a result of the Great Recession enterprises of all types found that in order

to be competitive they required much faster decision cycles for issues like pricing

determination, spot promotion definitions, etc. The mantra of reduced time to

market now has traction well beyond the marketing department and affects all

areas of business. Product development cycles may now be halved and customer

facing employee responses have become close to real time. In this context any

enterprise action underlying the timing of business decisions, including intelligence

development, must be accelerated or opportunities will be lost to a more agile

competitor.

The “Race to Zero “is another time to value related concept and term. It basically

states a desire by equity/commodity traders to reduce the latency of their trades

to zero. Already measured in nanoseconds this trading latency reduction has been

the subject of numerous projects and major Big Data investments. Certainly, the

process of actually deciding to make the trade (often based on near real time

situational analysis) is also a valuable element of the overall trade process cycle

and another area for acceleration. Time to Value is a common concern for all

types of financially valued activities and the overall trade cycle is another key

target for near real time intelligence provisioning with Mobile Big Data.

As an emphasis of the need for improving the timeliness of current data analysis,

and thus improving the “ freshness “ provisioned business intelligence, Figure 3.4

displays current business intelligence analysis cycles.

Figure 3.4: Current Business Intelligence Refresh Cycle

Source: Several Recent Surveys

Page 16: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 16

Sure, real time analytics (less than 5%) can be costly but it’s fairly incredulous that

anyone could consider a monthly refreshed (35% incidence) analysis as

actionable business intelligence. Perhaps these analyses are status reports, but

even in long cycle manufacturing and construction all such information is strictly a

historical record by that point. It’s a small wonder the majority of today’s

knowledge workers use their own resources and informal means (smart phones,

internet access, spreadsheets, etc.) to supplement or replace the stale business

intelligence they now receive.

Real Time is Money: Halting Fraud and Enabling Mobile Marketing

In many real world situations, fraud being a good example, to be effective

decisions must be made while the actual transaction is in process. In a variety of

credit card and other financial transaction related processes (credit application

processing, claims adjusting, etc.) semi-automation is common, but such systems

are frequently too slow to be of assistance while the transaction is in process or are

based on limited data sources and ill-defined criteria (location zip code,

transaction total amount etc.) likely to be ineffective. With Mobile Big Data based

connectivity, specialized analytics and near real time delivery, it’s possible to

provide in-process transaction specific intelligence (tailored to the specific

situation and cross referenced from multiple data sources) for well informed

decision making, again while the transaction is in process. This Mobile Big Data

ability is especially valuable for halting in-process fraudulent transactions to

minimize the associated losses.

Conversely, there are transactional circumstances where real time intelligence

may be valuable not for the purpose of halting in-process fraud but rather for the

purpose of improving sales revenues. These upside transaction based situations are

often available to customer facing edge employees (call center staff, field sales

account reps., etc.) were the situation frequently includes temporary time-

dependent opportunities, either close the deal now or it’s gone for good. In such

situations Mobile Big Data can provide real time knowledgeable guidance ( based

on individualized intelligence from past preferences, usage behaviors, click

through history, market trending, etc. ) in-process during help desk calls, pricing

quote determinations , product feature and availability queries, etc. The ability of

the employee to knowledgeably offer a personally tailored promotion, a unique

“one-off “ pricing quote, etc. can improve close rates and increase business

volume, all with existing staff. An ROI based on increasing business revenues may

even be better (for its potential future impact) than one based entirely on cost

reduction. The referenced real world success examples include the Mobile Big

Data enablement of real time cross selling and mobile marketing.

Page 17: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 17

Democratizing Intelligence

Post great recession enterprises have pushed decision making out to the

organizational edges and significantly flattened themselves, reducing

organizational layers to get “lean and mean”. In this context it’s appropriate to

strongly consider democratizing business intelligence. This intelligence

democratization includes flattening and speeding all intelligence associated

processes and especially provisioning intelligence directly to first level

management. In many organizations the effectiveness of first level management is

critical to the organization hitting targeted goals. First level management

(especially on the customer facing edge) often has the highest customer impact

with a major opportunity to gain or retain business. Continuing to provision

intelligence to the organization’s senior management and marketing teams, often

within the corporate headquarters monolith, is as important as ever, but the SAME

type intelligence provisioned to them should also go to first level management at

the SAME time. Traditionally business intelligence may be initially routed through a

delay ridden management review (and unfortunately filtering) process prior to

being released. The associated adjustments and slowed delivery can make the

intelligence far less actionable where it may count the most.

The screening, filtering and senior management approval associated with business

intelligence should occur during the analytics algorithm and recommended

response definition stages. The experienced intelligence analysts should be

thoroughly instructed, given the analytics tools needed to do their job properly

and then trusted to produce and distribute actionable intelligence. Once the key

experience based business rules and alert/action definitions are in place and

working well, the same resultant intelligence should go directly to all appropriate

persons, at all levels of the organization, without further review and “ aging “.

Unfortunately the old saw that “knowledge is power“ is sometimes used by the HQ

staff as leverage over the field staff in an unproductive philosophy of “share

nothing “or share only bits of data. For maximum effectiveness, and to provide the

best ROI, the business intelligence produced by Mobile Big Data (without filtering

and adjustment) should be shared equally and shared ASAP. The powerful

processing, analytics and delivery capabilities of Mobile Big Data should be

applied to (1) develop the best business intelligence possible and (2) the

democratize this intelligence through immediate delivery, thus providing a true

business advantage.

C. LEVERAGING MARKET TRENDS AND USER PREFERENCES

In addition to the functional capability reasons to deploy Mobile Big Data, several

current market trends and general user preferences also contribute to its potential

Page 18: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 18

as a true business advantage. Instead of ignoring, or working around, these

external environmental issues, Mobile Big Data takes full advantage of them to

help gather data, lower deployment costs and improve user acceptance. For

example, per recent IDC reports, almost a billion smart phones were sold during

2013 and their growth is expected to continue with at least a 15% CAGR. Given this

general environment it’s totally logical for Mobile Big Data to ultilize smart phones

(and tablets) to deliver near real time business intelligence to knowledge workers.

This current market trend and several others encourage and facilitate the cost

effective deployment of Mobile Big Data.

Petabyte is the New Terrabyte and Gigaflop Blow-Out

Almost without exception, the huge amount of data that enterprises regularly store

and process for all purposes, including business intelligence, would have been

unimaginable only a few years ago. For years annual IT budgets have included

major CAPX increases just to accommodate these volumes and the associated

capacity for ever expanding storage, processing and communication. The

increased demand now means that for many enterprises the required data

storage capacity is increasingly being described in petabytes (Note; 1 PB = 1000

TB) with required processing capacity edging ever closer to petaflops (Note: 1 PF =

1000 TF). The huge capacity related costs have been a backbreaker for many IT

shops and a practical gating factor for advanced information system deployment

of all types, including business intelligence.

Fortunately to address this serious situation Moore’s Law (i.e. due largely to ever

greater chip densities, and other associated ongoing technical improvements, the

cost per unit of capacity will halve about every two years) has come into play and

has helped considerably. There has lately almost been a blow-out sale for

gigaflops (1000 floating point processor transactions per second), with associated

storage unit cost drops, and this combined with increasingly efficient chip level

hardware architectures and improved software have greatly helped the situation.

Figure 3.5 illustrates the ever lower cost gigaflop.

Page 19: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 19

Figure 3.5: GIGAFLOP Cost per Unit

Source: Wikipedia

This cost per unit lowering process is forecast to continue with the future

implementation of advanced processor designs combined with the potential

move to non-silicon based technologies. As with other processing intense activities

these continued cost per unit reductions are critical to future Mobile Big Data

upgrades and in many cases the issue will come down to “ how much is

affordable? “Rather than “ can we do it?”

In addition to cost concerns is the issue that the significant processing requirements

of performing business intelligence analytics can’t take precedence over an IT

operation’s ability to perform its other responsibilities, namely keeping the lights on

(i.e. accounting, supply chain, etc.). A general fact of life is that there will always

be a shortage of available IT resources; further emphasizing the importance of a

metrics based ROI approach for all Mobile Big Data deployments. Of note with

regard to IT resources issues are (1) increasingly the business intelligence functions

may NOT report through the IT organization but rather through functional user

group management who are closer to its value proposition and (2) cloud services

are increasingly available to at least partially offset potentially disruptive

processing loads by handling Mobile Big Data associated processing off site. Of

course, with a Cloud supported approach the source data (which can be

extremely proprietary) and the resultant business intelligence (which can be highly

confidential) are not in total control and thus become subject to significant

security and reliability concerns.

Page 20: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 20

Crowd sourced Data: Likes, Tweets and Twerks

When considering anything to do with consumer related business intelligence it’s

important to consider the following from USA Today.

- 96% of the U.S. internet connected population access social media.

- Mobile devices now outnumber the earth’s human population.

- Facebook has over 200 billion friends and does 6 billion likes/day.

- Facebook has processed over 8 trillion messages.

If those statistics don’t get your attention, nothing will. And yet it has been difficult

and slow for most enterprises to obtain much value from this virtual landslide of

consumer information. Sure, regular market trend summaries are available for

purchase and subscriptions to general analysis services are available, but cutting

edge business usage of social media data types for business intelligence still isn’t

common. Capabilities intrinsic to Mobile Big Data may be the key to facilitating

the practical business usage of social media derived intelligence.

As previously reviewed most social media data is only semi-structured and quite

often mostly text based. Admittedly the meaningful analysis of text information is

perhaps more an art form than a science and certainly best addressed by

experienced analysts. While key word and phrase identification is fairly simple,

unless care is taken to retain the context of such information any business

importance may be lost. Without the proper context mere reference occurrences

and count numbers may be misleading. To retain the context of social media data

can unfortunately entail retaining relatively large amounts of support data.

However, it’s definitely possible to produce meaningful social media sourced

intelligence (consumer sentiment trends, competitor trending comparisons, etc.)

with specialized analytics tools.

Practical Mobile Big Data enterprise usage of social media derived intelligence

has included the quick identification of emerging customer service/relationship

issues and proactively initiating responsive actions to address them. Some

organizations have used social media based information to trigger quality actions

and pricing adjustment decisions without waiting on direct customer feedback.

Other enterprises have gone to the extent of immediately sending a responsive

reply to customers expressing problems or concerns (hopefully, not in an

anonymous or derogatory manner). In some cases an attached file or address is

included to further explain correct product usage or appropriate service options.

In effect, the enterprise uses social media to take proactive actions in an attempt

to make real time customer situation improvements. Please recall that Mobile Big

Page 21: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 21

Data can be used to communicate directly with customers should such action be

warranted by recalls, alerts, etc.

BYOE Phenomenon

Please also consider the below phenomenon when reviewing market trends

having a significant effect on many enterprises.

- 2/3 of U.S. citizens own a smart phone and 1/3 own a tablet.

- 60% of U.S. employees now use a BYOD platform for business.

The term Bring Your Own Everything (BYOE) originated with employees being

allowed to Bring Your Own Device (BYOD), typically smart phones or tablets, to use

in their business related activities – the real world situation has now gone way

beyond that point. It seems most employees prefer to have their own familiar easy

to use personal device and don’t much care if they, rather than the employer,

pay for it. They like doing this so much so that the Yankee group found that 60% of

U.S. employees currently use BYOD platforms including a significant number of folks

doing so covertly because their employer specifically prohibits BYOD (typically for

security reasons). At a surface level BYOD seems fairly logical and is quickly

becoming the rule rather than the exception. More problematical is the BYOD

extension to BYOE where employees are increasingly using their favorite smart

phone applications for business and tapping into whatever information sources

they feel like to obtain business related information. Other than monitoring

employees closely (through Wi-Fi, etc.) and enforcing penalties for employees

caught doing BYOE there’s not much an employer can do. Yet employers are

typically considered responsible for the actions of their employees (especially

when performing job related duties) – when this BYOE situation causes security

breaches, incorrect actions, etc. the resultant liability and potential customer

problems will fall on the employer.

In many cases, BYOE is occurring because employees find it easier, and possibly

more timely, to go around their enterprise provided systems, where they even exist.

The best BYOE response may therefore reside in doing a better job of providing the

employee with true intelligence they can’t readily obtain elsewhere, in a timely

and easy to use manner. To do this means enterprise support systems require well

developed (not merely converted) applications that operate uniformly across the

various disparate BYOD Mobile platforms that are likely to be encountered. Once

the enterprise provides such applications experience has shown users will be quite

willing to use well targeted mobile applications. By incorporating providing mobile

BYOD enterprise applications through the overall Mobile Big Data system these

become a key element of effective end to end user intelligence support.

Page 22: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 22

D. “WHY” MOBILE BIG DATA TAKEAWAYS

The primary takeaway points from the WHY Mobile Big Data section include;

1. Over half of today’s employees are knowledge workers regularly making

business decisions and more than half of them don’t believe they receive

adequate decision support intelligence.

2. Many knowledge workers are customer facing edge employees regularly

making key decisions and more than half require a high degree of mobility.

3. About 60% of knowledge workers use their BYOD platforms (even where

prohibited), often using ad hoc applications to locate external data and

generate their own business intelligence.

4. Time to Decision and Time to Value are often synonymous so accelerating

all aspects of the decision making cycle is essential and doing so can have

a measurably value. The ROI metrics for enabling better quality and faster

decision making may include significant revenue increases as well as cost

reductions.

5. The knowledge intensity of business intelligence is directly dependent on

using the best available input data. Solid business intelligence typically

derives from multiple data sources with multiple data types and may

potentially include social media data and IoT liquid data input.

6. The analysis of social media sourced data and streaming IoT liquid data can

be challenging but is practical through the use of Advanced Analytics.

7. Mobile Big Data is a solid business advantage with the objectives of (1)

minimizing decision latency and (2) enabling better quality decisions.

Page 23: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 23

IV. WHAT ARE THE CORE CAPABILITIES OF MOBILE BIG DATA?

Because Mobile Big Data combines three largely distinct elements (Mobile

platforms, Big Data and Advanced Analytics) it’s appropriate to review each

component separately and in the context of their convergence. The Mobile

platform element provides intuitive and familiar user interfaces to deliver business

specific intelligence and applications. The Big Data element is directed at

efficiently performing large volume data handling and processing. The Advanced

Analytics components perform complex analyses on various data types to discern

significant patterns and cross-referenced business intelligence. The combined

Mobile Big Data system delivers on its dual goal of (1) significantly improving the

quality of complex business decisions and (2) reducing the delay, or latency, of the

decision making process.

A. INDUSTRIAL STRENGTH PROCESSING OF BIG DATA

Obviously, to handle the huge data files appropriate to Mobile Big Data and

perform almost real time complex analysis requires a major processing throughput

capacity. The required capacity is available in a reasonably economic manner

with a combination of Big Data derived technologies including large memory

footprint platforms, MPP, etc. Figure 4.1 displays the amount of “analytics only”

data (internally sourced data) that business analysts typically expect to be

regularly required for their organization in 2014.

Figure 4.1 : Estimated 2014 “Analytics Only Required Data” Volume

Source: Several Recent Surveys

Page 24: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 24

This volume of data handling and processing strictly in support of business

intelligence related analytics activities represents a significant requirement and

these volumes are anticipated to continue their rapid growth.

MPP, Cloud Option and HAL Jr.

Gone are the “Big Iron “processor days and in its place Massive Parallel Processing

(MPP) has become the future path of most processing intensive environments.

Where in the past huge transaction loads (banking, insurance, etc) drove large

scale processing requirements, now the liquid data stream handling and analytics,

all done in near real time, is a key factor for the move to MPP. MPP essentially

provides a huge collective throughput capacity derived from a massive array

(potentially several hundred) smaller and commodity type processor platforms. In

this manner each platform handles a portion of the processing with their resulting

output then being collected and coordinated into the final results. Efficiently

coordinating and managing these parallel processor arrays is an ever improving

technical process but when scaled they can provide a net processing throughput

definable in petaflops. Some specialized processor platforms can now quite

practically provide capacities approaching 2.5 petaflops. The processor

throughput cost per unit generally follows Moore’s Law but because of their size

MPP based systems can come with a hefty price tag. Mobile Big Data today, and

even more so in the future as it continues to ramp its analysis volumes, frequently

depends heavily on the MPP processors to perform complex analysis of large data

files in a timely and cost effective manner. These significant MPP costs again

emphasize the need for Mobile Big Data to specifically target essential business

requirements, rather than technical capability, as its deployment rationale.

A key concern inherent to the MPP approach is a significant potential for individual

processor platform failure. By using a large array of smaller and cheaper processor

platforms it is to be expected that the failure of one or more of these commodity

processor units may be a relatively “common “occurrence. To address this issue,

and eliminate the potential of a catastrophic single point of failure, a set of

advanced processor array operating systems and non-hierarchical component

architectures are utilized. Along these same lines data storage, and application

execution, is done in a redundant parallel manner. Processing overhead could

obviously be an issue with such complex synchronization and control methods but

near native instruction sets allow for net array efficiencies.

Of course Mobile Big Data can be initiated, and scaled in a limited manner on

existing non-MPP enterprise processing platforms. The concern here is that the

other tasks, essentially keeping the enterprise lights on, can’t suffer from the major

additional processing load required to support advanced data analytics, etc. In

many ways the Mobile Big Data processing decisions again come to a very

Page 25: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 25

practical issue of “how much can you afford ?”. In the future different processor

types (non-silicon) will continue to make multi-petaflop level processing ever more

affordable.

Also coming is improved machine learning or artificial intelligence, almost to the

level of the HAL 6000 processor of Stanley Kubrick’s “2001, A Space Odyssey “.

Today an almost HAL Junior type processor platform exists in IBM’s Watson, a huge

capacity MPP system, which is designed to proactively learn from its mistakes and

to minimize or totally negate structured programming instruction of any type.

Certainly an AI based learning machine with advanced self-correcting capabilities

could be of key long term importance to the continued evolution of complex

event Mobile Big Data.

Speed Dating Data: Hadoop/MapReduce

Somewhat analogous to the rapid fire speed dating process of learning a lot

during a short time period, the Big Data analysis of truly huge data files had to be

accelerated. It also quickly became apparent, even with high throughput MPP

platforms, that a more efficient data handling methodology was required for

unstructured data types. As a result Hadoop and Map Reduce were developed

initially to improve the efficiencies of data handling and processing for addressing

Google’s huge data bases. These tools were later generalized and made

available as open source software to provide very specialized environments. As

such they are important potential elements of Mobile Big Data.

With Apache Hadoop (the open source variant) as a data file is being entered

into the system it’s fed through a server that breaks it into user definable blocks,

replicates it and feeds it to 3 separate storage locations (ideally on 3 separate

processors) for batch analysis. These multiple data block locations are indexed

and records kept as to the correct sequence, etc. When it comes time to act on

the data blocks the application is also segmented (according to specifically what

is being performed) with replicated instruction elements being sent to the server

platforms containing the appropriate data. The various processing platform

processing results are then verified, cross-checked and may be summarized,

reprocessed again, etc. according to the specific requirements. This segmentation

of the data stream and the programming functions takes good advantage of the

MPP array architecture to drive throughput and minimize the impact of device

failure. In this manner the failure of a specific physical device is immediately

detected and there is no single point of failure for the overall processing. Due to

the fast processing speeds and the device specific instruction set the overhead

involved with Hadoop becomes less a factor and the array’s net throughput is

immense.

Page 26: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 26

Apache Hadoop can run on Linux or M/S platforms and has been available since

2005 but due largely to the nature of its open source license (without supplier

support, contracted upgrades and enhancements, dynamic security patches,

etc.) Apache Hadoop has been somewhat slow to achieve widespread enterprise

acceptance. To better address typical enterprise system needs the overall

Hadoop concepts ( major design elements, etc. ) have spawned a number of

proprietary varietal offerings, add-ons and lookalikes, all suitably “ hardened “ to

better meet commercial requirements. Whatever specific variant is used, the

Hadoop data handling concepts are an important MPP enabler for Mobile Big

Data. Its not perfect but for the right batch processing environment it may be the

best alternative.

MapReduce is a Hadoop paired operating environment and, when dealing with

distributed file systems (HDFS, etc), further reduces network traffic to significantly

improve overall throughputs. The Map portion optimally filters and sorts data being

entered and the Reduce element efficiently collects the various sub problem

results and defined final processing results. These MapReduce capabilities

automatically optimize the assigned location and processing sequence to speed

the overall processing. Although such improvements are measured in

nanoseconds, with the volumes involved this may be the only practical approach.

As an example MapReduce has been paired with the Apache Hive by Amazon

(using an Elastic MapReduce variant) on a more than 10,000 Linux processor array

and Facebook uses a Hadoop based system to operate on it’s massive 100+ PB

database. Other specialized and valuable Big Data related tools exist (SAP HANA,

Cloudera, NoSQL, etc.) but Hadoop, or a variant, can be key for the highly

scalable processing environment essential for certain varieties of business

intelligence analysis.

Page 27: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 27

Figure 4.2: Analyst’s Preferred Analytics Processing Platform(s)

Source: Several Recent Surveys

As shown in Figure 4.2 analytics users prefer their analytics processing be

performed in various locations with a growing number preferring that Hadoop be

included. The group indicating a preference for a Hadoop based analysis

platform typically prefer this approach for effectively handling disparate data

types. Also note that many of today’s intelligence analysts consider having

multiple analytics processing platforms to be acceptable approach.

Dark Data, Active Archive and High Density Processing

Most enterprises now have a vast amount of available data located at various

points of the organization and kept on a variety of storage systems. The issue is that

storing all this data is on Tier1 devices can be cost prohibitive and may lead to

totally off line storage (inaccessible for short-term requirements). Unfortunately this

valuable data resource then becomes “dark data” that for most purposes is dead.

The other concern is that of the Tier1 stored data today, approximately 80% hasn’t

been accessed in the last six months.Without an effective means to use Tier 2

(secondary disk based) or Tier 3 (typically active tape) the ever growing data

storage investment CAPX issues quickly lead to ever increasing dark data. The Big

Data answer has increasingly become active archive.

With an active archive approach data is actively managed and assigned to the

most appropriate storage type and location. Interestingly, none of the active

archive managed locations are truly off line, although some clearly have a longer

recovery cycle, so nothing ever goes dark and thus becomes of no value.

Preserving the value of data may be the 4th V of data.

Another characteristic of major data volumes is that in order to process them, for

analytics or whatever, they may have to be segmented due to processor platform

data size limitations. This data segmentation can be highly inefficient by requiring

multiple combinational processing, etc. The Big Data response is through high

density processing (actually high data density) platforms. Such processing

platforms now approach 100 TB of data for single pass under one operating

system, true “first time final” type processing. In effect this high density approach

enables valuable application consolidation that can greatly facilitate big load

analytics (huge image files, etc.) A good example of the value of these advanced

Big Data capabilities was the recent case where the enterprise was able able to

reduce a major 200 hour processing job to about 20 minutes, indeed a clear

business advantage.

Fabric-Like InfiniBand Topology

Page 28: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 28

The linkage or interconnection methodogy of processors is another focus for Big

Data throughput improvement. To address speeding these linkages InfiniBand was

developed to better support nanosecond cycle processing with active congestion

management and has resulted in further latency reduction and reliability

improvements. There are alternative connectivity methods, but InfiniBand is the

connectivity approach of choice for petaflop level processors when reducing

overall latency. InfiniBand logically allows no single point of failure while eliminating

the typical mesh based device connectivity latency (again measured in

nanoseconds) associated with their typically complex coordination and control.

InfiniBand methodology also reduces potential failure recovery impact on

throughput. InfiniBand design provides a chip level dynamic single pathing

approach for an intelligent fabric-like device interconnection. With this logical

(rather than physical) multiple interconnection pathway fabric, each logical path

is continually optimized for throughput and the net reliability of guaranteed

delivery is possible, a serious consideration for MPP based processing.

B. MOBILE BASED INTERFACES, ACCESS AND APPLICATIONS

The vital user interface of Mobile Big Data comes through the usage of today’s all

pervasive mobile devices, Wi-Fi connected smart phones and tablets. To be

effective, especially considering the time critical nature of many supported

decisions, these devices need to be running purpose built mobile applications.

Clearly the implied display limitations are critical as is the issue of having the mobile

applications work and operate in a similar manner regardless of the specific

mobile platform in use. In addition mobile devices themselves are also continuing

to grow and evolve with the addition of wearable watch based units and glasses

type display and data collection.

User Mobility with Preferred Platform

From numerous recent surveys it’s been shown decision making employees,

knowledge workers, now greatly value the mobility provided to them through their

smart phones and tablets above all other types of technology. Employees no

longer consider it appropriate to be required to utilize “standardized “devices and

to do so in a controlled environment. They want, and are determined to have, the

mobility to work from home, a Starbucks or any other site that’s convenient.

Supporting the often informal technical environment (frequently with minimal

security) has become a major enterprise concern.

Key among the employees specific mobility supported capabilities is their having a

direct, and easy to use, access to decision support information (including

proprietary pricing metrics, etc.) through their totally familiar personal BYOE

platform. They’re willing to use employer provided applications and information

Page 29: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 29

just so long as this remains as user friendly and intuitive as commercially available

alternatives (most with trivial pricing). The message here being an employer must

either offer its employees (especially edge workers) full mobility , with easy access

to enterprise provided business intelligence, or they will use their own BYOE

platforms and applications to access whatever data they feel appropriate.

Provisioning Actionable Intelligence

Current mobile platforms provide high resolution, but physically size limited, displays

and fairly HD image capture with audio. Also available are image projection

capabilities and logical keyboard input methodologies. Key for employee

acceptance and usage is developing applications specifically designed to work

well on such devices with their size and operational limitations. Again as a result

the mobile platform applications provided to employees must be purpose built

and specific to mobile platform use. Truncated standard device applications

(desktop, laptop, etc.) won’t compare well to commercially available apps and

will be a point of contention.

Practical guidelines for developing effective mobile apps include;

- Limit display information but provide numerous drill down features

- Design for touch screen user interface with user familiar icons, etc.

- Use display color and audio to emphasize alerts and/or key facts

- Avoid detailed small diagrams and dark on dark or light on light displays

- Minimize streaming volumes to speed downloads

- Require minimal user input with easy abbreviations, etc. limit keyed data

- Provide pop ups for key contextual information

- Include available reference links for follow-up (esp. with large files)

- Provide a “master long-term retention “option and offer a print option

- Always include proprietary data instructions and security warnings

- Include likely FAQ help instructions for any potentially confusing

functionality

- Don’t be dramatic but don’t be gameslike or cartoonlike

- Be as conversational as possible, using familiar slang, abbreviations, etc.

Page 30: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 30

The lessons learned from years of effective mobile app. development all serve to

make Mobile Big Data delivery processes highly intuitive and extremely user

friendly. Figure 4.3 displays the preferences of active users of business intelligence

as to what type decisions they believe are best supported by business intelligence

provisioning.

Figure 4.3: Business Decision Types Likely to Benefit from Improved Intelligence

Provisioning

Source: The Economist

From Figure 4.3 it’s obvious that most current users believe a number of different

type business decisions would benefit from improved intelligence provisioning.

Foremost among these opportunities are key business issues dealing with

identifying new markets, retaining current customers and responding to

competitive pressures.

Mobile Data Capture and Platform Transparency

In addition to mobile platform provided user interfaces they can also collect and

input some limited forms of data. This mobile data can include user activity logs,

but also extends to video and audio streams and spatial location specific

information. Retailers have already conducted tests with mobile spatial data to

determine user apparent interests (what display did the user pause to look at and

for how long, etc.) and a variety of independent subscription services can provide

mobile usage data (what books were read and when, etc.). The image and video

capture capabilities of smart phones are regularly being used for business related

purposes, including accident and adjusting records, in-store display set up, etc. In

addition, although the various internal features of mobile platforms are not

uniform, those equipped with accelerometers can deliver data on driving patterns

66%

55%

46% 41%

35%

29%

29%

26%

19%

New Market

OpportunityCustomer Satisfaction

Product and Service

ChangesCompetior Evaluation

Financial Review

CAPX Investment

Risk Analysis

Market Segmentation

Fraud Prevention

Page 31: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 31

(tendency to weave in traffic, frequency of sudden stops or rapid acceleration,

etc.) that may have significance for consumer needs or even law enforcement.

The use of mobile platforms for data collection continues to develop with the

continued enhancement of wearable devices.

In the process of providing easy to use business intelligence another key item must

be accounted for, the various non-standard, and quite specific, product

differences likely to be encountered in a BYOD device landscape. The mobile

applications must be able to look and work the same on an Android based mobile

device as they do on an iO/S platform. To address this cross-platform application

issue a variety of established utilities (including those provided by Fusion Labs) exist

to assist with development. Again, this potentially troublesome technical issue has

already been addressed by the established mobile platform support resources and

is thus available to help speed Mobile Big Data deployment.

C. ADVANCED ANALYTICS DERIVED INTELLIGENCE

Fairly advanced analytics in one form or another have been with us for some time,

the differential with Mobile Big Data is the speed at which analytics have to work

efficiently and the irregularity of the data being analyzed. Surely, given sufficient

time and attention analytics have long been able to discern the hidden patterns

and subtle meaning in even the most motley of data, the trick with Mobile Big

Data is doing this on the fly in nanoseconds. Liquid data streams can vary

continuously, even in never seen before, or ever to be seen again, manner without

a significant pattern. Determining what represents actionable information that

needs to be immediately delivered to users on their mobile devices is the

challenge.

Strategic Analytics and Analytics in Motion

Strategic Analytics largely include statistically based data analysis, primarily

analyzing existing static data though packaged routines, has been available for

decades and is accurate to determine the statistical significance of data

variances, identify variable correlations, identify change inflection points, etc.

That’s the good news, the not so good news is that these strategic long standing

statistical methodologies work better with well understood and structured input

data and that with large files they can be computationally intensive (multiple non-

linear regression analysis of a huge file, etc.). They wouldn’t be the analysis

analytics approach of choice for working on IoT streaming data or textual analysis

but can be effective for reviewing existing internal data files to identify significant

business intelligence. Statistically based modeling and analysis tools can also be

used for predictive trend definition and are the basis for stochastic models of

automated decision tools (including providing confidence levels). What they

Page 32: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 32

don’t do really well is discern the subtle significances of random appearing

consumer information typical to social media data streams. Statistical analytic

tools are available off the shelf and can thus provide economic analyses of MDB

structured data. For the purpose of contributing to actionable business

intelligence, the statistical analysis results should be taken out of their typical

statistical jargon (large correlation coefficient, etc.) and put into terms (high

confidence trend, etc.) better understood by decision makers.

Analytics in Motion typically refers to the best analytics for understanding the data

types more typical to IoT data streams and social media sources. To discern

significant business intelligence from text a combing type review is done based

primarily on the ability of the analytic tool being used to (1) identify critical words

or strings of words and (2) define the likely “ sentiment “ being expressed in the

tweets and (3) tally these according to predefined query limitations. For practical

reasons the input data being analyzed is typically defined to focus on most likely

sources of needed intelligence (the tweets from a geographic area for a specific

period of time instead of the whole U.S. for a year). This type of sentiment analysis

can be effective for quickly spotting customer problems with products,

understanding the general public perceptions from institutional advertising, etc.

Frequently consumer sentiment trending verses that of direct competitors is used to

evaluate advertising campaigns, pricing and promotions, etc. Where impractically

large scale background is desirable, a statistical sampling technique may be

applied.

The strategic and in motion analytics can be visualized as being centered on

different type data (static, liquid, text, etc.) so it’s critical to be able to specify

what data is to be pulled in, what query or analytics tools are to be applied and

what the blended output intelligence is to consist of. Recall the analytics are

focused on identifying early-on consumer preferences, non-random sensor

deviations, purposely disguised programmed trading alert algorithms, etc. In many

cases the results are presented to users with general text statements, graphical

plots, etc., care must be taken however not to imply a degree of confidence that

doesn’t actually exist. Certainly, the analytics derived results represent valid

business intelligence but the decision makers should always fully understand the

intelligence, and its reliability, and use their best judgment prior to acting.

Complex Event Processing

Complex events processing could almost be viewed as a hyper drive type real

time IoT, liquid data sampling and monitoring. Complex Event Processing (CEP)

analytics typically extend the input data monitoring into real time (or near real

time) as a result of the importance of the situation being monitored. CEP

simultaneously monitors several data inputs (often including liquid data) to

Page 33: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 33

continuously verify they individually and collectively are within predefined

boundaries. It may well be that each variable is within an acceptable range but

that a certain combination or grouping of the monitored variables is not within an

acceptable range. The complex event situation being monitored for is typically

made up of a fairly complicated layering of contributing smaller events and

activities, often with each occurring in very rapid fire manner. Quickly recognizing

the compounding of a complex event and properly responding in a timely

manner can be a major challenge. However, specialized analytics tools can help

to recognize and address these serious and time urgent situations. Complex event

analytics typically include;

- Situational experience based monitoring and evaluation rules

- Multiple live data feeds of potential CE situation contributing factors

- Automatically adjusted data stream sampling and filtering may be used to

focus on key variables

- Event correlation analysis used to evaluate significance of developing

situation

- Includes alerts as to unexpected trending and “ out of bounds “ situations

- Full real time in-memory processing may be required

- Algorithm derived complex event responses are predefined

- Alerts and recommended response actions are immediately transmitted

- Feedback for potential rules adjustments is provided for permanent variable

shifts

Complex event processing analytics are obviously a resource intensive approach

but are justified where there’s a requirement to respond to time urgent and

continually changing environments having a high value. Typified in this type

solution set deployment are financial market situations, critical process

environments (rocket fuel blending comes to mind) or life and death medical

situations. CEP may indeed be a part of overall business intelligence, but is often

handled separately by dedicated specialists.

Soft Science: Market Listening and Data Visualization

Market Listening refers to the use of a regular ongoing monitoring of market

sentiments and preferences. As described, social media listening can be done as

a part of Mobile Big Data analytics through text based tools that discern the

sentiment of customers or prospects and are also able retain the apparent

Page 34: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 34

context. This type market intelligence may also come from a variety of subscription

services but in that situation the same business intelligence is likely to be available

to others, including key competitors. The intelligence from regular market listening

is primarily intended to provide early on alerts for market shifts or newly developing

trends. For example such intelligence could be triggered by warranty claim

related tweets that would result in the system checking recent field service logs, a

review of related in-process engineering changes, etc. In addition to external,

social media based, market listening, some enterprises “listen” to internal text data

for alerts and may include text from field reports submitted by their edge worker

employees. As an example a major insurance firm automatically reviews recent

field adjuster reports and may use this intelligence for account negotiating. For

cases where social media trending can be key, such as fashion dependent retail,

relevant portions of social media data (RSS feeds filtered for color comments, etc.)

may be appropriate for inclusion. The point being that Mobile Big Data is ideally

suited to provide employees making complex decisions with the timely support

intelligence they need as derived from multiple sources and disparate data types.

Data Visualization is another seemingly soft science pattern identification analytics

method where apparently random data points are displayed or printed (typically

on a colorful 2 axis plot or other highly graphic format) and then manually

examined to identity patterns leading to potential intelligence. Although quite

basic, this type analytics approach can be very helpful as a first pass to define

following,and more rigorous, analysis. The data visualization process obviously

requires trained and experienced analysts in order to be highly effective. Data

visualization can, however, be quite effective in explaining the basis for derived

intelliegence and once properly identified and verified these data patterns can

be quite striking.

Self-Serve Analytics

In many cases the effective use of advanced analytics is not a common skill of the

recipients of business intelligence. While the users are often capable of making

normal queries, etc. fully understanding the usage and output of complex

analytics is likely best left to experienced analysts. In situations where intelligence

users are highly skilled there exists a variety of packaged off the shelf analytics

tools that can be used from remote locations. This may be especially important to

those enterprises where the business intelligence function itself reports to

operational management and/or is being outsourced to cloud based resources

with minimal involvement of IT support. Although self-serve analytics are frequently

promoted, it’s doubtful this is appropriate to highly complex environments or those

with extensive edge worker populations. It’s a trend to watch but should be

approached with caution. Not having knowledgeable and experienced analysts

Page 35: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 35

involved in the business intelligence process may be false economy resulting in

costly mistakes and a loss of confidence by the users.

D. “WHAT” IS BIG DATA TAKEAWAYS

“What” are the Mobile Big Data capabilities section takeaways include;

1. Mobile Big Data is a business advantage tool that provisions business

intelligence to user employees at all levels and locations of the

organization and which is based on existing and proven technology.

2. The Mobile Big Data functional capabilities derive from a seamless

convergence of Mobile platforms, Big Data processing environments and

Advanced Analytics.

3. The Big Data processing capabilities often derive from fast and reliable

MPP processors operating in specialized high throughput environments.

4. The Mobile platform based features provide user mobility and include

intuitive interfaces, business specific applications and data collection.

5. Advanced Analytics capabilities include the efficient analysis of multiple

data types (including liquid data and social media) and the blending of

resultant intelligence into a fully actionable format.

6. The Mobile Big Data deployment rationale should be based on

addressing defined business needs rather than technical capability.

Page 36: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 36

V. “WHERE” DOES MOBILE BIG DATA FIT BEST?

As to be expected, the cost effective application of Mobile Big Data is highly

dependent on the organization’s specific business requirements and the targeted

user’s typical activities and capabilities. In enterprises today a considerable

proportion of employee are “knowledge workers”, who on a regular basis are

using the best available information to make vital decisions. This situation is partially

due to the post great-recession economic pressure that causing many enterprises

to push decision making out to the organizational edges and greatly empower

their remote knowledge workers. Always recall that a key Mobile Big Data usage

rationale is its ability to facilitate better quality decision making while speeding the

decision making process, thus reducing decision latency.

Clearly the intrinsic value of intelligence is enhanced through the use of the best

available source data. Figure 5.1 displays the rather considerable estimated global

value still available to major vertical market segment enterprises by expanding

their business input data types to include available international market data.

Figure 5.1: Estimated Upper End Annual Value of Improving Global Intelligence by

Including Available Multiple International Data Sources

Source:McKinsey and Co.

It should be noted from Figure 5.1 that the upper range of overall potential benefit

from “globalizing” business intelligence exceeds $5.3 Trillion per year. Certainly

such global benefit estimates are difficult to substantiate but nonetheless indicate

a quite significant potential ROI through improved intelligence provisioning.

Page 37: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 37

To better focus on the types of enterprise environments most likely to be a good

Mobile Big Data application point (e.g. those environments most likely to provide a

solid measurable ROI) the “Where “discussion will first briefly categorize the typical

high ROI potential type environments followed by a few real world examples.

These early on success examples best serve to emphasize typical Mobile Big Data

benefits and to generally point out its likely direction. There will obviously be many

other types of successful Mobile Big Data implementations as it matures technically

and becomes increasingly cost effective.

A. ROI RICH ENVIRONMENTS PROFILE

Just as the Periodic Table of Elements categorizes materials according to their

basic atomic structure so does this ranking of likely enterprise environment related

ROI based on typical decision making processes. It should be recognized,

however, that the following categorization of ROI opportunities for Mobile Big Data

is not nearly so precise or scientific, and is based on the best currently available

information and the results obtained from a fairly limited number of to-date

implementations. The potential Mobile Big Data ROI ranking is in reality a set of

empirically based general guidelines. The identification of where Mobile Big Data

fits best begins and ends with the defined business needs and specific objectives

of each enterprise. There are indeed some emerging “best fit “characteristics

where user needs are similar, however, the specific business requirements of the

situation take clear precedence over any such general application guidelines.

Big Data Extension

Clearly a likely place to identify highly cost effective Mobile Big Data deployment

opportunities are those environments already invested in its biggest CAPX intense

element, Big Data. A solid ROI is likely when extending these existing Big Data

capabilities (out to the edge worker decision makers) where the primary cost

elements are likely limited to upgrades of existing infrastructure and new

application development. Even where a reasonable low improvement may be

projected for outcome metrics (3% customer contract renewal rate increase, etc.)

such environments can find a good rate of return for Mobile Big Data deployment.

Of course, from the Mobile Big Data implementation side those enterprises with

well-established Big Data (often, energy sector, medical, R+D, etc.) also tend to

have significant in house technical skills that can accelerate the deployment and

support of Mobile Big Data. Where new skills are required these typically can be

outsourced by an internal staff experienced in managing such efforts. The largest

skills shortage for such organizations is likely regarding Advanced Analytics

implementation and specific Mobile platform application development. In many

ways the ROI driven rationale for deploying Mobile Big Data is based on the fact

Page 38: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 38

that so much of what is needed already exists. With some fairly minor cost

upgrades and additions, it’s ready to be deployed.

Financially Valued Decisions

As experienced trading floor folks express so well in their appropriately labeled

“Race to Zero” efforts seeking absolutely no trading transaction latency, time is

money. Nanoseconds in that environment can have a readily defined value,

perhaps small individually but collectively huge due to massive trade volumes.

Other financial decision based organizations, including all varieties of trading, but

also banking, credit card transaction based, various types of underwriting, etc.

also have a clearly assignable value to concluding all their employee’s required

background decisions as rapidly as possible. Certainly speeding such decisions is

an intrinsic result of Mobile Big Data. Where as the time value calculation may be

somewhat obtuse for being able to accelerate marketing decisions for corn snack

foods, the value calculation for reducing the latency of a corn commodity trade is

fairly direct. Also, financial transaction based enterprises are information based

and thus typically have the ROI metrics in place needed to justify their often

massive investments in IT – Mobile Big Data is no exception.

Complex Event Response

Complex event driven enterprises that require employees to regularly make real

time complex decisions are another spot where Mobile Big Data can produce

solid ROI. Put yourself in the position of having to make time critical operating

decisions when directing the large and complex operational network of a major

airline, huge rail system, ocean going freight forwarder, etc. This complex event

driven situation is affected by many significant and ever changing variables

including weather, equipment and employee availability, etc. Providing the real-

time multi-source and fully actionable intelligence is a perfect application of

Mobile Big Data. Improving the operational metrics by just a few percentage

points for these high volume and low margin enterprises offers all the ROI needed.

Visualize an airline’s personnel being able to (1) have the support intelligence to

proactively make complex passenger rerouting decisions in response to a

developing weather event and (2) contact the passenger’s mobile phone to

inform them of the developing situation, and the proactive draft decision made for

their rerouting, and asking them to approve this plan – all without anyone talking

on the phone or passengers standing in line at an airport desk. The same type big

data based analytics driven complex network management with proactive

rerouting/notification delivered to employee and customer mobile devices is now

being done by Canada Rail.

Page 39: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 39

Mobile Big Data has a huge ROI potential for these difficult decision making

environments and indeed an entire set of off the shelf analytic tools is specific to

efficiently supporting complex event decisions. Complex event driven

environments may be where Mobile Big Data does best, especially with customer

facing edge employees. The ROI may not have the easy or direct metrics of

financial organizations, but the business results (lower cost and operational

improvements) can be quite solid.

Customer Facing Edge

Where employees regularly deal directly with customers in a remote location

(customer site, accident or disaster scene, kiosk location, etc.) it is appropriate to

consider a Mobile Big Data approach to enable better informed decisions. In this

situation the employee may be provided with competitor information, such as their

recent pricing and promotions, active product recalls, new services offerings, etc.

The end result is an edge worker with the information to be more effective in

working one on one with a customer or prospect, The practical business value from

better empowering edge worker employees may include higher sales close rates,

up-selling an increased ticket size, better contract renewal rates, etc. Without

increasing the expensive edge worker (and possibly reducing them) a revenue

increase is possible that may be accompanied by an improved customer

satisfaction rating, all due to enabling better informed and more responsive

representatives. Edge workers may also include emergency responders (EMTs, etc.)

where being able to receive immediate assistance, possibly based on interactive

information exchange, can be critical. Of course, the recipients of edge support

can also be the customers themselves or the citizens affected by governmental

agencies. Given the pervasive mobile platform availability (in some cases the only

available linkage) this type edge support is likely to expand dramatically.

B. REAL WORLD EXAMPLES

The following real world enterprise level Mobile Big Data examples are highlighted

for their results in successfully addressing business needs and producing

measurable ROI. They should be considered as reasonably representative but are

subject to continued refinement. These Mobile Big Data examples have mostly

been documented in various media so additional background information is

generally available.

Minimizing Disruption in Complex Transportation Network

ENVIRONMENT: Major Commercial Airline

ORGANIZATION; Delta Airlines Inc.

Page 40: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 40

The management of an airline network is a perfect example of a complex event

driven enterprise with many critical and highly changeable variables including

hard to maintain and expensive equipment assets, weather, skilled employee

resources, etc. Overlaid on this situation are regulatory constraints, union rules and

the obvious safety considerations, etc. Many airlines, Including Delta, have

implemented Big Data based CEP analytics to define up to the minute, and best

practices based actionable intelligence and recommendations. These outputs are

typically near real time and may be delivered to a variety of devices including

mobile platforms. The information may include schedule changes, routing revision,

equipment and crew assignment revision, baggage problem alerts, etc. A portion

of this CEP produced data may be routed to edge worker devices and even to

the customer’s mobile phones. The results have included reduced delays,

operational efficiency improvement driven savings, and better customer relations.

Edge Worker Communication and Citizen Mobilization

ENVIRONMENT: Large Metropolitan Government

ORGANIZATION; City Of Minneapolis, Minn.

Many governmental agencies at all levels use a Mobile Big Data approach to

more cost effectively communicate with and instruct their dispersed edge worker

employees and to alert and mobilize citizens. This specific example uses CEP type

analytics based on a wide variety of input data types and sources (including EDW,

external data feeds, sensor type “liquid data”, weather service input, etc.) along

with a set of best practices based algorithm decision rules and recommendations

to optimize event response. Alerts and recommended action instructions are sent

to both edge worker employees and to the mobile devices of likely to be affected

citizens. The potential event situations addressed include public safety issues

(weather, terrorist threats, etc.) traffic management, time critical municipal

permitting, and management of entertainment venues. To optimize ease of use a

variety of custom tailored dashboards (including pie charts, etc.) and intuitive

displays (including mapping, etc.) are utilized along with reference links and

accessible portals. The results have included savings from improved operational

efficiencies, significantly fewer public safety disruptions and increased citizen

participation and assistance.

Improved Customer Retention and Cross Selling

ENVIRONMENT: Major Insurance Carrier

ORGANIZATION: Metropolitan Life Ins. Co.

Page 41: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 41

As a part of its ongoing push for improved customer service and higher insurance

renewal rates MetLife introduced a social media inclusive analytics approach to

identify potentially dissatisfied customers (without any formal complaint) and

proactively approach them with proposed improvement proposals including cross

selling for appropriate additional products. The potential customer satisfaction

analytics use a combination of internally sourced data (EDW based) along with

social media input to prepare recommended actions. This information is forwarded

to the appropriate edge worker account manager who then contacts the

customer (as a routine satisfaction survey) and then takes the opportunity to

propose improvements. The customers have typically appreciated this positive

approach and as a result policy renewal rates have climbed and product cross

selling results have raised about four fold. The Mobile Big Data ROI basis of

increasing revenues has proven to be correct.

Accelerating Product Claims Response

ENVIRONMENT: Large Pharmaceutical Maker

ORGANIZATION: Confidential

A large pharmaceutical firm’s customers when making various types of product

related claims (in-transit damage, incorrect amounts, wrong SKU, etc.) were

displeased with the claim resolution delays they experienced, which they

considered as indifferent customer service, and this situation negatively affected

customer reordering and cross selling. The root cause of the claims resolution delay

problems was the claim processing employee’s typical difficulty in accessing and

understanding the sometimes complex information required for them to make a

claim resolution decision. The decision support information often had to be

obtained from a variety of external and internal sources, often without cross

referencing or contextual content. To address these decision delay issues an

advanced self-serve analytics system was established and a set of best practices

algorithms based recommendations defined. In this manner employees trying to

resolve open product claims (whether edge worker or central employees) are

able to efficiently access the data they need for their specific product claim and

receive guidance on recommended resolution actions. As a result the typical

product claim resolution period has dropped from a little over a month to less than

a week. The resultant positive customer service experience had a significant halo

effect on increasing customer reorders and cross selling success.

Page 42: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 42

VI. RECOMMENDED MOBILE BIG DATA BEST PRACTICES

As stated, Mobile Big data remains in its early stages and continues to evolve at a

rapid pace so as yet there is no definitive manual of best practices. The

experience gained from initial Mobile Big Data deployments and the operation of

its primary component elements has, however, been adequate to identify some

initial best practice recommendations. Although preliminary in nature these

suggested general recommendations are worth serious consideration.

A. ORGANIZATIONAL RECOMMENDATIONS

In that the technical foundation of Mobile Big Data is based on three solid and

proven components there are limited technical factors to be considered as true

best practices. In fact, most Mobile Big Data deployment concerns and

recommendations generally relate more to the preparation of the organization

than to technical factors. The root cause of these organizational issues is the fact

that the business intelligence prepared and delivered by Mobile Big Data, is often

a new concept. As a result the implementers and technical support teams have to

understand their role and the recipients need to know how to best utilize this new

resource and what is expected of them.

Management: Stand and Deliver

The biggest initial organizational hurdle is the need for visible senior management

sponsorship. As with any major business practices change, senior management

must lead the way and be very clear about why this effort is being done, the

expected impact and the roles of all involved. In the eyes of business intelligence

recipients this issue may be the most important deployment success factor. The

sponsoring senior management needs to fully understand and support the value

proposition and leave nothing vague or ill defined. The kick off needs to include a

senior management “here’s why we’re doing this and here’s is what expected

from each of you “type introduction. Obviously, the clear definition of solid

business driven ROI metrics for evaluating results will go a long way toward taking

this Mobile Big Data business advantage initiative out of any potential “science

project “type status.

The next issue might well be exactly WHO will be the sponsoring senior executive in

charge of the Mobile Big Data initiative. In many organizations, largely due to its

technical nature, Big Data logically fell under the purview of the CIO. The CIO was

responsible for the EDW so extending the user management responsibility to all

aspects of Big Data made sense. Mobile Big Data isn’t quite so clear. The mobile

device delivery platform is frequently BYOD (if not actually BYOE) based and the

Page 43: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 43

users, and their management, deal directly with those issues. Security is always a

concern but that seems to have been sidestepped and can be minimized at the

Big Data interface levels. The business intelligence user management

(sales/marketing, claims processing, etc.) may be closer to the Mobile Big Data

value proposition and thus better able to see that the provisioned business

intelligence is used as intended and that results occur as planned. Figure 6.1

displays that largely for these reasons, in many enterprises today the actual

business intelligence analytics operation (as distinct from the processing platforms)

may report outside the IT organization.

Figure 6.1: Location of Business Intelligence Analytics Management

Source: Several Recent Surveys

It’s of significance to note that a specialized separate analytics management

function is currently the most common business intelligence analytics

management location.

Empiricists vs. Analysts: MIA Skills

The second key organizationally related issue for Mobile Big Data deployment that

often needs to be addressed is the fact that experienced intelligence analysts

may not exist. Frequently there has been little to no reason previously to have fully

knowledgeable analysts to perform analytics processes or intelligence gathering.

Those are distinct skills from typical EDW management and may include dealing

with different data types (social media information, liquid data, etc.) many of

which are semi-structured or unstructured. In addition, the ability to define business

needs driven heuristics and decision model algorithms may not have been

encountered and the interpretation of social media derived output can resemble

Page 44: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 44

a “soft science”. Given the importance of the business decisions that may be

based on the intelligence provided, this analyst work may be something best not

left to amateurs.

Directly related to the Mobile Big Data requirement for specialized analysts is the

difference between data empiricists (often DBA specialists) and true analytics

experts. An empiricist may tend to defer to statistically proven methodologies and

analysis outcomes, even when such approaches are impractical. Whereas a true

analyst in such a situation may cross reference the information obtained from

multiple data sources and types to produce actionable recommendations.

Effective intelligence development often requires an experienced interpretation

going well beyond professorial types of theorem proofs. A clear example of the

value of these analyst interpretive skills is the use of data visualization analytics

where the analyst first examines the visual pattern of a huge data array to identify

possible patterns. These are the valuable personal skills of the flexible analyst versus

the more rigid approach of a empiricist.

User and Tech Staff Preparation

Another organizational issue regarding Mobile Big data deployment is user and

technical staff training and preparation. An early on effort to evaluate user needs,

capabilities and skills should drive the definition of a well thought through training

experience for these employees. In large and spread out user organizations it may

be necessary to adopt a train the trainer type approach but there’s no substitute

for having a knowledgeable person of some variety in close proximity to users

during start up. The training involved must closely resemble the real world

environment and include decision situations that users will likely encounter and not

be theoretical or overly technically based. The users need to understand enough

about how the intelligence was derived that they can believe in its validity and be

comfortable using it (more than they apparently do today) but they don’t need a

PhD in statistics to reach this comfort level. They need to use good judgment in

applying business intelligence, but they don’t require a copy of all the algorithms

being used.

With regard to tech staff preparation it’s often possible to send the operational

staff, and analysts as required, through training classes as specific to the technical

tools included. Most suppliers of such software and hardware components offer

canned training programs and supplemental webinars on a regular basis. The

technical staff’s full familiarity with these systems likely will come mostly through

actual set up and rigorous testing.

Page 45: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 45

B. IMPLEMENTATION BEST PRACTICE

As with organizational Best practices the implementation related

recommendations for technically related issues remain a work in process. For the

most part the key point to Mobile Big Data implementation is spending the time

and effort needed for very thorough pre go-live testing and adjustment. Identifying

the appropriate input data types and sources and defining the detailed analytics

decision rules and algorithms can be time consuming but can’t be short cut, these

steps are critical to the accuracy of the resulting intelligence.

Make vs. Buy vs. Both

As a general rule the ability to license proven off the shelf analytics components

typically outweighs the development of such routines. The same applies to the Big

Data related elements. These two key Mobile Big Data elements can be obtained

off the shelf in a fairly easy to set up, integrate, and use form. The mobile platform

apps specific to a business situation may be another matter. In some cases the

required mobile applications may be available, but most likely these will require

significant customization to better match specific conditions and to be totally user

friendly. The user interfaces must be as understandable and intuitive as possible so

should be almost conversational. This means including the business terminology,

and jargon, already familiar to the user set. Please recall the intelligence now

received is not felt complete or easy to use and for that reason is often ignored. It’s

likely in a major Mobile Big Data installation to encounter both standard off the

shelf type elements as well as custom developed or heavily modified components.

Iterative Piloting with Metrics

Iterative piloting with metrics relates to the best practice of starting small, then

growing, when introducing a new technology. That doesn’t mean the starting

point won’t be an important situation, it may be just the opposite and the

beginning point may well address a critical business need. The iterative piloting

process allows for careful front end analysis and solid testing prior to entering

production status. Especially important during this effort are user perceptions and

feedback and close metrics monitoring. Mobile Big Data represents a new

concept and can be disruptive – rigorous front end testing and fine tuning is

absolutely warranted to be sure it’s working as planned and expected.

Page 46: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 46

Stepwise Extension

To avoid deadly Big Bang Theory start up issues, it’s always best to start with a

limited scope and then expand as appropriate. This means beginning with the

most likely to succeed sub-set of users – perhaps those with the most related

experience or those most convenient to training and start-up support. The less

skilled or less interested users, and possible those most remote, can be added

once things are running well. Obviously, that approach may be difficult to do

where the prime target Mobile Big Data opportunity is with edge workers - in that

case a region by region roll out plan may be appropiate. Starting with a limited,

and success likely user set, then expanding with a step-wise deployment including

published schedules is the preferred Mobile Big Data deployment approach.

Page 47: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 47

GLOSSARY OF TERMS AND ABBREVIATIONS

A glossary definition of the specialized abbreviations and terminology contained in

this white paper includes.

- IoT; Internet of Things is a term referring to the internet connection of

devices rather than people.

- MPP: Massive Parallel Processing as performed by an array of smaller

platforms aligned and controlled to operate as a single huge throughput

system.

- NoSQL : Not Only SQL is an indication that several data base types and

structures will be used ( in addition to SQL format DBs )

- CEP: Complex Event Processing, the typically real time (or near real time)

processing of several data streams or variables that contribute to the

situation being managed.

- EDW : The Enterprise Data Warehouse primarily refers to the internal data

repository of the enterprise and the data is typically kept in a hierarchical

and/or relational manner

- Apache Hadoop; The open source and licensable version of Hadoop.

- MDB; Master Data Base typically refers to the structured data contained in

the EDW.

Page 48: Mobile Big Data

Mobile Big Data

Fusion Labs, Inc Page 48

For further information please contact:

Brian R. Blackmarr Martin Ward

Chairman Vice President of Sales

Fusion Laboratories Inc. Technical Services Division

214.217.9783 (office) 214.217.9763 (office)

214.435.4433 (cell) 214. 909.7709 (cell)

Two Lincoln Centre

5420 LBJ Frwy. Ste. 850

Dallas, Texas 75240

214.739.5454

www.fusionlabs.net