predictive data analytics to help your customers

93
#DataTalk Predictive Data Analytics to Help Your Customers

Upload: experianus

Post on 21-Apr-2017

10.183 views

Category:

Data & Analytics


1 download

TRANSCRIPT

#DataTalk Predictive Data Analytics to Help Your Customers

Join our #DataTalk on Thursdays at 5 p.m. ET

This week, we tweeted with Michael Beygelman, Co-founder and CEO of Joberate, Berry Diepeveen, Partner and Enterprise Intelligence Leader at EY, and Chuck Robida, Chief Scientist for Experian Decision Analytics.

Check out all tweets from this Twitter chat:

ex.pn/predictive

What is predictive analytics?

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Predictive analytics is extractinginformation from data sets to determinepatterns, predict outcomes and trends.

Chuck RobidaChief Scientist, Experian@ExperianDA ex.pn/datatalk

#DataTalk

Predictive analytics is the abilityto use data to predict future behavior

based on past behavior.

Berry DiepeveenPartner, EY@Berry_Diepeveen ex.pn/datatalk

#DataTalk

I think it is as old as business.Nobody can perfectly predict the future,but you want to be more accurate about

what is likely to happen.

Chuck RobidaChief Scientist, Experian @ExperianDA ex.pn/datatalk

#DataTalk

It’s the ability to analyze data in a waythat can scale, be reproduced, and

provide unbiased results.

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Clever marketers are redefiningpredictive analytics into whatever suits

them today, so we need to beware.

Chuck RobidaChief Scientist, Experian @ExperianDA ex.pn/datatalk

#DataTalk

Some techniques get more attentionthan others like machine learning,

but all are used to solve business problems.

Berry DiepeveenPartner, EY@Berry_Diepeveen ex.pn/datatalk

#DataTalk

It is about being able to intervene.What is the point of finding out we

lost a customer after he left? We need to prevent losing one before it happens.

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

I’ve always said that predictive analyticsneeds to be actionable like a brake system

in a car. When you press, it does something.

Chuck RobidaChief Scientist, Experian @ExperianDA ex.pn/datatalk

#DataTalk

Predictive analytics isn’t a crystal ball, but the value comes in identifying

the propensity of certain behaviors.

How trustworthy is predictive analytics?

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

Predictive analytics is very trustworthy,but not when used in pure isolation.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Predictive analytics has a natural lifeand models need to be continually

validated and aligned to changes in behavior.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

It’s not just about the predictive modeling,statistics and the algorithms. It’s also about

the play, experiments, intuition and innovation.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Behaviors change so should your models.Economic change-inflation, housing,

unemployment. Life change: marriage, kids, job ...

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

In terms of relevance, if associated withsome decisions of value or have meaning,predictive analytics can be very relevant.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

And you must deal with in a sensible way,especially around sensitive use cases

such as fraud detection.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Simply put, predictive analytics still comesdown to a cost/benefit decision.

Use it as your compass.

What type of data do companies use for predictive analytics?

Depends on the business goal.Generally a mix of fit-for-purpose internal

and external data types, structuredor unstructured data.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Sometimes companies start by usinginternal data. In my world, payroll data, promotions, performance reviews, etc.

#DataTalkMichael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

Depending on how much success they have with internal data, and how quickly,they’ll usually broaden out to third-party data.

#DataTalkMichael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

For lenders: asset evaluations for loans,address change for collections -- all good data,

in compliance with regulation.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

For marketing: social, contact history,profile data, all good data.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

If we go back to the fraud detectionuse case; you’d have to rely on

internal, external, structured and unstructured data.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

The beauty is that there is no limit aboutwhat data sources you want to tap into.

It’s always driven by the business and use case,not the other way around.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

Only limitations are legal, complianceand your imaginations.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

How much data preparation needs to be done before

executing predictive analytics?

It requires a very tight collaboration betweenbusiness and data science in order

to determine the iterations.

Berry DiepeveenPartner, EY@Berry_Diepeveen ex.pn/datatalk

#DataTalk

Data preparation is arguably as important as the rest of the process.

ex.pn/datatalk#DataTalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Garbage in, garbage out.Data preparation is the most important step.

Incorrect or insufficient data equalsbad business decisions

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

We see three phases in any predictive analyticsprogram: 1) strict data management, 2) building and applying advanced analytics models, and

3) using data visualization to bring the insights back to the end user.

Berry DiepeveenPartner, EY@Berry_Diepeveen ex.pn/datatalk

#DataTalk

If the sample size is massive, it mightbe more practical to sample the data;else you can use the whole sample.

ex.pn/datatalk#DataTalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Without strict and rigorous data management,you should question your investments

in data science.

Berry DiepeveenPartner, EY@Berry_Diepeveen ex.pn/datatalk

#DataTalk

Decide what to do with incomplete data,discard it or take guesses at missing data points

by looking at other data in the sample.

ex.pn/datatalk#DataTalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Be careful before tossing any data. Bias!

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Many activities like selecting, combining,and aggregating data are important,

especially when defining the form for training.

ex.pn/datatalk#DataTalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

How often should models get updated?

It’s more of a business decision.If your data is updated quarterly,

no point in updating a modelmore often than that.

ex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

#DataTalk

Frequent model evaluation or validationis critical + results should be taken

in context of other solutionsand external factors.

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Building good models is the science.It involves experimentation,

sufficient quality data and is time consuming.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

If data is updated daily, and you choose to update the model quarterly,

you might have to live with somebad assumptions.

ex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

#DataTalk

Expect a model to naturally deteriorateover time. Predictive analytics

needs to be continually validated forfit for purpose.

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Regardless of the use case, you needto update models regularly and

structurally, but additional ad hocupdates depend on use case.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

Models are fit-for-purpose and considerthings like economy, home values...Tests + benchmarks exist to ensure

models are robuts.

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

What is often forgotten is that newmodels have to be retrainedwith the updated data sets -

and results verified.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

What are the best ways totest the effectiveness of

predictive analytics?

There are many scientific ways to test,but the real question is did the analytics

provide you with actionable insights,at the right time.

#DataTalkex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

Splitting data at the outset could bea good idea so you’re not accidentally

creating a super model that onlyworks on one set.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Deploy them in a manner wheretheir impact can be measured in a

controlled environment likechampion-challenger testing.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Use a majority of the data (say 65% or so)for the build of the model, anduse the 35% of the data for the

test of the model.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

There are numerous ways to testmodels, and some people swear

by some approaches almost like religion.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Test models by using data notused during development.

Validation won’t yield same results,so benchmarking plays a big role.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

One can use lift charts, decile tables,some people like to use target shuffling.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

What are ways companies can use predictive analytics

in new ways?

Possibilities are endless,but business focus is key.

ex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalk

Predictive analytics used to scientificallypredict anything from the future state

economy + weather to spread + cures for disease.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Newest technologies allow youto quite efficiently translateunstructured into structured

such that it can be included in models.

ex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalk

#DataTalkMichael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

I spoke to a gentleman atGoldman Sachs. They were using

predictive analytics in the hiring process.

#DataTalkMichael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

Goldman Sachs used analysis ofincoming CVs to compare to

top performers and those who hada cultural fit.

What are the challenges when working in predictive analytics?

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Challenges? Too many :)But making sure you have ample

relevant data is important,and making sure you have tested models

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Challenges working with predictiveanalytics: data availability,

quality, volume, and statisticallyrobust sample size.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

We need to seriously considerdata ownership and data privacy

in every single engagement.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Compliance and control over permissible purposes presentchallenges, especially when

rich data can’t be used.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

Too many examples of not treatingdata as a real asset that

ultimately belongs to the customer.

Berry DiepeveenPartner, EY@Berry_Diepeveen

#DataTalkex.pn/datatalk

Another challenge is resourcesand finding professionals who have

skills around technology, techniques,modeling and business acumen.

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Globalization of predictive analyticsis another challenge.

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

In more mature markets, the uptakeis “simpler” while in other marketsless so, which creates challenges

for global organizations.

What trends are happening in predictive analytics?

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

In terms of trends, machine learningto automate the analytics process

itself is certainly one of thebigger trends.

ex.pn/datatalk#DataTalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

We are predicting the future ofpredictive analytics now.

We need an algorithm and a model. :)

ex.pn/datatalk#DataTalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Available data + advanced statistics+ new processing tech = businesses

+ can build more meaningful + relationships with consumers.

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Another trend hard to ignore is thedatafication of our lives;

basketballs to tennis rackets, and Lumo Lift to help you stop slouching

Michael BeygelmanCEO, Joberate@beygelman @joberate ex.pn/datatalk

#DataTalk

Along the datafication continuum,data privacy laws are severely lagging

and will need attention.

ex.pn/datatalk#DataTalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

Look at fantastic innovations fromenterprise technology providers

and new ventures that revolutionizethe markets

Any final tips for companies working in predictive analytics?

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Become more community basedrather than managed by centralized

IT at big companies, or siloed insome underfunded organization. :)

#DataTalkex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

Absolutely! The internet of things iscreating great opportunities where we

have seen completely new business models.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Business new to predictive analyticsmaintain robust model validation

methodology. Using a broken modelwill cost you money.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

The input of community into the evolution of predictive analyticscan have profound open-source

like benefits.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Sophisticated users of predictiveanalytics, remember to start with

the business problem and work backwards.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

My best tip for working inpredictive analytics is walk,

don’t run. Make sure you take verydeliberate first steps.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Predictive analytics serves to existfor solving complex business problems.

Start with end in mind.

#DataTalkex.pn/datatalk

Michael BeygelmanCEO, Joberate@beygelman @joberate

Decide that you will have a cultureof analytics and then move

into that area. Don’t “test” analyticsto see if they’re “for you.”

#DataTalkex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

Do not underestimate how importantthe data visualization is to end

user adoption of predictive analytics.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

The world of analytics and datais exploding. It’s critical to prioritize

analytical opportunities to stay ahead of competition.

#DataTalkex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

And do not underestimatehow important the data visualization

is to end user adoption of predictive analytics.

#DataTalkex.pn/datatalk

Berry DiepeveenPartner, EY@Berry_Diepeveen

The business needs to work withthe insights and it is not about

developing the most accurate andcomplex model.

#DataTalkex.pn/datatalk

Chuck RobidaChief Scientist, Experian @ExperianDA

Simply put, predictive analyticsstill comes down to a cost/benefitdecision. Use it as your compass.

Join our #DataTalk on Twitter on Thursdays at 5 p.m. ET.

experian.com/datatalk