the future of community insights
DESCRIPTION
Online Communities are changing. They're getting larger, sprawling and multifaceted. At the same time, they're also getting smaller, and more specific. They're become more diverse, interconnected, multiplatform - and the tools needed to interact with, find and manage these communities are having to change with them. We present three fundamentally important technologies: Machine Learning (A.I.), Network Analysis and Text Mining - that we believe will underpin the future of community insight. What they are, why they are important, and how you can use them to bring your brand and its community closer together.TRANSCRIPT
The future of Community Insights
Michael ConroySenior Analyst, Tempero @mickyconroy
So who am I then?
Mick ConroySenior Social Media AnalystTempero
Tempero’s Clients
Let’s get started!
1. Network Analysis
2. Machine Learning (A.I.)
3. Natural Language Processing
CM and Insights together
Community Mapping
(Networks)
Conversation Analysis
(Text Mining)
CommunityEngagement
CommunityOutreach
(M. Learning)
Community Insights
Community Management
Network Analysis
Community Analysis
Thinking in “networks”
Social networks are networks. Duh.
If we start analysing social media like networks new and useful insights arise
This is the New York Times
...and this is Forbes
Positive
Negative
Neutral
Fans @mention Barlow their love of the new track
“Barlow can stick those Percy Pigs...” gets RTs
Behind the scenes photos shared
Barlow like “slashing the Mona Lisa with a platinum machete”.
Conversation Analysis
The power of networks
Influencers can be found by their place in the network
Spread content in a way that it gets seen
The power of networks
Virality can be measured
Going viral can be induced
Going viral is a science
Going viral is a science
Find the influencers that connect different communities
Subcommunities: Find any interests groups within your community to determine how to seed your content
The tools
mappr
Start with a tight community
Influence the connectors
10% is enough
Be relevant
Your message should be closely related to the purpose that brings the target network together
Acknowledge the group’s mindset by bringing in some novelty
London’s Social Media Map
mappr
UK Fashion and LifestyleLuxury Fashion Brands
Home and Decor
Magazine/ Design
companies
Art auction houses ,
publications and a
museum
Luxury Travel Blogs,
sites
Luxury watch
brands, blogs
mappr
Summary
Network analysis helps us find the people and conversations that matter, and take the right actions
By knowing the structure of the network, you can better model how to influence it
Machine Learning
This is the Internet
You look at this much
Intro to Machine Learning
Train artificial intelligence on a big dataset
AI makes a prediction based on what it currently “knows”
AI recalculates what it “knows” based on the new information
Get feedback on whether the prediction was right or wrong
It’s everywhere!
It’s also handy for Social
Social media monitoring 2.0
AI can be trained to look for concepts or themes in text, independent of keywords
Human level intelligence – at scale
Humans and Algorithms...
A few tagging examples
The new way
Train the AI around a bespoke classification taxonomy
One trained, the AI is able to tag mentions autonomously
Keep human experts on hand for minor corrections
Sample Taxonomy
Message-typeNewsTechnical QueryComparisonAutomated shareAdvertContestOpinionProduct FaultCustomer Service
Product CharacteristicsPriceEase-of-useCompatibilitydesign/buildAvailabilityEnergy Efficiency
PersonCustomerRival customerPressPublicStaff/Sony
Product featuresoundset upinternal softwareappsScreen SizeResolution/Picture QualityMedia PlaybackMobile/tablet/PC connectivity
BrandSony BraviaSamsungLGSharp
Product-type3D Smart/Internet TVLED TVPlasma
The end result
Sony Samsung Apple
Customer Service Sentiment by Brand Worldwide
The end result
Purchase Intent by Competitor
28-Aug 29-Aug 30-Aug 31-Aug 01-Sep 02-Sep 03-Sep 04-Sep 05-Sep0
5000
10000
15000
20000
25000
Sony Samsung Panasonic
Deep insights, at a glance
Summary
Human-level intelligence at scale
Deep insights; Fast.
The basis for a laser-targeted engagement strategy
Natural Language Processing
We’re all drowning in text!
CM’s have their fair share
Twitter status updates
Facebook comments
…there’s interaction data everywhere
It’s not just “us”…
Introducing Overview
overview.ap.org
And these drones…
Acquire your data
How Overview works
Activists use drones to track endangered wildlife
Apple rejects iPhone App
Overview in action
Represent the data
Pakistan D
rone St
rike
Parrot A
R Dro
ne
Surve
illance
Bradford
Protest
Haqqani
Imra
n Khan
Afganistan D
rone St
rike
US Dro
ne Enquiry
Robotic Horse
Australia
n Miss
ions
Yemen Dro
ne War
Apple Rejects App
Debates about m
orality
Israeli d
rone enters
Lebanon airs
pace
Black O
ps 2
Drone H
acking
Win a dro
ne
Drone Jo
urnalis
m
Police U
se
Immigra
tion Control
Citizen Su
rveilla
nce
Conserva
tion
BBC discusse
s priv
acy
Airspace
safety
conce
rns
Alan Sugar m
isspells
"dro
ves"
Drone as i
nsult
0
50
100
150
200
250
300
350
Represent the data
Football fans at a match
Summary
A new toolset is needed to analyse the sheer volume of conversation we all have to deal with
Natural Language Processing algorithms are an ideal way to find meaning in the clutter
In closing...
Pulling it all together
Community Mapping
(Networks)
Conversation Analysis
(Text Mining)
CommunityEngagement
CommunityOutreach
(M. Learning)
Community Insights
Community Management
There’s a lot more to it
We can help!
Michael ConroySenior Analyst, Tempero@mickyconroy
Thanks!