from data to insights building accurate consumer insights from online conversations francesco d’orazio - research director, head of social media, face jake steadman - brand and social media insights, O2
May 09, 2015
from data to insightsbuilding accurate consumer insights from online conversations
francesco d’orazio - research director, head of social media, facejake steadman - brand and social media insights, O2
the business context
1. To plug social media monitoring into the heart of businessand brand planning
2. To help maximise the potential of executions in terms ofSocial Networking
3. To understand how we can target media and seed better thantoday
4. To help with gaining proactive insight – early discovery ofemerging trends / issues / opportunities
5. Brand reputation management of both consumers and themedia
The overriding research objective was to utiliseone tool for Traditional media and Social media,covering:
1. Media and consumer social media monitoring2. Buzz monitoring and sentiment tracking3. Brand reputation management and performance4. Media performance5. Messaging success
what social media means for research
first, understand what users want from the web
the web as...
mirror projector constructionfrom ‘Solipsistic Branding’ | Lawes Consulting | 2007
identity - who I am
contacts - who I know
activities - what I like
...now add ‘social data’
ongoing conversations, rather thenQ&A or one-off discussions
a window intonaturally occurring behaviors
far less obtrusive
re-introduces social context
a real-time stream of data
this makes social mediathe richest insights field ever
for the first timewe can do
qualitative researchon a mass scale!
impressive...
but there are still some open challenges:
demographicsgeographysentimentsourcesspam
this is why social media analysis is all about...
reducing complexity
partitioning
hierarchy
independence
mining for patterns
decoding the shape of phenomena,rather then micro values
what we did
RTO2 is a social media monitoring andanalysis solution designed for O2.
RTO2 it’s software +strategic consultancy
outputs:
1) bespoke dashboard2) weekly reports3) monthly conferencing4) quarterly workshops
web crawler
A web crawler browses theWorld Wide Web searching for
specific key-phrases in amethodical, automated manner.
It creates a copy of all therelevant pages and stores themin an online database for later
processing
analytics
Software analytics tools andhuman analysts process theinformation harvested by the
web crawler in real-time
online DB
dashboard
A live interactive dashboardgenerates real-time data
visualization and customizedreports
the real-time dashboard
real-time tracking and measurement to enableadaptive brand management
visual mining and natural interaction with datato support pattern recognition
and iterative thinking
volumes are nothing without weighting
a channel-specific ‘visibility’algorithm takes into account
reach, engagement andfeatures of each messageto assess its ‘influence’
human sentiment analysis is key:we crowd-sourced it
a) access to a on-demand, 24/7, scalable workforceb) volumes: thousands of messages analyzed in a few hoursc) lower cost per message coded
Benefits of this approach
Analysts Company
embedded KPIs tracking
overall volumesvolumes by dayvolumes by typeoverall sentimentoverall visibilityvisibility by dayvisibility by type
qualitative visibility
outputs
volumes/day
volumes/share of voice
volumes/channel
volumes/reactivity
visibility
visibility / bands
sentiment
qualitative visibility
topics/wordclouds
topics/treemaps
trending topics / strata
brand social graph
influencers
geography
from data to insights
Human-codedSoftware-coded
FACE
Human-codedSoftware-coded
this iscool..
PICTURE
BLOG
noise reduction
targeted monitoring
refined mentions
analysis
web scan
CHAT
sawthis on
fb
did youhearthat..
love thisad...
VIDEO
thepackaginglooks cool
but...
checkthis out
POLLS
reporting
feedback
feedback
feedback
feedback
action action
FACE
FACE
analysis funnel
step #1landscapesummary of the mostimportant data for anyspecific search
step #2leadsidentify the leadsthat might bepointing towardsrelevant phenomena
step #3drill-downdifferent types ofleads requiredifferent types ofdrill-down action
step #4eventsexplain whatcaused a specificphenomenon
step #5insights
tactical“reactions show that speed is not the main concern andthere is a solid audience for propositions that aretailored on usage levels”
“identify the most influential sources that have beendiscussing the topic and engage them about the launchof the new broadband packages”
strategic“providing original research-based information whichsources and users can embed within their contentsdrives high profile mentions and viral engagement”
step #6trends+models
bring together events + insights in acomprehensive trend that tells the story ofthe brand over a specific period of time
identify recursive patters and suggest actionswhich will cause a trend to emerge again
insight/innovation: topic-trend explorationbrand/pr: brand reputation - share of voicebrand/pr: wider landscapebrand/advertising/pr: product/service/campaign trackingcustomer care: targeted customer engagement / social CRM
it’s new and exciting,and we are already using it:
1) Key word search continuity2) Scale3) Education/training
but there are immediate practical challenges:
thanks