I presented this at ICT Spring Europe 2015 in Luxembourg. The presentation highlights the way in which big data investments are not always delivering on their promise and why brands should consider taking a 'human-centred' approach to big data analytics.
3. Fundamentally changes business models:
Health clubs become
DIY fitness?
Death of the
shopping basket?
QSR becomes
personalised?
Transport
becomes social
60% of business executives believe big data will disrupt their industry within 3 years
Capgemini Consulting
4. The stakes are high:
McKinsey has estimated that a retailer using big data
can potentially increase its margin by more than 60%.
5. Fuelling a technology arms race:
AT Kearney forecast
value of big data tech
market will be $114B
by 2018
6. BUT:
Through 2017, 60% of Big Data projects will fail to
go beyond piloting and experimentation and will be
abandoned Gartner
72% of business and analytics leaders
aren’t satisfied with how long it takes to
retrieve the insights they need from data
Alteryx
65% of CEOS think their organisation is able to
interpret only a small proportion of the
information to which they have access
The Economist
Only 27% the executives surveyed
described their Big Data initiatives as
successful Capgemini
7. Even with all the capabilities and
tools in place, we are drowning in
data and starving for insight
Global bank quoted by Forrester
8. Some of the solution is organisational:
Scattered data lying in silos across the organisation
Absence of a clear business case for funding &
implementation
Ineffective co-ordination of big data and analytics
teams across the business
Dependence on legacy systems for data processing
and management
Sourced from Cracking the Data Conundrum : Capgemini Consulting
9. Some good use cases of how to manage
analytics:
Nordstrom Data Lab:
Multi-disciplinary team of data scientists, mathematicians,
programmers and business professionals. Continuous build and
test prototypes to take new products to market rapidly
AT&T Foundry:
Innovation centre that draws on network both within and with
partners using data sources to review and refine developments
P&G Decision cockpits
A single ‘point of truth’ for all decision makers across geographies
and business units – dashboards that aggregate complex data
with drill down facilities. Used by 58k people weekly, speeded up
decision making and time to market
Sourced from Cracking the Data Conundrum : Capgemini Consulting
10. The numbers have no way of speaking for
themselves. We speak for them. We imbue them
with meaning. Before we demand more of our
data, we need to demand more of ourselves.
Nate Silver
But is this enough?
11. So just what is a data scientist?
Data scientist job typically involves:
• Mathematical modelling of human behaviour
• Mainly predictive analytics
• Drawn from numerate disciplines
“As the amount of data goes up, the importance of human judgment should go down”
Andrew McAfee, MIT Sloan School of Management
“I have lost count of the times I have been presented with some amazing fact that
data has told us through the use of some incredible new technology, to be left
thinking “so what?” or “isn’t that obvious?”
Caroline Morris Sky IQ
12. Human side of analytics:
29 different teams of analysts asked to determine
whether soccer refs more likely to give red cards to
players with darker skin tones.
• Each team was given an identical dataset.
• 21 different sets of variables chosen for analysis.
• Different teams used different statistical models.
No surprise that teams came to fundamentally
different conclusions
Subjective judgements are embedded in the way in which we generate,
process and analyse data
Equip teams with cognitive and behavioural scientists who understand how
people perceive problems and analyse data
13. Separating the signal from the noise:
Our predictions may be more prone to failure in the era of Big Data
• In a big data world statistical significance is no a longer
reliable means of discrimination
• Modellers and statisticians may well be ‘getting it wrong’
• Studies suggest that as much as 90% of published medical
information that doctors rely on is flawed
Marketing & consumer insights need to be embedded in the team to disentangle
signal from noise
14. Made more complicated by:
Privacy backlashUncanny Valley
Need to understand the context in which data analytics plays in the real
world
Feedback loops
15. Measurement of outcomes:
Challenges in determining
attribution of advertising effect:
• Advertising blocking
• Advertising fraud
• Teasing apart background from
campaign effects
• Teasing apart retargeting
outcomes
The story the data
tells us is often the
one we’d like to
hear, and we
usually make sure
that it has a happy
ending
67% of business executives do not have well defined criteria to measure the success
for their big data initiatives
Capgemini Consulting
17. A more rounded view of the consumer:
SCVs are often limited to the brands’ data assets & data brokers
Help brands by:
• Identifying new sources
• Doing due diligence on the data assets
• Integrating (at a consumer level)
• Identifying value exchange for
consumers
Rapidly emerging personal data economy creates opportunities and
threats for brands
18. Train team in thinking about thinking:
Expert judgement as susceptible as the layperson
Cognitive pitfalls
Over-reliance on
statistical significance
Confusing correlation
with causation
Fallacy patterns
Role of theory
Data does not speak for itself
Danger of implicit models
What explicit models to
consider
Organised mind
Distinction between lab
and factory
Defining the questions
Avoiding vanity metrics
Data provenance
Representativeness
Sources of bias
Caveman effects
19. Use data for new sources of insight:
What ‘soft’ attributes can be derived from
behavioural data?
• Personality attributes
• Cognitive styles
• Satisfaction with purchase
• Intention to purchase
• Copying behaviours
Insight is often considered to be purely hard behavioural but the real need is to
understand the soft issues - attitudes and needs
Exciting area: but need to understand limitations as well as opportunities
20. Closing thoughts:
• Technology and organisation investment a necessary but
not sufficient condition for successful data analytics
• Explore how the right value exchange can enrich your
customer view
• Understand the consumer context of how you use customer
data
• Recognise and address human strengths and fallibilities in
big data analytics
• Involve marketers and insight professionals as part of core
team