This presentation was given at the festival of marketing 2014. How grown up is your analytics? This slide deck will help you understand what you need to achieve optimum business benefit from your data analytics.
2. The Data Party
2
2012
http://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-century/
3. The Data Party is Over
3
2012
2014
http://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-century/
https://gigaom.com/2014/03/19/the-sexy-era-of-big-data-is-over-as-vcs-turn-to-fund-unexpected-industries/
4. What Does Good Look Like?
Insight
4
Data
01
Action
02 03
Turning data into
insight and
action?
Transforming
data into
actionable
insight?
5. The Reality: Lots of Moving Parts
5
One size does not fit all
Pricing
Customers
Competitors
Business Plan
People
Skills
Tools
Platforms
KPIs
Data
Structure
6. Maturity Is Not One-Dimensional
6
Data
Reporting
Analysis
Optimisation
Personalisation
7. Risk of Box Ticking
7
Data
Reporting
Analysis
Optimisation
Personalisation
Got some
data?
Is the data
in a report?
Hired an
analyst?
Done a
test?
Switched on
personalisation?
8. Risk of Oversimplification
8
Data
Optimisation requires
granular data
Reporting
Analysis
Optimisation
Personalisation
Reporting requires
high-level data
9. MVT
Visualisation
Analytics Maturity
9
Time to Grow Up
Modelling
Big Data
Predictive
Business
Intelligence
Test & Learn
DMP
Streaming
AVOIDING THE HYPE
SPEED, SCALE & “SHINY”
ACHIEVING BALANCE
DATA VERSUS DECISIONS
01
02
03
PRIORITISING
SKILLS, TOOLS AND PROCESSES
10. MVT
Visualisation
Analytics Maturity
10
Time to Grow Up
Modelling
Big Data
Predictive
Business
Intelligence
Test & Learn
DMP
Streaming
AVOIDING THE HYPE
SPEED, SCALE & “SHINY”
ACHIEVING BALANCE
DATA VERSUS DECISIONS
01
02
03
PRIORITISING
SKILLS, TOOLS AND PROCESSES
11. Avoid The Hype 1: Speed
11
Speed is not in itself an outcome.
Humans rarely make (good) decisions in real-time.
Speed often arises only from
automation, which comes after
analysis.
The need for speed becomes a
barrier to investing in longer term
analytics with much higher ROI.
12. Avoid The Hype 2: Scale
12
Rubbish In = Rubbish Out
Applies at every size of data
Critical success factors are the
same regardless of size:
Massive
data
Small
Data
Relevancy
Accuracy
Coverage
Structure Rubbish
Medium
data
Big
Data
14. MVT
Visualisation
Analytics Maturity
14
Time to Grow Up
Modelling
Big Data
Predictive
Business
Intelligence
Test & Learn
DMP
Streaming
AVOIDING THE HYPE
SPEED, SCALE & “SHINY”
ACHIEVING BALANCE
DATA VERSUS DECISIONS
01
02
03
PRIORITISING
SKILLS, TOOLS AND PROCESSES
15. Balance is Key
Analytics maturity is two-dimensional
Constant trade-off between:
Sophistication and availability of data
Capacity to make effective decisions
15
Data
Maturity
Decisions
16. Data Famine
Trying to run sophisticated models and make extremely granular
automated decisions…
… without investing in getting decent data to feed the process.
16
Data
Maturity
Decisions
17. Drowning in Data
Overinvestment in data capture and
systems integration…
…lack of capacity or understanding to
respond to the data flows
17
Data
Maturity
Decisions
18. Maturity Model
18
Understanding
Benchmarking
Educated Guesswork
Competitive
Advantage
Optimisation
Data
Decisions
19. Maturity Model
19
Understanding
Benchmarking
Educated Guesswork
Competitive
Advantage
Optimisation
Data
Decisions
Ad Hoc Capture Measurement Design Real-Time Flows
20. Maturity Model
20
Understanding
Benchmarking
Educated Guesswork
Competitive
Advantage
Optimisation
Ad Hoc Capture Measurement Design Real-Time Flows
Automation
Test & Learn
Statistical Analysis
Business Intelligence
Basic Reporting
Data
Decisions
21. Maturity Model Example
Measurement Design 21
Statistical Analysis
Data
Decisions
Understanding early
behavioural triggers to
purchase
e.g. a key signal of being
in the market for a
holiday is performing
multiple searches with
different dates on travel
sites
22. Maturity Model Example
Optimisation with critical link
between analytics
(segmentation) and testing
(variant) data
e.g. quickly understanding
how a test performs for
different audience/customer
segments or varies across
devices before it is too late
22
Real-Time Flows
Test & Learn
Data
Decisions
23. MVT
Visualisation
Analytics Maturity
23
Time to Grow Up
Modelling
Big Data
Predictive
Business
Intelligence
Test & Learn
DMP
Streaming
AVOIDING THE HYPE
SPEED, SCALE & “SHINY”
ACHIEVING BALANCE
DATA VERSUS DECISIONS
01
02
03
PRIORITISING
SKILLS, TOOLS AND PROCESSES
25. Analytics Nirvana?
What do you think is the optimal balance between technology, people
and process to achieve analytics nirvana?
2014 Econsultancy/Lynchpin Measurement & Analytics Survey
25
26. Refocusing the Debate
26
35%
45% ?
20%
What are the best
processes to invest in?
What are the right
technologies to buy?
Who are the right people
to hire?
27. Technology
Digital is increasingly Just Another Data
Source for visualisation and modelling tools
Success not always about having the best
tool
More often about using the right tool for the
right job
27
Technology
10
5
0
I q II q III q IV q
29. Process Determines ROI on Analytics
What are the objectives?
What is the most appropriate methodology?
Is there a genuine balance between:
Improving data quality/relevancy/accuracy
The capacity to respond to that data?
29
30. ROI Comparison
30
Digital Analytics Tool
Re-platform
return
0%
Rare to break-even in first
year once cost of change is
taken into account
Customer Retention
Model
return
+25-30%
Average 25% - 35%
performance improvement
from targeting customers
that are likely to lapse
Attribution Project
return
3-6%
Average 3 - 6%
performance improvement
from evidence based
reporting and some budget
reallocation
31. People: Critical Skills
• Strategy
– Objective setting is 20% of the time but 80% of the importance of
most analytics projects.
31
• Engineering
– Measurement, data flows and databases needs to be
designed, not hacked together.
• Analysis
– Understanding the business model is as important
as understanding a statistical model.
32. Critical Skills
Analysis
Strategy
Ad Hoc Capture Measurement Design Real-Time Flows 32
Automation
Test & Learn
Statistical Analysis
Business Intelligence
Basic Reporting Engineering
33. Analysis (is not) Magic
33
Experience plays a massive part in the process
Finding new
avenues of
exploration
Knowing how far to
go down an avenue
before pulling back
Fitting data to
purpose and
questions
Preparing and
manipulating data
so it’s in the right
shape
34. 5 Take Home Points
02 03 04 05
Beware false
dawns of speed,
scale and
shininess
34
Analytics Maturity
01
Process
determines ROI:
choose the right
analytics
projects wisely
Maturity is not
one-dimensional:
one size does
not fit all
Balance
between data
and decisions is
vital for analytics
success
Invest in the
critical
engineering,
analysis and
consulting skills
35. 35
Understanding
Benchmarking
Educated Guesswork
Competitive
Advantage
Optimisation
Ad Hoc Capture Measurement Design Real-Time Flows
Automation
Test & Learn
Statistical Analysis
Business Intelligence
Basic Reporting
Data
Decisions
Editor's Notes
This is the purest analytics presentation you’ll see all day: it’s essentially an entire presentation about one graph.
But, first, to get a sense of the room: how many people here have analytics as the primary focus of their role?
And how many people got into analytics because of this article?
Well: bad news.
Apparently the data being sexy phase has lasted less than 2 years. Poor Tulisa and Amy Willerton have had to abort their data science careers in favour of the X Factor and I’m a Celebrity Get Me out of Here.
So if – tragically - sex appeal is no longer a benchmark for analytics and we can’t do a line-up of beautiful data scientists, what does “good” analytics look like?
The classic definitions are solid, but fairly open to interpretation. “Turn” and “transform” are fairly broad metaphor here for a lot of things. Over the past 10 years we’ve seen some good and bad transforming!
And the reason there’s not a straightforward benchmark is that even within industries, e.g. if I think about all the different financial services organisations we work with, the business stage/strategy/objectives and context around teams, tools and platforms is highly diverse.
So the classic maturity model looks great; but is really hard to actually use (not very actionable). The words are very open to interpretation. Is there anyone that doesn’t do some reporting? Would any business really ‘fess up to having never done any analysis?
Think of the best decision you made. Was it the fastest?
Now thing about the decision you most regret? You get the picture.
Speed is often really about automation. Not about feeding real-time dashboards into the brains of superhumans.
I’m hoping I’m allowed to say that it really doesn’t matter what size the data is.
Structure is a critical success factor. Big data is technically unstructured.
There are lots of shiny buzzwords in analytics. Based on The Digitals last night you could have composed a pretty good awards entry by stringing all of these together with a decent twitter hashtag… but that doesn’t automatically make them the most useful analytical tools or techniques.
So let’s acknowledge that it’s more complex than a simple one-dimensional path.
But it’s not that complex, and the key unacknowledged issue is that data is tricky to manage, analytics is tricky to resource and there’s a balance required here.
So, here’s the graph. Constant balance. More sophisticated the analysis, the more sophisticated data you often have to feed it.
Decent data is not about it existing. E.g. ASOS have had decent data from scratch, but before they had a critical mass of customers that wasn’t in itself an asset. Major banks have loads of customer data, but it’s not necessarily all easy to access.
Often a classic symptom of this is lack of ROI from technology solutions.
So, the graph.
Ad Hoc: It’s there. Or it’s not. And it might be right. Or it might not. Good place to be if you like surprises!
Measurement Design: examples, mobile app, customer service calls (cancellations).
Real-Time Flows: it happens in one system and is automatically available in another one as fast as you need to make the decision (real-time is often an aspiration rather than a requirement).