The Knowledge Economy
Setting up for Analytics
Delivering Pizza Before the Phone Rings What is Customer Analytics?
What Can Customer Analytics Do?
Dr. Paul Bracewell
Understanding the Customer
9th November '06 Summary
Perspective The Knowledge Economy
Statistician By-product of the knowledge economy is lots
Empirical Methodologies (data driven, simulation, and lots of data.
approximation – close enough is good enough) Data is the currency of the knowledge economy
Interpretation is key (business context) How to cash in?
Comes down to how the data is processed.
Focus is fully utilising the data (valuable business
asset – numbers don’t have a political agenda) Just like any rush, there are plenty of people
jumping on the band wagon (think of students at
Pizza – fast food – not a lot of substance… Pizza-Hut, all-you-can-eat)
What is Analytics? Understanding Analytics
Customer Analytics is not extracting customer numbers To understand customer analytics, requires a
– that is simple data extraction and reporting (slicing & glimpse behind the scenes.
dicing – how pizza is prepared)
It is understanding Customer Behaviour
Purpose of Analytics
Think of customer analytics as the interest to be gained
from your currency
Importantly – no interest is yielded by stuffing your
investment under the mattress…
Slicing and Dicing is more like counting your investment
Before Analysis Data Collection
Purpose of Data Every phone call,
Data Context Every purchase,
Access to Data Every search,
Cleaning/Manipulation (Known Data Problems?)
Make sure the project isn’t set up to fail (it may generates data and is often
sound as if customer analytics can deliver the world, but an expression of behaviour
without preparation, it won’t come close) understanding that behaviour
is your business advantage
Data Storage Data Manipulation
Out of Sight, Out of Mind: Like that piece of pizza left under the couch
Cheap storage means that lots and lots of data after a night with the lads, data manipulation can
can be collected and stored without causing too often be forgotten.
many problems. With vision, data manipulation can be
Often this means data is stored in warehouses minimised for future projects (modelling mart)
with no doors, no windows.
Even in an air-tight container, pizza goes stale
Data Types What is Analytics?
Long Thin (one row per event)
Short Fat (one row per entity) More than simple querying and reporting,
Customer Analytics is the use of statistical
Most work in preparing data for analyses is techniques to identify pertinent information.
converting long thin data to short fat data.
Sponsors often neglect this important step
Where are the results? Why are you taking so long?
What is Analytics (2) Effective Analytics
Customer Analytics enables the following sorts of Customer Analytics requires five key elements:
questions to be answered – Someone who knows the data (domain expert)
Who will respond to a campaign? Someone who knows how to analyse the data (analyst)
How many calls will a call-centre get?
Someone who knows how to interpret the results in the
What is an acceptable waiting time at a call-centre?
context of the problem (end user)
What flavour of pizza will customers want?
What are the triggers of customer churn?
Someone who can sell the results to the business (sponsor)
NOTE: Future Tense – prediction (pro-activity) Tools
– using rigorously tested statistical techniques (need to Find the useful information and get that out to the
look beyond the cheese) people who need to use it.
Two Broad Types of Modelling – Predictive and Descriptive The task is to turn databases into sentences
Using Analytics to Make Decisions Making Robust Decisions
Statistical techniques are there to make Is there a real difference between
working with numbers easier – show us $1677.58 and $2268.17?
when to look and where. Consider the distribution of the data
Humans are poor at detecting patterns in
large data sets. Correct use of statistics
avoids mistakes and bias.
Understanding Customer Behaviour Role of Slicing and Dicing
Rugby, Pizza and Beer
1. Customer Analytics
2. Slicing and Dicing
Bracewell, P.J. (2002). Implementing Statistics in a
Diagnostic Coaching Structure. Research Letters in
the Information and Mathematical Sciences, 3, pp. 79-84.
Fraud Fraudulent Behaviour
New Zealand Insurance companies loose $150 With a credit card, you are responsible for repaying the
debt. A fraudster is not responsible for the debt. Any
million to vehicle fraud each year. They estimate funds lost to fraud are recouped in bank fees and interest.
that only 10% is picked up… A nice example is a type of fraud called a BIN attack
(BIN = Bank Identification Number)
You pick up the rest in your premiums. Look at your credit card – 16 digits or if AMEX 15 (Visa
For fraud to be profitable, it has to be abnormal. starts with 4999, Mastercard 5403)
Software on Internet allows #’s to be generated. Fraudster
Customer analytics can detect abnormal takes the generated #’s and tests them at various websites -
typically for purchases less than $30 to find genuine cards.
behaviour Then those genuine cards are hit – without the card holder
Fortunately, with the use of customer analytics BIN attacks
can be identified.
Minimising Workload Acquisition Risk Scorecard
Potential Bottleneck if large number of cases
Improved hit rates
Increase value for Money of Investigation
Individuals Involved Cost of Fraud
Marketing Interpreting Results
Another similar field is marketing – all about understanding Intention of the data must not be neglected
Selecting customers for campaigns. Recent study on the Results are only as good as the:-
North Shore at reducing junk mail showed that only 2% of 1. quality of data
people read junk mail.
2. understanding of data and process
Better targeting, better response rates, better profits
Listening to the customer – actions speak louder than 3. “translation” to audience
words – this can be used to create campaigns. 4. audience uptake
Work out which customers are likely to leave, but also
establish why they are going to leave so measures/offers
can be put in place to keep those customers.
Information Simplication Marketing Automation
Intended audience Once these ideas/thinking have been embraced,
Ease of communication the next logical step is further improvement of
Pin-pointing areas of “real” interest process…
Reduction of work-load …this improvement stems from ease of
Solid platform for acting on insight
…paving the way for solutions like Marketing
Greater understanding Automation.
Summary (1) Summary (2)
Those organisations which are open to finding Insight leads to innovation only if action is
previously unknown but useful information are well
placed to take advantage of the wealth of data in the
modern business environment Action can only occur if insights are robust and
The power of customer analytics is only applicable practical.
when an organisation is receptive to the underlying
principles of this process Insights are only robust and practical if suitable
tools are used, right personnel are involved and
“It is hard to look at both sides of the story when there is frequent and open dialogue between all
you are coming from one perspective” parties.