|
Solution
After considering a number of
solutions including industry
heavy-weight SAS, Microsoft SQL
Server Analysis Services and open
source solution R, Trade Me decided
on a suite of solutions from 11Ants
Analytics consisting of 11Ants
Customer Response Analyzer, 11Ants
Customer Churn Analyzer and 11Ants
Model Builder.
When asked why 11Ants Analytics'
solutions made the cut, Trade Me
Analyst Suhel Mangera identified
several factors. "Deployment speed
was one - how quickly could we have
it all up and running out of the box
and in use? Self-sufficiency was
another - we wanted a tool we could
run in our Analytics team and not
have to rope in the Tech guys every
time we wanted to do a piece of
work. The third was accessibility - we needed something that didn't
require us to become total experts
in data mining and algorithms.
11Ants does the thinking for you."
Suhel adds: "The four step modelling
approach is very intuitive and easy
to use. It's
basically a zero learning curve,
which amazed us when we looked at
how complex many of the other
solutions were. All the others
seemed to demand steep learning and
more resources."
The culture at Trade Me is one of
making things better, and the
culture Megan is creating for the
Analytics team is all about being
proactive and taking ideas to the
business - rather than being in
response mode.
Megan says:
"We work in this strange
position where requests for work
come to us from all over Trade Me to
support their initiatives.
11Ants' solutions are good in that
they allow us to have a voice and
identify opportunities
for our teams that they haven't
thought of. We were coming from a
situation where we didn't have a
huge amount of resource, and we
didn't have extensive predictive
analytics knowledge so the solutions
have been very helpful. It's allowed
us to drive ideas of our own too".
Benefits
Rapidly, Trade Me put 11Ants
Customer Response Analyzer to work
with very satisfying results. The
initial initiative was to identify
which of Trade Me's users would be most interested in
Trade Me's 'Pay Now' service.
Suhel explains: "We wanted to use
this tool to model the existing Pay
Now sellers and their attributes.
Not just
demographics such as gender, but to
consider factors such as the type of
items they were selling, the values
of the transactions, what kind of
shipping options they offered, the
sell-through rate, how quickly items
were selling, number of listings per
week, and the number of items they
were actually selling. A simple
segmentation could
never accommodate something as
complicated as what we were after.
We wanted to build a model and then
apply it to our database and
identify other customers that
'looked like' existing Pay Now
customers - so we could provide them
with info about Pay Now.
"Megan says: "Community is key for
us - nothing is more important than
our relationship with our members.
It's dumb to promote new services to
those unlikely to be interested in
them, so we want to ensure we're
talking to the right people. The
11Ants model ensured we could nail
this aspect. "Those users identified
by the model responded a staggering
166% more than the control group.
Relevant information about existing
Pay Now customers was loaded into
11Ants Customer Response Analyzer
along with an equal number of
non-Pay Now customers. In order to
evaluate the effectiveness of the
model, a fully controlled experiment
was run.
Two groups of members were
selected - one at random, and one
consisting of those "most likely" to
respond (as identified by the
model). Both groups were sent
identical information about Pay Now
and given the
opportunity to sign up. Members
identified via the model responded a
staggering 166% more than the
control group! This is a great
example of predictive analytics at
work, both in terms of slashing
marketing costs, but even more
importantly of helping to shield
customers from irrelevant
information. |
