In today’s digital age companies need to harness their
consumer data effectively to create and protect customer value. One
increasingly common approach to do so is the use of next best action modelling
to drive personalised marketing. Companies that use this type of modelling will
predict what the best thing for an individual customer might be, spanning the
gamut of potential possibilities—product, action, non-action—all whilst
accounting for the customer’s browsing or purchase behaviour.
If it sounds complicated, that’s because it is often notoriously
complex, which makes NBA a difficult thing to get right even for large
companies with many resources at their disposal.
TMAC specialises in increasing customer lifetime value using
data-driven marketing, so NBA algorithms and marketing campaigns are a common
delivery for our consultants. They often demonstrate that the companies that
can nail the right approach—which usually blends clever data modelling and
machine learning with bespoke marketing—consistently reap more value from their
customers.
The Challenge:
A FTSE 250 insurer had recently expanded their product
portfolio and invested in a NBA algorithm to embed the new products into the
market. Ultimately, the insurer wanted to reduce churn and increase market
share to maximise customer lifetime value. Insurance is an industry with
relatively low margins since there is often little direct control over price,
so once a customer is acquired, even on a low price, it was cost effective for
the insurer to cross-sell products where appropriate.
Unfortunately for our client, they went to a big four
consultancy firm to build out the solution and after 16 months, had no clear
indication of the benefit in terms of ROI versus statistically efficient
control groups.
TMAC’s work:
Our work was clear cut. Our client was 16 months overdue on
delivering value to the bottom line so we had to identify both the issue, solution
and build a quality alternative, quickly.
We split our work into the following streams and spent the
next 8 weeks building and deploying:
·
Data Processing and Selection: we looked at how
the data was processed and what data was being looked at: a red flag in the
former firm’s model was the lack of previous purchase data and use of different
marketing channels.
·
Rebuilding the model: predicting the likelihood
that a customer would buy a certain product based on who they are, their
previous behaviour and the channel they were most likely to buy from, whether
dotcom or outbound marketing.
·
Optimising outcomes for best profit: In our
experience, most companies that tried to do NBA get this step wrong with the
highest frequency. It isn’t enough to look only at probability; firms need to
account for profitability as well. A customer that is 80% likely to buy a
product worth £10 is not as valuable as a customer that is 60% likely to buy a
product worth £20, and marketing campaigns and overall level of effort can then
be adjusted accordingly to optimise marketing return on investment (MROI) and
profit.
·
Campaign and Campaign Selection: We ensured that
we created statistically robust control groups to ensure that there was no
skewing of the results. It was important given the context that the senior
leadership team understood that any new results could be fully trusted.
·
Training and Deployment: As we deployed the NBA
tool and marketing collateral, we ensured that call centre agents were given
the correct training on how to use the approach most effectively.
The results:
A whopping £750k in incremental revenue for the insurer;
and, since we deployed very quickly our client was able to reap these benefits
in year. Outer year benefits are still being realised!
Get in touch with TMAC here.