A Tale of Optimal Customer Targeting
Greger Ottosson 2700028ERR firstname.lastname@example.org | | Tags:  brms marketing predictive-analytics mathematical-optimization
0 Comments | 1,159 Visits
It's 8am, and we’re in the car. On the radio, an ad for our kids’ school is running and from the backseat my 5-year-old daughter wonders out loud “why are they telling us about my school when we already know about it?” She’s got a point. She’s been going there for two years now.
As we get a bit older, we accept that most one-to-many media channels, like radio and TV, have some inherent limitations when it comes to reaching us with the right message at the right time. They are inherently “one size fits all” channels because there’s no way for the broadcaster to know who’s tuned in and to tailor the message. In a more personal medium, the school ad could’ve skipped the pitch and instead reminded me about the Tuesday afternoon show-and-tell I signed up for or – more likely – passed the opportunity to sell me something relevant to another advertiser.
Other channels are more suitable for one-to-one marketing (e.g., Email, Mobile
and Web). However, it’s a familiar fact that the average B2C company has yet to
fully leverage the information sitting in its data warehouses to take advantage
of these new channels created by the Web and mobile revolutions. We still get
promotional mailings and calls about products we already have, don’t need or want. But some leading retail, banking and insurance companies are paving the
way for more personalized promotions and, during the last decade, have started
to apply analytics to optimize their product recommendation decisions among
Here’s an interesting example. A large
international retail bank needed to better communicate with its customers,
propose more targeted product offerings and – ultimately – improve the lifetime
value of each customer. To do so they deployed a new business process: Each
night a marketing plan is updated for each customer, detailing what communications
will occur in the next six weeks and what offerings will be proposed through
which channels. The plan ensures that all communications are relevant,
accessible and well-spaced; in other words, effective. Both online and offline channels
are managed in the same process. Online channels include in-person, internet
chat and the call center. Offline channels include email, snail mail and ATM ads.
How did they do it? In addition to the standard operational data stores and application servers, the bank is using a trio of emerging Decision Management technologies:
The whole decision process is deployed in two different modes – the nightly
batch mode described above and then an “attended” simulation and testing mode
where business analysts perform what-ifs to determine if there are ways to tune
business parameters and rules to improve the outcome of subsequent batch runs.
Did it work? After initial deployment in a
national bank with just under 10M customers, the solution is now rolling out across
additional international brands. So, while I still might hear the same radio ad
tomorrow morning, it’s good to know that millions of people are now getting
targeted product information instead of an offer for a credit card that they