Predictive Modeling And Your Inbox

We started this year’s annual winter Email Insider Summit with the familiar "Bring Out Your Dead" bit from “Monty Python and The Holy Grail,” because I can think of no channel in digital media that has been so identified with a single movie line as email has with “I’m not dead yet.” For most of the time I have been covering digital media (1993), the enormous impact of email has always been generally underappreciated and expected to dwindle soon anyway. Challenged by user exhaustion, spam or preemption by social, chat, IM, generational shifts in media habits, or now mobile messaging, email has, amazingly retained its potency against obvious competition.

And yet email remains undeniably unrivaled when it comes to ROI. Custora's analysis of 100 retailers showed that 23.9% of all sales over the Thanksgiving to CyberMonday weekend were driven by email. Message volume was up 33%. Retailers on average were sending one email a day between Thanksgiving and Cyber Monday. Walmart, Saks, Estee Lauder,  Overstock, and HSN all sent over ten messages over the long holiday shopping orgy.



More than a decade old, email remains ROI king. The DMA contends that every dollar spent on the platform generated $42 in sales for a 4300% ROI. And yet like the rest of digital media, the real challenge for such a successful platform is integrating into other digital and offline channels. We are still in our infancy in knowing how, when and what sorts of messaging people really want or need in the context of their journey to buy something.

At the Summit we started with a panel of email veterans offering some of their latest tips.  Not surprisingly, there was a lot of talk about using predictive modeling and data on the back end to make messaging a lot more personalized, targeted to specific effects and business goals.

Mark Bloom, vice president of business development and marketing at ShermanTravel Media, says his brands are tracking all user behaviors on the site to target interests by segment -- but just recently started using these online behaviors to trigger emails specific to how someone was interacting with the content. “Out of the gate, we saw incredible open and click rates,” he said.

The company's next project is to use more predictive modeling (via the Salesforce platform) to track user opt-outs. “The technology will be able to evaluate on a lookalike basis which of our users are likely to opt out, and [we can] change the messaging or suppress frequency or be more aggressive with those not likely to opt out."

Using email to reduce churn was another theme. Sue Cho, email marketing manager at The Honest Company, is using predictive modeling that she says is within 80% accuracy of forecasting customer cancellations. That is the time to make a special offer to get these fence-sitters to stick.

Both Amazon and eBay have among the most sophisticated email programs in the world, and they have been leveraging data to personalize the experience for years. But Amazon Marketing Manager Vicky Ge warned against reading the data too broadly: “When people analyze behaviors, they tend to fall into the law of averages and make decisions based on an average.” She suggested instead looking at the volatility within a data set to identify the strongest segments to focus on rather than aiming for the middle. “If you have people engaging at two different rates, don’t send to the median point.”

eBay’s Senior Marketing Manager for Specialty Retail Rishi Mahalaha discussing using propensity models to parse audiences. “We saw a 10x lift by segmenting people by price point for high priced or low priced phones," he said. The company is expanding this experiment into other categories.

One of the themes so far in this Summit appears to be the greater integration of email with richer data and reporting to fine-tune messaging. Until recently, email marketers tended to rely on blunt instruments -– opt-outs, CTR, time since last response, etc. –- to “know” the customer. But data gathered from a number of other behaviors across that user’s many interactions with a brand can be extrapolated by predictive analytics into models of future propensities.

As we have seen so many times in the discussion about leveraging data, this is science in the service of humanizing brand relationships with users. In person-to-person live communication, we use a series of subtle visual cues to adjust our end of a conversation according to whether our partner seems interested, bored, distracted, ready to bail. Data gives the email marketer the opportunity to mimic face-to-face communication by understanding consumers' subtler responses to messaging.

Next story loading loading..