Learn more about leveraging predictive analytics for email and also hear social media strategy lessons from retailers like Macy’s, Harry’s, and One Kings Lane from our webinar co-host Percolate.
Predictive analytics allows quick, reliable conclusions about how a person will behave by inferring how similar they are to other people: Daya is a top customer who shops business chic shoes and loves free shipping. Olivia just signed up and has similar characteristics to Daya (came from Facebook, browses footwear, from the Midwest) — better send Olivia an email about shoes that come with free shipping.
Throughout a customer’s lifecycle, predictive analytics helps retailers better understand who their customers are and what they like, and then tailor messages to better suit those tastes — from activating email subscribers (sometimes called “members”) into customers, to growing relationships through repeat purchases and interactions, until the point of retention, when it looks like that customer might stop shopping with that store, and beyond.
Recently, we co-hosted a webinar (with Percolate) and shared stories of how predictive analytics is helping e-commerce leaders (like Guess, Nasty Gal, and Sole Society) transform their email marketing.
We covered case studies from each phase of the customer lifecycle:
Catch the entire webinar recording at the bottom of the page, or read on for highlights.
Activation: Convert more members into customers
Once someone signs up for a store’s email list, they will typically receive the same email or “welcome series” that everyone else gets and maybe a standard promotional offer. Perhaps that company has tested to see which welcome emails work better compared to each other, but that is typically about it.
Applying predictive analytics to demographic and behavioral data enables retailers to segment customers based on their likelihood of conversion. With these segments, they can then test different messages and offers to see which approach resonates best with each group.
Once it is determined which message is best for the different segments (cat vids forever) taking into account the cost of promotions vs. the expected revenue, teams can automate the most effective combinations. Sole Society made big improvements with this method, nearly doubling member to customer conversion rate and boosting combined revenue 3.5% — watch the webinar to learn more.
Growth: Send emails that match individual customers’ tastes
Email segmentation splits customers into groups based on behaviors, characteristics, and/or needs. This lets companies send these segments only emails they are interested in, not just the same thing to everyone (learn more from the segmentation course on Custora U).
The simplest approach to email segmentation relies on individual variables related to customer behavior and demographics, while more advanced segmentation looks to build defined customer “personas” that incorporate a variety of customer traits and product categories (e.g., “accessory shoppers” or “jeans and sweater lovers”). A big limitation of both of these approaches is that they are typically built historically, meaning they sort customers by products they have already purchased: After all, somebody who bought shoes in the past is probably receptive to an email featuring shoes.
A predictive approach, in contrast, identifies what product to recommend based on what a shopper is likely to buy next (and is able to incorporate a multitude of different variables, including acquisition path, age, order amount, style, etc.). A retailer with these insights might know that customers who buy certain types of shoes are also likely to purchase jewelry and pricier belts, and should be sent product recommendation emails accordingly. Guess recently started incorporating these types of predictive insights into their email campaigns (learn more).
Retention: Win back customers who are moving on
Eventually, a customer stops shopping at a store, but how can a retailer tell when that day has arrived?
Many retailers do so by coming up with a standard amount of time (e.g., sixty days) since last purchase, and if a customer goes that long without a purchase they are considered “churned” and sent a win-back message. This is a bedrock strategy for e-commerce marketers, and it is effective, but this blanket churn designation lets customers slip through the cracks and means retailers sometimes hand out unnecessary discounts. Take the two shoppers below:
By the time based measure one would think that Mike, the shopper on top, should get a “We miss you” email (with a discount) as it has been over 90 days since his last purchase. But in fact, Nora is the churning customer according to the predictive algorithms, as she is a more frequent shopper who is more likely to move on quickly and never look back. With a the time-based churn rule, Nora would have already started shopping elsewhere before receiving an email, while Mike may have been given a superfluous promotion as he’s just a less frequent, but still loyal, shopper.
Predictive analytics lets companies adopt a more personal approach to keeping customers, and move away from a more rigid method that only considers time since last purchase. Nasty Gal saw a 17% increase in customer re-activation and an 11% revenue lift by sending automated churn emails powered by predictive analytics: learn more.