A Beginner’s Guide to E-Commerce Retargeting
All major retargeting companies globally have a programmatic buying platform, which takes real-time decisions on every impression. And almost all the players claim to have sophisticated big data algorithms that take real-time bidding decisions on users’ impressions.
Have you ever wondered how these algorithms work? What insights go into these massive platforms that can crunch through terabytes of data and pick the most valuable impressions from billions of ad requests?
Here are a few insights used in e-commerce retargeting and how advertisers are using them to drive performance.
More Recent Users Convert Faster
Recency refers to the user’s last visit to the advertiser website and it plays a big part in delivering return of investment (ROI). This is fairly intuitive. Users that visited the advertiser’s website recently are much more likely to convert than those that dropped off earlier.
The “earlier is better” argument is true only until a certain point, though. There’s an inflection point beyond which bidding based on recency will likely give worse results. Statisticians call this phenomenon the “inverted U-curve.” Malcolm Gladwell covers the concept extensively in his book David and Goliath, where he says that almost anything worthwhile in the world follows this curve — whether its wine, student class size, or exploiting recency!
To drive higher conversions, advertisers exploit recency at or near this inflection point consistently.
Recency as a variable should not be looked at in isolation. It’s strongly correlated with other variables described below. It is also often combined with the “depth” of the user’s visit — did it stop at the home page or did the user add a few products to the shopping cart?
Users That Frequently Research Are More Likely to Buy
Research frequency of the user refers to two things. First, how many times has the user visited the advertiser’s website? In website analytics lingo, this is also called a “session.” Secondly, across these sessions, how many products has the user seen?
Clearly, the more often a user visits the advertiser website, the more valuable the user. But that’s just half the puzzle. The key is to understand the number of sessions it takes before the user buys. This varies a lot from advertiser to advertiser. For instance, from our internal stats, a majority of baby products buyers often complete the purchase in the same session. Fashion products shoppers, on the other hand, visit the advertiser site more often before making up their minds.
Closely linked to session frequency is the number of products viewed by the user. Here again, there’s significant variance among different types of advertisers. Apparel buyers can easily look at a couple of dozen different products and their sizes, colors, and availability before they purchase. Cosmetic buyers visit three or four products — typically comparing brands, prices, and benefits for a cosmetic product they intend to buy.
Image source: Vizury
Users Buying High-Value Products Drive Better ROI
While it’s important to know how many products the user visited, it’s equally critical to know what these products are and consequently, how valuable is the user’s transaction likely to be? Taking the baby products website as an example again, targeting users that saw a low margin product like diapers may not help drive great returns, so going after users that checked out toys, which have higher margins, may be better.
One way in which advertisers can tap into this insight is by setting different prices for different categories of products. By doing so, they can better control investments by category and the overall ROI from retargeting.
The three insights explained above — recency, frequency, and expected monetary value of user — form a generic framework called the RFM model that is used to determine customer value not just in retargeting but across multiple industries.
Loyal Users Are Much More Likely to Transact Again
Users that have already bought before on an advertiser’s website are much more likely to transact again — since they would expect to have a familiar and predictable experience. Assuming, of course, that the price is comparable to that offered on other sites.
Advertisers often use non-PII information (typically encrypted CRM IDs) about previous buyers and use this information effectively in real time — ither to aggressively target the user next time or to upsell products after a recent transaction.
A related but exactly opposite example involves advertisers using the same data to focus their retargeting campaigns only on new users to lower costs of acquisition.
Smartly Mixing Product Recommendations Creates Better Engagement
The age-old saying associated with retargeting refers to “that pair of shoes which keeps following me around the Internet.”
While showing products the user has already seen is definitely helpful as a gentle reminder, it makes sense to keep refreshing the banners from time to time with other similar and relevant products. This prevents user fatigue, helps the user understand all options available, and drives higher engagement.
Another technique advertisers used is to have a set of products they want to proactively promote to users, usually because there’s a discount or sales promotion on them. These products are “mixed” with regular recommendations.
Some of the best retargeting companies consistently outperform others because of two reasons. They are able to work closely with advertisers to identify many such insights on a wide variety of e-commerce advertisers and apply rigorous mathematical models on large data sets to exploit these insights to the hilt.
Original article published on ClickZ on February 4, 2014