It’s hard to say that one attribution model is the “right” one. Each marketer has their own set of concerns and challenges, so vendors need to maintain room for flexibility. However, site retargeting is one of the few marketing channels where it’s actually quite straightforward to measure incremental lift, and too few marketers take advantage of this.
There’s been plenty written about the weaknesses of last click-based attribution models, so we’ll just touch on the topic briefly. With retargeting in particular, it’s dangerous to focus on last click-based attribution models since you’re only targeting people who’ve already been to your site. Many of these site visitors would return on their own if you did nothing, so measuring retargeting solely on clicks means you’re definitely paying for site visitors who would have come back on their own for free.
To properly account for this, advertisers can use a “hold out” control group to measure lift. This is a population of randomly selected visitors to an advertiser’s site that are not shown retargeting ads (ideally, they’re shown control PSA ads). This group can be compared to the behavior of a default group that does see the advertiser’s retargeting ads. It’s then fairly straightforward to measure the incremental conversion rate between the group that does get retargeted compared to the group that doesn’t. This ensures that you’re only measuring the true lift that retargeting delivers, and not counting customers (whether for clicks, views, engagements, etc.) who would have returned to your site on their own.
For this type of A/B test, the optimization techniques that drive more incremental conversions vs. lower cost-per-clicks (CPC) and last-click cost-per-action (CPAs) are quite different. Here are some tactics that focus on generating lift over shifting credit from other (possibly free) marketing channels.
Perhaps the most underutilized optimization lever within a retargeting campaign is based on time cohorts. We’ve found that with many advertisers, if they did no retargeting whatsoever, 80 to 90 percent of conversions would occur within five days of a person’s first visit to the site. If the focus were to simply drive the lowest last click CPA, it would be ideal to include retargeting efforts during this time period in the results in order to get credit for as many conversions as possible. However, this would probably be taking credit for a large number of conversions that would’ve happened anyway. If we’re optimizing for incremental lift, we have to explore other time cohorts and techniques to reengage dormant users and ensure retargeting generates enough incremental conversions to justify the media spend.
On-Site Behavioral Patterns
Who are the most likely visitors to click on an ad and convert once they’ve left your site? The shopping cart is probably the best place to start. When trying to optimize for the lowest CPC and lowest last click CPA, focusing on the people who put items in their cart is the lowest hanging fruit. However, aren’t these people also the most likely to come back and convert on their own? When optimizing for incremental lift instead of cheap clicks, retargeting vendors have to explore other intent signals and other areas of a site that might signal high potential customers. This could include specific products for an online merchant, or a specific feature page for a B2B company.
Creative Designed for Lift, Not Clicks
If you want to generate lift and incremental conversions, you need to inspire desire in consumers that are on the fence. It’s amazing how much this basic principal of advertising is forgotten when we strap our direct response hats on too tightly. Lift-oriented creative requires attractively designed units that reinforce a value proposition and feature the best possible product imagery and branding. On the other hand, if clicks and a low CPC are the only goal, something that looks more like a shopping widget might make more sense. You’ll want to maximize interest by showing a variety of products.
Retargeting often gets a bad rap for showing too many of the same or similar ads to website visitors. Using multiple vendors for the same cookie pool is often a culprit. However, this problem can be exacerbated by the attribution model. If a retargeting vendor is getting compensated and measured by clicks, it’s only natural that they err on the side of being in front of a user more times in order to get that click. If incremental lift is the goal, the vendor needs to tread more carefully. A lift-oriented strategy will focus more closely on the average number of impressions prior to conversion and prioritize exposure during that period, while tapering off after the point of diminishing returns.
Brands and their agencies have a variety of reasons for why they attribute different channels in different ways. However, with site retargeting, you can clearly measure the incremental lift it delivers, so why not take advantage of these techniques to optimize lift?