Breaking Free from Last-Click Attribution
SEM has a proven track record for delivering ROI by efficiently fulfilling demand and routing traffic precisely when users are actively looking for something to click on. Last-click attribution works for paid search because users are more likely to convert right after clicking an ad. Conversely, display and social marketing doesn’t require user-initiation and can generate demand by reaching potential customers at any time during regular internet browsing and content consumption. This enables many unique targeting tactics for finding prospects and staying top-of-mind, but makes measuring performance far less straightforward.
The true influence of display impressions is hard to determine because you cannot simply rely on metrics like CPC, CTR or post-click conversion rates. In fact, optimizing for clicks can be detrimental to the performance of your display campaigns. Simply put, most people don’t click on banner ads, and those that do aren’t likely to be your most valuable customers. But display ads are proven to lift brand recognition, boost branded search activity and increase consideration and intent to purchase. Display ads help drive conversions even if the user ultimately navigates to your website directly or through another channel. As such, post-impression (view-through) attribution is critical when evaluating the success of display campaigns.
Last-click attribution has remained the industry standard for a long time, but there are better, alternative options. Basic alternative attribution models include “Last Interaction,” “First Interaction,” “Linear,” “Position-Based,” and “Time Decay.” Using these as a starting point, you can go a step further and create your own custom conversion model based on your own set of assumptions, to adjust credit by interaction type or apply custom rules that fit your specific media mix and business goals. For example, using the Position-Based model as a starting point, you may want to adjust your allocations to give 30-percent credit to the first interaction, 50 percent to the last interaction, and 20-percent credit split between all middle interactions within the lookback window. Google’s most advanced algorithmic model, data-driven attribution (DDA), uses all available path data (from your floodlight configuration) — including data from both converting and non-converting users — to evaluate how particular touch points impact the probability of conversion and assign partial credit to each touch point. If you’re not ready to fully commit to an alternative attribution model, you can still compare how each model looks using the Attribution Modeling Tool in Campaign Manager.
Selecting the Right Conversion Window
Post-click and post-impression conversion windows are typically set at the advertiser level and will apply to all channels and tactics the same whether they’re display or search, prospecting or retargeting. In this case, you must select a conversion window that makes the most sense for your digital marketing efforts as whole. To do this, consider the purchase cycle for your product. Products like insurance or B2B enterprise software require more education and consideration than something like movie tickets and should have longer conversion windows. The “Path Length” tool in Campaign Manager can help you analyze how long it typically takes between ad exposure to conversion.
To take your attribution model to the next level, consider using “Campaign,” “Site” or “Placement” level overrides. These lookback window overrides allow you to shorten the post-click or post-impression window for certain audiences or partners on your media plan. For example, you may want to shorten the post-impression lookback window for your remarketing efforts but keep a longer lookback window for prospecting. How far back should you look? Historically, 30-to-90 days has been the standard adopted by many advertisers, but you have to consider the attention spans of human beings. Is your audience likely to remember your ad several weeks after exposure? The impact is likely to be strongest when the ad exposure is fresh in their minds. Consider shortening your post-impression window to one-to-seven days, which has become standard in some platforms including Facebook. A longer, seven-to-28-day window can be used for post-click attribution, as the ad interaction is very likely to have contributed to the conversion and leaves no doubt with regards to viewability.
Calculating Lift and Taking Viewability into Account
Viewability and view-through attribution go hand-in-hand. You can take your custom attribution model to the next level by weighing view-through conversions by viewability rates to factor in the probability that the ad being credited with the conversion was actually seen. Below is an illustrative example of this method applied to two display partners with very different viewability rates:
|Display Partner A||Display Partner B|
Adjusted Conversions = CT Conversions + VT Conversions x Viewability %
In this example, Partner A had a much higher CPM and was originally attributed with less VT conversions at a higher CPA, but ultimately achieved a lower CPA after adjusting for viewability. Failure to factor in viewability could leave you susceptible to overvaluing low-brow tactics such as “cookie-bombing” – heavily retargeting users with cheap, unviewable inventory at a high rate to cheat the system and steal credit from truly influential partners. For more information on viewability measurement, see my last blog post: “Viewability: The Importance of Adequate Measurement in Display Advertising.” To determine that your display impressions are truly driving incremental conversions rather than simply getting in front of users who were already going to convert anyway, consider running a conversion lift study. You can do this by serving PSA ads to a control group and comparing the conversion rates of your control vs. exposed groups.
Measuring View-Through Conversions for Social
For paid social and other ad types that don’t allow for third-party impression tracking, it’s unfortunately not possible to attribute post-impression conversions in Campaign Manager. It’s essential to use platform data to get the full picture of how these ads are performing. These conversions will not be de-duplicated from other digital channels, but you’ll still get an idea of how many conversions were influenced by social ads, and this will be essential to optimizing your paid social campaigns. Facebook has become an attribution platform and boasts “people-based” attribution as opposed to the cookie-based attribution used by Google’s Campaign Manager. Cookie-based attribution suffers from things like cookie deletion and incognito browsing, whereas Facebook user logins serve as more persistent identifiers for both targeting and attribution. This deterministic method is arguably more accurate than probabilistic models, especially when it comes to cross-device measurement.
Even with cookie syncing, device ID matching and other sophisticated ways of mapping multiple data points to an individual user, view-through attribution remains an imperfect and ever-evolving science, due to shifting regulations, technical limitations and the intangible aspects involved in human psychology and behavior. Nevertheless, advertisers cannot afford to ignore the impact of display and social impressions with or without clicks. Countless studies conducted over the past 10 years by organizations ranging from ComScore to Stanford University have consistently arrived at two basic conclusions:
- Approximately 999 of every 1,000 programmatic display impressions don’t get clicked.
- Somewhere around 8 percent of internet users account for 85 percent of display clicks.
- Up to 50 percent of mobile clicks could be accidental.
- Over 40 percent of all web traffic is composed of bots, half of which are bad.
- Consumers are more likely to conduct a branded search after being exposed to display ads.
- Site visits increase after ad exposure when compared to control groups that aren’t exposed (even if they don’t click).
- Display drives significant lift in conversions when users search after being exposed.
- Lift in search activity can last up to a week after exposure and can extend to friends and family of users exposed to display.