From the first click to the last: why attribution models matter
When you bought your coffee on the way to work this morning, why did you choose that particular brand? Was it a magazine ad, recommendation from friends, the posters near the station, or did you just happen to pass by and drop in? Online advertising has used ‘pass by and drop in’ as the standard attribution model for digital marketing for a long time. Referred to as ‘last click’, this method is simple to measure and report as it gives all the credit for the site visit or purchase to the last ad a user clicked on to bring them to your site. However, there is now a general consensus that it is unsuitable for measuring marketing performance and alternatives need to be considered.
Although there is still a place for last-click in a minority of campaigns, the majority of attribution models should be updated. Last-click focuses spend on the bottom of the sales funnel, on areas such as retargeting – where visitors to your site are sent repeated ads to try and encourage them to return – so more valuable opportunities further up the funnel, including prospecting for new customers, may get neglected. Last click encourages aggressive and off-putting retargeting as your retargeting partner throws everything at trying to acquire the final click before conversion.
There are other ways to measure campaign performance besides the last click – some of them based on the accepted view that simply seeing an ad has some branding benefit and plays a part in the conversion path for any subsequent sale:
Post view or post impression attribution - If 90% of your site traffic comes from searches that include the name of your brand, do you think that Google (or another search engine) should get the last-click credit? Clearly, they were already aware of your brand and just used the search engine to get to your site. If you also have retargeting strategies in place it is likely that the user would most likely have bought anyway – so better attribution ensures that credit for sales is attributed to the right place. Post view makes the assumption that any user served with your ad who subsequently arrives on your site is down to, in some part, the ad they were shown. Many more people visit your website than the tiny percentage of users that actually click on ads, so it makes sense that viewing the impression may have had some effect
First click attribution - This works on the assumption that the first ad a user clicked on gets the credit. They may well have left the site, been retargeted and returned at a later time, but it’s the first interaction that was most important. The big problem is, how do you know that they would have come back without being prompted? Surely the last click deserves some credit
Fractional attribution - This methodology works on the principle that it’s not just the striker who scores the goal who gets the credit, as it is a team effort in which everyone has played their part. So ad spend is attributed to various touch points along the path to conversion by splitting it across all points in the process in predefined proportions. The advantage of fractional attribution is that it is relatively simple to implement. Although it is an improvement on last-click, its biggest drawback is that the split is usually arbitrary, with little justification for how spend is attributed. For example, an attribution value of 20% for last click, 20% for first click, and 60% for the rest of the user journey could be used. While this attribution has its benefits, it can be rather arbitrary as it’s difficult to know what to award each stage.
Converter/non-converter analysis - This is a more scientific approach based on assessing the extent to which the conversion rate among a target group improves when a new provider is added to the plan. It uses Bayesian mathematics and game theory to establish the impact of each partner and attribute spend accordingly. Although this is a superior approach to fractional attribution, in that it rewards media that performs well, it is a complex process.
Sticking to a single attribution approach won’t work, as different models will suit different products and marketing objectives. The key consideration is the length of the conversion path and the average number of clicks, or impressions needed to make a sale.
In verticals with a large degree of consideration, such as retail or travel, there will be numerous touch points, which is why a converter/non-converter model would add real value. However, in verticals where there is a shorter conversion path, the ROI may not justify a converter/non-converter model, making fractional attribution more attractive.
The best attribution models work on the basis that spend should be assigned to each touch point in the conversion pathway, based upon the sale towards which it contributes. As a result, brands can shift the focus away from the bottom of the sales funnel and take advantage of opportunities further up the funnel, which could introduce a whole new group of customers and far greater levels of ROI for their marketing investment.