Why Google Is Looking Beyond The Last Click
In recent years, Google has invested heavily in building attribution functionality for search across our platforms, most recently launching the Attribution Modelling tool for Google Analytics Premium. This focus on attribution might seem odd given Google's traditional strength in direct performance marketing. So why have we done it?
The answer lies in our desire to help advertisers unlock the hidden value in those earlier clicks.
Back in 2010, shortly after launching Search Funnels - a tool that helps advertisers to understand the entire search path leading up to a conversion, including search ad clicks and impressions - a team of analysts at Google conducted a large multi-advertiser study looking at advertisers who increased spend on generic search terms from one quarter to the next. The study found that subsequent growth in brand search, and not generic search, accounted for the greater proportion of new conversions. So where did this growth come from? It turned out that almost all of these new brand search conversions had been 'assisted' by generic clicks earlier in the search process. Brand term conversions which were preceded by generic clicks grew at a rate which was many times greater than that of brand only terms.
Furthermore we found that this effect grew over time - earlier generic searches later led to brand-related queries, and new customers were acquired who later became repeat customers.
This analysis shows the critical importance of looking beyond the last click. Customers may make purchase choices at varying points during the search process, and therefore you need to ensure that your search advertising is effective before they have made these choices and not afterwards.
We are not saying that last click is always wrong. Quite simply there is no right or wrong marketing attribution model. However, some models are more likely to be relevant to your business than others.
If your marketing aim is to defend market share by retaining existing customers, then optimising to last click may be a good strategy for you. For advertisers who want to grow their business by influencing customers who are new to their brand or product, and who are shopping around, then it might make more sense to value the earlier clicks and click assists which generate leads.
We also believe that attribution modelling cannot ultimately solve the problem of exactly how your ad spend causes sales to grow. Attribution is one way to develop hypotheses about growth, but testing these hypotheses is a whole different problem. It requires a systematic approach that accounts for the limitations of attribution and which measures in aggregate how new marketing strategies can drive new sales. Simple control / treatment testing plans are a good way to determine the true incremental impact of your spend.
Huge challenges remain for attribution. A key issue is that advertisers cannot see the role of earlier clicks if they are not investing in them. When budgets for generic search are tightly capped, advertisers should not be surprised if their data shows lots of single-click brand conversions. The early clicks being missed out and won by competitors remain invisible to the advertiser.
Another major challenge for attribution, lies around the growing phenomenon of multiple devices and sessions. In addition, many advertisers do not factor in the hidden value of search for customers who research online and then purchase offline (ROPO).
It's hard to determine the precise effect of these measurement challenges on each advertiser, but we do know the direction of the bias: they all work to make user conversion journeys look shorter and less complex than they really are. Once again it's those long-undervalued earlier clicks that are most likely to be left out of an ROI analysis. So for now, our message to advertisers is simple: make sure you are not falling into this trap and ensure that you find ways to measure all of your digital marketing activities beyond the last click.
Read more on the attribution and buying cycle here