One of every five beauty purchases online is made via the Amazon marketplace, according to a new report.
Julian Chu, operating partner at private-equity firm Castanea Partners, argues that retailers can drive growth by focusing on channels that generate sales most effectively.
"Attribution” refers to the method a retailer uses to allocate sales credit across different marketing channels and campaigns, and most online retailers continue to use “last-click” attribution as their primary way to assess marketing performance. This approach leads them to underinvest in those activities that will win new customers and drive long-term growth. What are the shortcomings of last-click attribution and what steps can the average online retailer take to unlock growth?
Digital marketing has always been based on exploiting customer data, starting with search engines. Yet, despite the fact that retailers are able to capture an immense breadth of data about online consumers’ behavior and preferences, most digital marketers remain relatively unsophisticated in using that data. Catalogers and direct response marketers have used holdout tests and sophisticated statistical models for many years, but such techniques are not in the tool kit of most small and medium-sized e-commerce companies, and even some larger ones. Many are still organized by channel and focused on optimizing these programs individually.
A recent survey conducted by the CMO Club found that nearly half (47%) of the respondents (the survey included several industries, not just retail) use last-touch attribution, and about 20% use first-touch. Only 18% use some form of multi-touch attribution, be it based on business rules or computer algorithms. And these are not tiny companies—more than 85% of respondents spend more than $5 million a year on digital marketing, and 40% spend more than $50 million annually.
What’s wrong with last-click?
If you only use last-click attribution, you’re only getting a partial picture. In particular, last-click favors “lower-funnel” activities such as branded search, retargeting and affiliates that are typically close to the conversion event. To a certain degree, you’re paying multiple times for transactions that are going to happen anyway.
Worse yet, relying on last-click makes it difficult to track the contribution of, and justify investments in, “upper-funnel” programs that expand brand awareness and acquire new customers, such as display advertising and non-brand search. The result is marketers end up favoring retention marketing programs and may even run out of productive ways to increase their marketing spend. I’ve seen numerous businesses where the budget for branded paid search is uncapped, and the return on ad spend looks great, but there simply isn’t any additional search volume to pursue.
By comparing figures based on different attribution methodologies, the bias inherent in each becomes clear. For example, the first chart above shows the revenue contribution of different channels for a Castanea portfolio company, using both last-click and first-click attribution. On a last-click basis, it appears that direct load is most important. But looking at things from a first-click perspective, natural search takes the top spot. Note how much more value is attributed to display advertising using the first-click method—over four times as much! This could make a big difference in the ROI calculation.
A little-known fact about Google Analytics
Many online retailers use Google Analytics (GA) as their primary analytics tool, but there’s an important aspect of GA that is not widely understood: Its default reporting is based on “last non-direct click” attribution. That means it is not actually presenting last-click data. If the last touch is direct load, but there is another marketing channel involved earlier in the consumer’s purchase path, the other channel will receive credit instead. The second chart is for the same company as above, but uses GA’s default reporting for the last-click figures. It tells a different story and may lead to different conclusions about the relative value of each channel.
The point is, the perspective matters. Online marketers shouldn’t just blindly accept the numbers produced by their analytics tools, but need to understand their inherent biases and look at channel performance in a more sophisticated way.
Take it one step at a time
Change is hard—I get it. There’s only so much time in the day for additional analysis. More sophisticated tools cost money. You may not have the right skills on the team, and senior executives glaze over when you talk about weighted versus linear versus time decay attribution. But this is not about throwing out your current tool set and starting over. Rather, it’s about layering on additional information to provide greater clarity about what’s happening in your business.
First, look at channel performance from at least two to three different perspectives. If you use Google Analytics, the Attribution Model Comparison Tool makes this very easy. You can toggle between a variety of attribution methods in seconds and see how the picture changes. You can even define custom attribution rules to better align your methodology with your business objectives.
One of these methods should be a multi-touch attribution model; this will help you understand which channels serve primarily to attract new users, which help to guide customers down the funnel, and which act to close the deal. You may even find that certain programs contribute little incremental value but merely ride others’ coattails.
Second, invest in a third-party attribution tool, preferably one that enables direct measurement of display ad views and how they influence downstream demand. (Yes, I’m talking about the much-pilloried “view-through conversion.”) We know ad impressions have value, but we often have no way of assessing their contribution. Such tools will enable you to measure the extent to which upper-funnel activities, such as display and non-brand search, help lead to increased brand search volume, e-mail click-throughs and site visits. Retailers with any material level of display ad spend ought to have an independent way of measuring its impact and not rely just on vendor reporting.