By Simon Schofield, Analytics Lead, Crimtan
How does viewability impact on the conversion rate of digital advertising? And, as a result, affect ROI? Find out how we tested industry assumptions with one client – and showed them the true value of their digital display spend.
Here are two things we all know about viewability:
- Getting 100% viewable impressions is unfeasible.
- The higher the percentage viewability, the more the inventory costs.
This knowledge begs the question – how do you accurately determine the CPA for viewable impressions? And once you know this, there is one more piece of the puzzle that needs to be solved: what is the optimum viewability percentage that drives the best ROI?
We decided to test how viewability affected one client’s ROI
We wanted to find answers to these questions, and are lucky to have clients who are prepared to experiment and try new strategies to maximise ROI.
We felt there was a real opportunity to work with one particular client to establish how viewability affected the CPA. We were particularly keen to look at this because the client used the industry benchmark of 50% viewability as a reason to attribute just 20% of post-view conversions to us.
We felt that this was an underestimate – and undervalued the role digital display advertising plays in the marketing plan – so we wanted to demonstrate how viewability impacted on the conversion rate and how this affected ROI.
We began our test by including viewability as a KPI
To do this, we started to include viewability as a KPI, and compared performances across campaigns. We incorporated viewability, spend, seasonality and media type into our regression model to view the true impact each variable had on conversions.
As expected, a higher level of viewability does increase conversions, which led to us establish it as a hard KPI alongside CPA. From this, we adjusted inventory costs upwards to enable the higher bids required to secure more viewable impressions.
If you accept that an impression that isn’t seen has no impact on a conversion (it’s hard to argue against that!) the next step was to untangle the performance of viewable and non-viewable impressions and identify the conversions that happened only after a user saw the ad.
Next, we examined which factors affected viewability and conversions
To do this we examined viewability data and conversion data at a user and impression level – and went further to establish which strategies, domains and other factors affected viewability and conversion rate. We could then see which conversions came from viewable activity and review the client’s attribution model, which only assigned 20% of conversions to post-view.
While a simple ratio model clearly showed the uplift caused by more viewable impressions, it failed to consider spend levels. The ratio model was also too far removed from actual user activity, as focusing on the conversion rate doesn’t consider the role of frequency and the mixture of viewable and non-viewable impressions a user is likely to be delivered.
Finally, we experimented with how viewable impressions affected conversions
Modelling viewable attribution needs time to optimise towards the best results, just like trading does. So the final step was to experiment with the percentage of viewable impressions delivered and see how that affected conversions.
We could see that higher bids bought better viewability that, in turn, resulted in more conversions. But, as we pushed viewability higher, the excessive costs of viewable inventory reduced the number of impressions we were able to buy – and the increase in conversions began to plateau.
It’s important to note, though, that other factors besides viewability can have a significant impact on conversions. Contextual environment, time of day and many other factors affect conversions. For example, a high viewability site might attract non-relevant traffic and get few conversions, while a low-viewability site might attract just the right audience and convert well.
So the impact of viewability must be examined more deeply rather than used as an overall benchmark – and the optimum viewability will vary from one campaign to another depending on the main KPI and the product.
The results proved that 75% of conversions came from viewed impressions
Eventually, testing established the most effective viewability percentage target that produced the best ROI. And using only post-view and post-click conversions that resulted from users who were delivered a viewable impression, we were able to show that 75% of conversions were associated with viewed impressions.
This represents a three times reduction in CPA when compared to the original 20% attribution model.
When presented with these results, our client immediately recognised the significant part that display plays in reaching their marketing goals. This allowed them to more accurately allocate their budget and see an overall increase in ROI across their whole marketing strategy.
If you’d like our help with your digital advertising campaigns, please email jweeks@crimtan.com.