In the great attribution debate of display advertising, it often feels as though we have click-through metrics in one corner and view-through in another. While many polarizing articles, with tag lines such as “Kill the Click,” have been written to shine light on the need for other conversion metrics, click-through metrics remain alive and well. So, we wondered, who really wins in a cage-match battle?
Click-Through
According to a March 2009 comScore study, only 16% of Internet users clicked on at least one display ad. This was down from the 32% of clickers recorded in a similar 2007 study. The study also notes that “Heavy and Moderate” clickers, who make up only 8% of the entire Internet population, account for 85% of all clicks.
We tried to figure out what these findings mean for us here at Triggit. By digging into three months of data from a retail client, we measured the relationship between click-through rate and conversion per thousand impressions for over a thousand different creative units and multiple levels of data targeting. We saw what many before us have seen: no sign of positive correlation between clicks and conversions. By plotting the click-through rate of each of these individual ad units against the conversions per thousand impressions per each ad unit (using a 15 day view-through attribution), we can easily see the (lack) of a relationship between these two variables.
The graph produced is the archetypal “scatter shot” graph. Put simply, it looks as if someone took a thousand darts and threw them randomly at a board. The line of best fit is fairly flat and actually has a slightly negative slope. A correlation analysis between the two variables and a regression analysis also controlling for variables like ad size, eCPM, user level (previous customer, etc.), and above the fold targeting, claim a negative but statistically insignificant relationship between click-through rate and conversions per thousand impressions.
View-Through
Based off our above graph, we can expect that when moving from a 15-day to a 30-day attribution window, this lack of a statistical significance surrounding click conversions only becomes more pronounced (the p-value jumps from .23 to .32; statistical significance is usually declared at .05 or .01). This is where view-through metrics become truly important.
In the same March 2009 comScore study, researchers note that “exposed non clickers” are more likely to visit advertised sites in subsequent months than clickers. Users who view an ad but do not click on it are still statistically significant, but the evidence of their conversion is more difficult to measure, as it doesn’t occur with the same immediacy. This knowledge, paired with our 15-30 day attribution window for click measurement, helps to illustrate that time plays a significant part in determining conversions. And, when we factor in research such as that from Madan Bharadwaj Visual IQ article, which shows that view-through accounts for 99.87% of all conversions in display advertising, it’s clear that view-through indeed does matter. A lot.
Conclusion
Although the click is a bad indicator of conversion activity within a 15 or 30-day window, it is NOT a completely useless metric in display. The key to understanding the utility of the click in display advertising is removing the thought that it is a good measure of long term conversion activity and accepting the click for what it is: an indicator of immediate intent and interest generation.
For everything else—conversion intent, brand awareness, and long term conversion activity—view-through is key.
Read the full comScore study or the Visual IQ article
Also read: The Data Behind “Bigger is Better”