• by Chen Wang , September 28, 2007

In interactive media space, one of the biggest controversies falls in the area of viewthrough conversion attribution. For those who are not familiar with the dichotomy of conversion metrics, there are two types of conversions that are believed to be derived from online ad exposure: viewthrough conversion and clickthrough conversion.

Viewthrough conversions are often called post-impression conversions. They occur after a user is supposedly exposed to the ad and converts afterwards without actually clicking on the display. The other side of the coin is clickthrough conversion, where the conversion happens after exposure as well as a click on the ad. Marketers have long questioned the legitimacy of attributing viewthrough conversion back to advertising, arguing that there is no evidence that the user has even seen or paid attention to the advertising. Meanwhile, there seems to be a tacit agreement in the industry on clickthrough conversions. The conventional wisdom is that as long as a conversion happens after a click, that conversion should be credited back to ad exposure. A pertinent but almost never asked question is whether such glorious treatment of clickthrough conversion is really justified.

To answer this question, think of a real campaign you have run. Say that you have a month-long online campaign that has generated 5,000 clickthrough conversions. If we are crediting every single clickthrough conversion back to our online ad (as we would normally do), we are literally saying that if we hadn't run this campaign, NONE of the 5,000 conversions would have happened. Well, the word "NONE" has sure made it sound quite a lofty conclusion to make.

In fact, the only type of campaigns that can/should be credited with every single clickthrough conversion is run in a "contextual vacuum." What I mean by contextual vacuum is that the campaign is conducted in such an environment where the ONLY way a customer might land at conversion is through the campaign ads. Think of a product/service related campaign that is advertised via Web ONLY. No offline advertising, no direct mail, no PR blast is run concurrently with the online advertising. Furthermore, the promoted product/service does not have any memory effect from the past. Hence it has to be brand-new and unheard of before the campaign. In addition, there should not be a way a customer would be able to stumble into the conversion by him/herself. So a product listed at an online store that has a Web address that people can easily remember and type into their browser probably won't make the cut. Only when all these criteria are satisfied can we conclude that every clickthrough conversion is derived from the ad.

As marketers living in Web 2.0/3.0 space, we all know that the type of campaigns I just described are rare entities. A multichannel approach is the dominant force in today's advertising world. Most ad campaigns, online or offline, are embedded in a much larger marketing endeavor. It is very likely that your target audience is exposed to the same message via different media vehicles. Conversions (clickthrough or viewthrough) we have seen are partly a reflection of this trend. What this means to us is that some of the clickthrough conversions we have seen through the online channel would have happened anyway because some of the clickthrough converters, who are interested in our online ads are also likely to be exposed to or seek out the message through other means. Yes, without doubt the online campaign did add an extra round of exposure, but for those people the extra exposure would likely have happened through another round of TV or print exposure. Unless we can demonstrate that the extra round of exposure from other media channels is not possible in reality, it would be quite hard pressed to continue our current practice: to attribute ALL clickthrough conversions back to the online ad.

Empirically, there are telltale signs that support the hypothesis that not every clickthrough conversion is as valuable as we think. For the campaigns I have been investigating, I have always found instances of clickthrough conversions where the actual click is way ahead of the subsequent conversion. I am not just talking about hours, but days, sometimes weeks, between the time of click and time of conversion. Although the majority of the clickthrough conversions occur within a day after the ad exposure (the actual percentage varies by campaign, placement, creative, etc), the perennial existence of those far-out conversions does leave some trace of suspicion. It is a little hard to argue that conversions happening three weeks after the clicks are still 100% attributable to the original ad they saw. This is not to say that none of those conversions should be counted. I would agree that some of the creatives are quite memorable and/or some of the customers who clicked have great memory. But this is still far from the underlying assumption we are using in treating clickthrough conversions: EVERY creative is memorable and/or EVERY customer has great memory.

One possible challenge to my line of criticism is that maybe a few of the clickthrough conversions would have happened anyway. But that number has to be very small. Therefore, to ignore such effect probably isn't going to have material impact on the results we have been compiling.

My answer to that is that this might be true for some campaigns but certainly not true for all campaigns. I personally have seen campaigns with over 30% of the clickthrough conversions being labeled incidental. What is more significant is that without investigating this issue fully you will end up with suboptimal media plans. This is the case because the percentage of converting anyway over total clickthrough conversion may vary across sites, dayparts, and other media components. The metrics we have been using for media optimization are built upon our crediting 100% clickthrough conversions to the ad. So a site that has highest clickthrough conversion rate and lowest cost per conversion is our favorite. But what if we found out that this site generates a large percentage of conversions that would have happened anyway, much larger than other sites we have been using. By factoring out the incidentals, suddenly our No. 1 site is no longer our favorite. Whereas a site with a seemingly lower conversion rate may be something worth pursuing.

To recap, for the majority of online ad campaigns, we have been overstating the ad effectiveness by crediting every clickthrough conversion back to ad exposure. The fact is that there are some instances where such conversions would have happened anyway. To optimize media plans without taking these instances into consideration runs the risk of setting something up on a wrong footing. Therefore, a more constructive way of looking at the clickthrough conversion is to think in percentage terms. Remember, 100% is a dangerous thing to say.