Friday, February 13, 2009
Of all the ticklish challenges being faced by digital marketers today, especially those buying search keyword exposure, one of the recurring points of debate appears to be the attribution problem. As more and more seasoned offline and online marketers compare notes in what is an increasingly cooperative effort, they all lament the fact that the "last keyword search" typically gets 100% of the credit for a sale or lead.
While Yahoo has tinkered with a means of crediting prior keywords with "assists," we're still far away from closing the loop on this issue. It will never be fully closed, of course. What influences me to purchase cannot be a 100%, slam dunk, exact science... and I hope it never is.
The problem is: once we accept that our attribution models aren't perfect, we're left stranded on Assertion Island. Seasoned offline agency types will simply asssert that offline ads are vital for "demand creation" and that search is "well down the funnel." The latter is by and large true. As for the former, sometimes offline ads are good at demand creation, but at what cost? How much is enough? Too much? Shouldn't the parts further down the funnel, the ones that don't fumble, and indeed "close the deal on," nascent awareness and demand, get more than their apparent share of credit for that ability to take the pass and put the puck in the net? And don't some kinds of demand and awareness "creation" actually do more harm than good?
Sticking to what we can know more about, you wonder who will come out and leapfrog Yahoo's basic assist measurement functionality, towards offering a product that attempts to estimate the contribution of a variety of media buys, and a variety of keywords, on a purchaser's ultimate decision. I suppose much of that would require significantly more acceptance by web users of the practice of being "followed around" throughout their entire online lives. (Panel studies that do look at "cross-attribution" issues do follow people around, as they volunteer to be followed. But that's just a sort of dog and pony show to attempt to prove a certain point about the power of different kinds of campaigns, with a relatively limited sample size. The real sticking points lie in getting past that point towards following nearly all users more closely. On that issue I am neither pro or con, or maybe I am con.)
And even then, being an industry where many entrenched interests are at stake, alternative scientific hypotheses are unlikely to be fully explored. Causation is subtle. If apparent causation is really just a spurious correlation, is the seller of weakly-effective media exposure likely to admit that? Even sellers of strongly effective media are likely to arrive at an entente with their less-effective brethren, in a compromise type of pitch that essentially says: "It's all good." (But is it?)
What spurious correlation, you ask? I'm still wrapping my head around the data showing that online display ads create a long term lift in sales volume, especially in combination with search campaigns, etc. But how much can we attribute that lift to the advertising, and how much to the fact that the ads show to users who are signaling a prior interest by visiting appropriate content sites? If we're attributing causation to impressions, as opposed to clicks (which are more reliable), what about the impressions of the written editorial contents of the visited sites? What about the impressions of written reviews, comments by friends behind the social network's wall, the very presence of like-minded friends in that milieu? The result -- more sales of goods in that channel to those people -- is surely not purely attributable to the ad impressions as opposed to participation in that milieu and digestion of (unpaid) content in that niche. Seeing a disproportionate lift in sales of one brand of (say) jeans or cameras over other brands not showing ads there would be stronger proof of that causation, of course. Unless the content, sites, or networks were expressly devoted to those brands (SonyLoversForums.com), which again nullifies any attempt to measure the ads' independent impact. Ads on brand-focused sites that pull buyers away from sites devoted to other brands would be very strong proof of causation, IMHO (you see a Toyota ad on AudiLoversForum.com, and buy the Toyota, for example).
Perhaps the uncertainties associated with full-blown causation models in purchase behavior is why we tend to cling onto what is most certain: what ad or link did the searcher or online application user click on most proximally to their purchase?
No question about it: as a keyword search advertising guy, I'm much more impressed by goals than by assists. And assists from keywords are easier to credit than assists from other media, in terms of both extent and cost.
Let's see if any analytics vendors are willing to take a stab at improving what vendors like Yahoo have already contributed to so-called "conversion tracking." It sounds complex enough that it probably can't be expected to roll out in 2009.
View Posts by Category
Andrew's book, Winning Results With Google AdWords, (McGraw-Hill, 2nd ed.), is still helping tens of thousands of advertisers cut through the noise and set a solid course for campaign ROI.
And for a glowing review of the pioneering 1st ed. of the book, check out this review, by none other than Google's Matt Cutts.
Posts from 2002 to 2010