Multichannel Measurement Might be Premature Optimization of Sorts

I regularly receive offers from recruitment agencies, and judging by the somewhat erratic job descriptions they send to me experience with multichannel tracking is considered a key qualification even in small or medium sized businesses. I have to say that, in my experience, you should really think twice if multichannel tracking will actually help you with your commercial endeavors.

The idea behind multichannel tracking is that, if you advertise for products on your website, the eventual buyers will not only have seen the ad they clicked on immediately before their purchase. Instead, so the theory, there will have been a build-up of sorts with multiple touchpoints, i.e. multiple ads seen, each of which will have contributed to convincing them to spend their hard earned money (or never mind the hard part, or the earned, as long as its money) on your merchandise. This is the kind of thing that sounds nice when told to you by a well dressed smooth talking sales assistant with armed with a persuasive prowess in powerpoint presentation. I do however have some issues with this theory.

One problem with multichannel tracking (admittedly one it shares with other kinds of tracking) is that it only ever records positives. For example this Saturday I have watched about 30 clips on Youtube on three different devices. This means that I have now seen the very same beer commercial some 27 times in one afternoon. After about the 17th time I was so fed up I secretly decided to sabotage the company by bottling my own bodily waste products and mixing it into crates with their product, and I only abstained from the plan because by take 23 I had had enough opportunity to study the texture of the beverage to conclude I didn’t need to bother (since somebody apparently had already done so). Multichannel tracking would be so much more useful if it could tell you after how many repeats your ad turns cheerful ale aficionados into misanthropic teetotalers.

Another problem is that multichannel measurement puts a long way between ad views and eventual conversion – e.g. in Google Analytics the suggested lookback window (the time during which touchpoints are considered relevant) is 30 days. So we measure the ad views and ad clicks during the last 30 days and from that we build a model to see what enticed the visitors to become a buyer. However during that 30 days the user also did a lot of other things (like eating, sleeping, working, loving, hating and generally exploring the ups and downs of the human condition) that are to 99,999% unrelated to your brand, so to derive┬ácausality from a few scattered data points seem just a tad┬áspurious.

Of course there are ways to make amends for both problems. The first can be solved by applying frequency capping, i.e. show the same ad for a limited number of times only, and common sense should be good enough to work out a number. For the second, well, we will have to test if the results are somehow significant by creating allocation models and test them against a comprehensive dataset, which brings us straight to the third issue.

For multichannel tracking to make sense you need complete data for views and clicks over all your marketing channels (after all that is what “multichannel tracking” – now sometimes marketed as “omnichannel tracking” – means). Web Analytics tools will not usually give you that. For example Google Analytics integrates nicely with Adwords and other Google Products, but does not a lot to provide view data from your display publishers, retargeting vendors etc. So you need an additional tool, which on a largish e-commerce website can easily cost you in the four figure range per month, plus a one time setup fee.

The main purpose of multichannel tracking is to help with budget allocation. Maybe you spend already more on your last click channels than can be converted into sales, and would be better advised to invest in the assisting channels that slowly gnaw away at your users patience slowly convince your audience that your merchandise is worth their money. Now, it is unlikely that up until now you have done everything wrong when setting up your marketing channels, so after deciding on a tool and setting it up and waiting for baseline data to flow in and selecting a reasonable allocation model you will probably end up shifting some ten or twenty percent of your budget between channels. And this budget shift will have to trigger enough many additional conversions to cover the cost of the tool and then some (although taking sales away from the competition might a reward in itself). So if you need to generate tens of thousands in additional profit (which depending on your margin means hundreds of thousands in revenue) by merely shifting marketing budgets you need a hell of a budget to start with. But even if you do not spend nearly enough for a significant budget shift there is a high probability that somebody tried to convince you of the merits of multichannel tracking. Why is that ?

As far as I can tell there is (in this respect) no fraud in the industry, so the main reason is that vendors are convinced that they bring something relevant to the table. And of course, since by now anybody will feel familiar with the word, “multichannel” is a door-opener that allows vendors to market their other products. In fact stand-alone solutions are by now rare or extinct, and the tracking capabilities are part of marketing automation suites (which is good because it means tight integration, and bad because tracking is no longer an independent control system). We implementation guys (and gals) like multichannel tracking because there is some seriously cool technology involved (and while of course we should work for the love of data cool technology is really where an implementers heart is). And finally there is the somewhat less-than-honorable reason: we cannot go around and sell you the same stuff year after year; we want a bit of growth for our industry, and that works only if we push ever new tools to our customers. For you this will be at bit of a vicious circle, because after six months (twelve months top) somebody will tell you that your multichannel thingie is not set up properly and don’t you want to try that new exciting tool that will remedy the problems you did not have before.

I do not actually want to discourage you from buying web analytics, multichannel tracking, marketing automation or anything else we try to sell you (after all that is my job that is on the line here). All I’m saying that you probably shouldn’t lead with that. Have your SEA and ad management teams optimize your campaigns, do proper onsite SEO (but don’t buy links, that’s fraud), check and improve data quality in your analytics tools, find and remove bottlenecks in your conversion process, conduct user surveys and in general start with all the things that do not require you to allocate more and more resources just to manage your toolchain. As long as there are still simple and cost-effective things you can do bringing out the big guns is just a sort of premature optimization.

Leave a Reply

Your email address will not be published. Required fields are marked *