One of the most simple questions asked in analytics is, “How much money are we making from our paid search campaign?” The problem is that there are many ways to answer this question, as well as many factors from the Web analytics side that can created different answers.
As a Web analyst working within a team of more traditional SQL-using, data analysts, explaining how an analytics solution answers the above question can be challenging. The 3 primary variables that are a part of a revenue attribution methodology include:
- The order with which the campaign credited with the sell occurs in relation to other campaigns
- The length of time that a campaign may receive credit for a sale
- How attribution is split (or not) among multiple campaigns
As for order, the most common approach is last touch. In other words, if a visitor clicks through your email campaign today and then through your Google ad tomorrow, the Google ad will get all of the credit because it was the last campaign touched before the purchase. The problem of course, is that even though the email campaign was clicked first and might have impacted the sale, the email receives no credit. One alternative to last touch that gets around this is linear attribution. Basically, linear attribution would split the previously mentioned sale 50/50 between the email and the Google ad. But should it really be 50/50? In addition to last touch and linear, you can also have something like first, or original, touch, where the email would get all of the credit. So there are a lot of choices to mull over.
Now that I’ve talked about a few of the methodologies around the order of attribution, the variable of time needs to be added. Using the previous example, and assuming last touch as the order of attribution, how long after coming in through a Google ad should the ad get credit for the sale? Only if they buy within the visit? 7 Days? 30 Days? The most typical solution is 30 days. However, this could very well extend out to months if your Web site is one of lead generation where the sales cycle is weeks or months long. Also, if you send out daily emails, is 30 days really a good choice for attribution? If you don’t have the choice of a custom solution, then 30 days is probably your best bet right now since that seems to be the standard. But, just keep in mind that you might have the option of changing your attribution to any time preiod (or maybe even event on your Web site).
Earlier, I mentioned linear attribution as a method of splitting revenue between multiple campaigns. Aside from this even split among campaigns, there are not many other options out there. This is one of the biggest challenges in revenue attribution. One way around this is to export all of your analytics data by visitor ID for every visit (that’s a ton of data to say the least). Once you have this data, you can create your own methodology to tie a sale back to every visit by the visitor, and every campaign that they touched (and the time between) prior to the sale, all the way back to maybe even the first campaign code ever touch by the visitor. We’ve done this at my current job, and I can tell you that it is not something that is easy to recreate on an ongoing basis.
It would be great if there was some solution on the Web analytics vendor side that would let you create a truly custom attribution methodology. However, the problem there is that if you can customize every aspect of attribution, you might end up creating a self-fulfilling prophecy. What I mean here is that if you want to weight the last touch before a sale as being worth more than the campaign touches between the first and last, then you might be over valuing paid search if that is most often your last touch marketing channel.
So what is the solution?
As far as I am concerned, it is short sited to value everything as last touch. You can’t just ignore the fact that other campaigns have in someway influenced/impacted your visitor prior to making a purchase. To ignore this is to miss out on understanding and optimizing your marketing efforts from beginning to end. So, ditch last touch (in a perfect world, if you can).
Next, you’re going to need to create your own solution as to how to tie all of the different campaign touches together and appropriately attribute them to the sale. This is an easy thing to say, but not so easy to do obviously. In a later post I will try to flesh out an idea to actually do this.
Do you have any ideas as to how to improve upon existing ideas of revenue attribution? If you’re doing something custom yourself, let me know about it.