Attribution: Five Tips For Gathering Accurate & Complete Customer Source Information
‘Attribution’ is a word I have been seeing with increasing frequency these days. This post is not intended to be a contribution to the academic furthering of this important concept; rather it has the simple purpose of providing a few common-sense ideas about gathering complete and correct information in anticipation of your need to eventually do some sort of attribution analysis.
Let’s start with the definition. ‘Attribution’ in the marketing sense is defined in Wikipedia as “the process of identifying a set of user actions (“events”) that contribute in some manner to a desired outcome, and then assigning a value to each of these events.” In other words, attribution is an attempt to quantify the various ‘touches’ that lead to a ‘conversion’ or purchasing decision. Marketing attribution is becoming a higher priority with the expansion of digital marketing, which lends itself to tracking customer behavior in ways never possible in the offline world.
It is important to know is that there are several attribution models. No matter which one you choose to use, you should gather complete customer data. Mainly because you never know when you will change attribution models AND you can never go back to get complete and accurate information from your customers once that data-gathering moment has passed. So I’m advocating some simple guidelines so you gather complete and accurate information the first time.
First, a brief description of the types of attribution models:
- Single Source: with this model, you assign conversion credit to one ‘touch’ – in many cases the last event before the conversion occurs. This is considered the least accurate method of attribution.
- Fractional Attribution: credit is assigned to multiple touchpoints. Examples are the Equal Weight model that distributes value equally to multiple touchpoints and U-curve, which assigns weight only to the first and last touch, ignoring all progressive touches in between.
- Algorithmic Attribution: this approach can be the most accurate, and certainly is the most sophisticated, requiring some sort of statistical data modeling to come up with a distributive attribution allocation unique to your particular business.
Note that as you become more sophisticated, you will likely move to a multiple-touch model. And get this: well over a century ago, multiple attribution models were recognized! Check out Advertising Effectiveness, dated back to 1885:
I always smile when I read this because I realize that the some things never change, even from previous centuries…
OK, so now you are convinced of the importance of attribution analysis, here are some simple guidelines to keep in mind so you gather complete and accurate information:
- Whenever you ask customers how they connected with you, do not word it, “How did you hear of us?” because that may lead to specification of the very first touch (only), which is not what you want. Instead, ask, “How or where did you get information about us?” This will cause them to think about (multiple) active and passive touches with your advertising and marketing channels.
- Ask this information at every entry point, whether offline or online, and make sure you ask in the same way every time. A lack of consistency on your part could invalidate your data.
- Present a list of selections from which the customer may select MULTIPLE options. This is one of the biggest mistakes I see, because it creates the single attribution source inaccuracy described above.
- Your list should be exhaustive! It is important this list be inclusive of marketing you are currently using, have used in the past or will use in the future. It should be comprehensive because to change the list over time invalidates responses and does not allow you to evaluate trends. Also, please do your customers a favor and make the list easy to use. Easy to use means all the items are at the same hierarchical level and that there is some obvious organization, such as alphabetization.
- As much as humanly possible, remove the chance for human bias and error. So this means if possible, let the customer make the selections without prompting by anyone, and if at all possible have a digital device handy even in physical locations so the data capture is direct from the customer.
The truth is, getting to an attribution model that works for your company is important, and it is not very difficult to gather accurate information. It just takes a determined and committed effort by you and your employees. The payoff is that future marketing dollars will be much better spent and your cost per new customer acquisition will go down. Well worth the effort, in my book!