Measuring Attribution: How Your Numbers Are Lying To You and How To Fix It
Let us get the bad news out of the way first: you have an attribution problem. If you advertise on more than one channel, the (very real) possibility that a user interacts with your marketing in several places will complicate your data. A user might interact with more than one ad that goes to the same landing page before making a purchase.
In this case, your cost per conversion will be underreported. Several publishers will essentially take credit for converting the same user. This is why we have to learn to measure attribution. Attribution means crediting your conversions to a marketing channel and assigning a cost. It’s a way of determining what sources have the most influence over your customers’ conversion. It all comes down to tracking your users across multiple channels and understanding their buying behavior. It might sound a bit daunting, but with a little understanding and the right tools, you’ll be well on your way to getting better data about your campaign.
Track channels with UTM parameters
As Ryan Koonce explains on his “Marketing Attribution 101” webinar, good attribution is a combination of good channel tracking plus good conversion. You can only understand attribution if you track channels properly. For this, most marketers use UTM parameters. These are basically a way of creating tracking links that look for certain parameters in users. When someone clicks, they send the information to analytics tools. They’re stubborn little tags that stick to your links even across different media and networks. You can add them to your URL’s with Google’s URL builder.
The most important parameters are:
- Campaign source (where you purchased the ad)
- Campaign medium (where your ad is running: email, CPC, social, etc.)
- Campaign term or keywords
- Campaign content
- Campaign name (that you set up in your advertising account)
UTM parameters can be used together with auto-tagging. Auto-tagging is convenient because it will pass data directly from AdWords to Analytics. It saves time, eliminates errors and is easier than tagging every destination URL by hand. However, auto-tagging only works with Google products and it limits your use of UTM parameters in other tools.
Five models of attribution
There are several ways to define who gets the credit for your conversions. These are the five models of attribution:
- First touch: the first ad clicked gets all the credit
- Last touch: the last ad clicked gets all the credit
- Linear: all ads get equal credit
- Position-based: the first and last ads get more credit, usually 40% each, while channels in the middle get equal credit (usually 20% in total)
- Time decay or time-based: all ads leading up to conversion get some credit and the last ad gets the most
The first touch and last touch are the primary models. They can give very different information about your cost of acquisition.
For example, let’s say you spent $20 on AdWords. Then you have a person who:
- Clicks on AdWords and visits your landing page
- Clicks around but doesn’t complete any purchase
- Comes directly to the site later (through a search) and makes a purchase
If you measure by the first touch, your COA is $20 because the sale started with a click on the Google ad. But if you use the last touch, your COA is $0 because it’s attributed to direct traffic. Last touch attribution has an intuitive appeal – after all, it credits the source that “made the sale” – but it discounts a lot of the conversion process. When a client converts after searching for your brand name, it’s usually because they were exposed to your brand before. The ads that introduced them to your brand might have been what actually convinced them to buy.
This makes the first touch sound like a smarter choice. However, there are problems here too. Where do you draw the line in measuring the gap between the first click and final purchase? If someone clicks on an ad and, five days later, makes a search and buys the product, can you really say the ad was possible? What about after two weeks? Two months? Also, the first ad might have very little to do with the customer’s buying decision. It might be just their first exposure, and ads on other channels might have much more to do with why they make the purchase. If last touch attribution undervalues the awareness channels, the first touch undervalues the cumulative effect of marketing.
How do I know which model to use?
Any of these models (or a combination) might be right for your business. The only way to find out is to run the data. Start by making sure that you’re tracking your channels correctly. Then compare first and last touch attribution. You may or may not see a big difference, which will tell you something about how people are finding your site. After that, check out the other models.
Time decay is a good answer to some of the problems of first touch and last touch. It takes the approach that all of the user’s early interactions with your marketing contributed something to their conversion, but the last one takes the cake. After all, if one of the earlier touchpoints was so strong, why didn’t they convert then? Of course, I could come up with several scenarios off the top of my head, like the user wants to buy after clicking on an ad but doesn’t have time and is reminded to purchase later by seeing another ad. But overall, it passes the common sense test more than the others.
How do I measure attribution for my business?
In my opinion, these are the top three analytics tools for measuring attribution. Google Analytics: Google Analytics mostly works by the last touch, though a few years ago they introduced “Multi-Channel Funnels” to track users who go through more than one source. Their tool now supports all five models of attribution and it allows you to compare three models at once. However, it doesn’t do so well in identifying users.
Kissmetrics: It supports first touch, last touch, and linear models. Kissmetrics is great for showing you which users are doing what, and when and where they’re doing it.
Attribution: This app supports all five models for Facebook and AdWords, including cost data. It works well but with only two major channels, it doesn’t give a full view of your users.
Attribution is a thorny subject. It’s easy to fall into a statistical rabbit hole with it, lost in the what-ifs and unknown variables. Fortunately, you don’t need to understand the why’s and where’s of every micro-conversion in order to clean up your data. Get started with UTM parameters and a good analytics tool, and you’ll immediately start seeing the information that might change the way you run your campaign.