Attribution Is Mostly BS
Marketing attribution is BS.
I recently interacted with a CMO who is obsessed with attribution models. "If only I had a good attribution model, everything would be better."
I don't see how this is true. It's a false belief. Let me explain.
The default attribution is last click. They way this works with affiliate or paid ads is to implement a tracking pixel on your website or app that fires when the user makes a purchase. The pixel tracks the source coding (UTM values) that indicates what ad the user clicked on that brought them to this session. Standard conversion settings then attribute that conversion to the ad and use the results to calculate metrics like CAC (Customer Acquisition Cost), ROAS (Return On Ad Spend) and others.
But what if the user sees four adds before this one and then they finally purchased after being persuaded by multiple ad impressions? Aren't we misallocating the conversion to the last click and not giving due credit to the earlier clicks that led up to the purchase decision?
Well, maybe. But you can only know this by looking at user-level impression and click data at the ad level. In other words, it's a data science exercise. If you don't have the data to back up the assumptions in the attribution model, then it's just an untested, unvalidated hypothesis.
Let's think about this from first principles.
1. Is your product a high consideration purchase?
The theory that users need to see multiple ads to convert is based on the assumption that is how they behave. In the example mentioned above, the CMO's business is a simple content subscription model for $20 to $30 per month. What makes us think users need to see five ads before they are ready to buy? What data do you have that supports this theory? How do you know it takes five ads to convert? How many Customers subscribed after seeing just one ad?
2. What are the interactions between the ads and the sequence users are exposed to them?
Let's say it's true for a moment that users need to see five ads to make a purchase. Does it matter which ads they see? Does it matter in which order they see them? What if one of the ads has an offer for a free trial and the others do not? Does the conversion rate change when you show the free trial first? or last?
To fully understand attribution, you need to answer these questions. Even if you can, ad platforms like Google and Meta have no way to showing ads in a particular sequence. All your ads go into their black box and they do whatever they want.
So, in the event you find the perfect ads using a team of data scientists that determine the optimal ad frequency and sequence, you have no way if implementing it on an ad platform. You'd be much better off by capturing an email signup and creating a drip campaign where you are in control of the message frequency and sequence. In this case, any attribution questions should be focused on email signup conversions and ultimate LTV/CAC ratios from these users over time.
3. Do you have the data and the analytical infrastructure to develop an effective attribution model?
To do this, you need all the impression, click and conversion data at a user level. To calculate Customer profitability, you need all payment data over time. In addition, you need a data scientist and algorithm builder. 99% of marketing teams don't have any of these things. Instead, they rely on the models provided by Google and other ad platforms. These models are a black boxes and advertisers have no idea how they are built or what they are based on. Do you think Google is optimizing your profitability or theirs?
You can test it!
The good news is you can test Google’s models, or any other model. Here’s a simple way to do it. Take an initial read on CAC, ROAS and LTV/CAC. You’ll need to use predictive LTV initially and actual LTV for cohorts that are a few years old. We can help you do this if you don’t know how. Then run an A/B test. Create two campaigns with the same ads and user targeting. Allocate spend in the first campaign using last click attribution and allocate spend in the second campaign using whatever attribution model you like. Compare the key metrics.
What do you see? You will see last click works better, every time. I’ve never seen a case where this isn’t true. You simply cannot increase conversion, efficiency and Customer profitability by spending more on low converting ads.
Think about it. The premise is illogical. If you spend more on ads that don’t convert well, then your overall conversion rate must go down. If your conversion rate goes down, your ROI goes down. It’s just math.
The idea of using an attribution model came from ad agencies and platforms that get paid why you spend more on ads. This idea serves them because it convinces advertisers to spend more on crappy ads that aren’t converting well. Attribution models are a winner for the ad companies and are losers for the advertises.
What do you think? Do you have a counter example that proves I’m wrong about this? If so, I’d like you to share it so we can learn more.
David Linhardt