Single channel conversion attribution doesn’t work.
To paint a real picture of how your customers interact with your website, it’s time to look at how multiple channels contribute to your buying cycle.
The most popular analytics platform, Google Analytics, uses a “last touch” model; a conversion – whether it’s a lead, a transaction, or a signup – is attributed to the last interaction between your company and your customer. Let’s take a look at why this is such a poor model for evaluating ROI from each of your channels.
Here’s an example of the journey a customer went through before signing up:
Using Google Analytics, the signup would be attributed to the final stage of the process; a branded search. Now, anyone who works in marketing knows that a branded search represents nothing more than a user looking for a company they are already familiar with, primarily with the intention of converting.
Would you say the branded search deserves the credit for the signup? No. Just no. Before returning to the site using a branded search, this customer viewed a display ad, clicked a paid ad, and viewed a comparison website. Even if we were to assign value equally to each of these touch-points are they truly equal?
Two of these touch points – the product review and product comparison – were persuasive interactions; experiences that convinced the customer this product was right for them, so surely they deserve as much, if not more credit for generating the conversion?
Other attribution models
Now, it’s fair to accept that the last-click model is unjust and (whilst it pains me as an SEO to say this) branded organic search receives far more credit than it deserves. I’ll be covering how to differentiate between branded and non-branded search in my next post, but for now let’s look at some of the other attribution models available.
For the purpose of this example, we’ve sold a product worth £100.
First interaction
100% of the credit is applied to the first interaction. This is a better representation than the last-click model as it values the initial interaction more highly. Winners of the first interaction model tend to be paid advertising, be it display, search, or social. However, this still doesn’t address the persuasive interactions.
In this case, £100 is assigned to our first interaction.
Notice how our Paid Search Cost-per-acquisition (CPA) is £338, whilst our Display CPA is £1,193? Clearly Paid Search tends to be an initial touch-point.
Time decay
Weighting steadily increases for touch-points closer to the final conversion. In this model the paid ad, comparrison website, and branded search receive the most credit. This is particularly beneficial for companies with long buying cycles as it weights the final few touch points before the conversion. But do we still want a branded search to get the majority of the credit?
In this case, £50 is applied to organic search, £30 is applied to our product comparrison site, and £20 is applied to our paid ad.
Now take a look at those CPA’s again. Display has increased to £455, whilst our Paid search CPA has increased to £2,224? Clearly neither are close to the final point of conversion.
Linear
Value is assigned equally to every interaction involved in the process. If we have 5 touch-points, that’s £20 allocated to each channel. If you’re looking for an easy way to assess your attribution, this is your best bet.
Now we’re hitting the middle ground. Paid Search CPA has reduced enormously to £986, whilst Display has also decreased to £409. That’s because Display is often a mid-journey touchpoint.
Position based
Weighting is applied to the first and last interactions. In this case, the blog post and branded search. Whilst this may seem a little more balanced, it ignores the entire nurturing process between initial engagement and final conversion. If your buying cycle involves a long lead nurturing process, steer clear of this one.
In this case, £50 is applied to our blog post and £50 is applied to our branded search
Our Paid Search CPA has shot up; clearly it’s rarely the first or last touch-point, whilst Display has reduced considerably.
Custom model
The fact you’re reading this article suggests you may not be ready for a custom model. Using a custom attribution model you’re able to manually apply value to each interaction, but be warned, this requires an in-depth analysis of not only your “planned” buying cycle, but the actual pattern users go through. In terms of accuracy, this is the ideal. However, the overheads in producing an accurate custom model are significant, and not a worthwhile use of internal time for most small to medium-sized businesses.
Another option?
By now I’ve probably scared you to the extent that you’re happy sticking with Google Analytics’ default model, but fear not; there is another way, and it’s call assisted conversions.
Under the assisted conversion model, credit is assigned to the final conversion touch-point, but an assisted conversion is assigned to all channels which contributed to the conversion.
In this case, £100 is applied to branded search, whilst all other channels contributed to our sale.
So what’s right for you?
Sorry to disappoint. As you’ve probably already gathered, mapping out an accurate attribution model is a complex challenge, but it’s important in understanding your ROI from various channels. There’s no “one-size-fits-all” solution, but if you’d rather not invest too much time in mapping your buying cycles, I’d suggest using a Linear, First interaction, or Assisted Conversions models for simplicity.
Hopefully this article explains just how much our perception and interpretation of channels can change depending on the model we’re using. So before you reduce or increase investment in the specific channel, have a play with some of Google Analytics’ attribution models to make sure you’re making an informed decision.