Experiment and refine until you find your custom attribution model

The trouble is, things are seldom that simple in advertising. A customer might see your billboard on the street before searching for your product. After doing that, they might take a few days before deciding to make a purchase during which they are served targeted banner ads for the product. When they finally do make a purchase, to which channel should the sale be attributed?

Further complicating matters is the sheer number of attribution models in use, all based on different assumptions about the customer sales cycle. It should hardly be surprising then that many marketing decisions are still made on gut feel.

That said, how can CMOs ensure that they’re using an attribution model that allows them to make the clearest possible marketing decisions?

The evolution of attribution models

In order to answer that question, it’s worth taking a deeper look at the evolution of attribution models. The earliest, and most simple attribution models were rule-based ones. These models can broadly be divided into single and multi-touch attribution models. Single-touch models assign the whole value of a conversion to a single click – whether that’s the first click, the last click, or the last non-direct click. Multi-touch models, meanwhile, distribute the conversion across multiple touch points based on a series of set rules.

Newer attribution models add machine learning into the mix and rely on having large amounts of data for the model to analyse. The idea here is to get an overview of a large volume of conversions and journeys to provide the most accurate calculation on how each touchpoint contributes towards a sale.

Although such a model might seem like the ideal solution, it’s important to remember that the model is only as good as the data that gets fed into it – feed in garbage data and you’ll get garbage results. The thing is, it’s all too easy to get the data wrong.

Even if we restrict ourselves to online channels, an organisation might not be able to accurately track a customer’s full journey as they move between devices and environments (such as work, home, and commuting). And that’s without even going into the complexities of things like private and incognito browsing and dealing with legislation around browser cookies.

You can’t force the right fit

While it’s important to understand the shortcomings of certain attribution models, that does not mean that CMOs should stop trying to find ways of attributing leads and conversions.

Large organisations, for instance, often find value in marketing mix modelling, which allows them to generate big-picture trends around the effectiveness of certain channels and allocate budget accordingly. The important thing to remember with this approach, however, is that it doesn’t offer the kind of detailed, granular reporting that can clearly indicate to organisations which messaging actually works.

Alternatively, organisations can run A/B tests on the part of an audience to see which messaging works, splitting them by geography or demographics. It’s often a worthwhile exercise, but because many A/B tools are cookie-based, they tend to suffer from many of the same issues as the other models mentioned above.

So, what should CMOs do?

For CMOs, the answer probably doesn’t lie in trying to force any singular standard attribution model to fit their business. Instead, they should rather continue experimenting, adapting, and refining until they have a custom model that works for their business.

The bottom line: Trustworthy data

Adopting this nuanced approach means that CMOs can feel much more comfortable knowing that they can trust the data available to them. As a result, they would put themselves in a much better position to design campaigns that consistently deliver results, without having to revert back to a reliance on gut instinct.