Credit Where Credit’s Due – Do Attribution Models Work?

by John Wilkes Somebody Digital

The thing is, few customers make a purchasing decision based on a single ad or marketing message. Most of the sales you experience have touched upon many different marketing channels before they ‘settled’, and the customer may have interacted with your company through TV, social media, and internet search results (both paid and organic) before they decided to buy.

So which channel gets the credit for the sale?

What Is ‘Marketing Attribution’?

Marketing attribution is the process of deciding how much of any one sale is counted towards each marketing or sales channel so you can judge which ones are more or less effective.

It’s like the old say says: ‘Half the money you spend on advertising is probably being wasted. The trick is figuring out which half’.

Attribution involves identifying a common set of customer interactions which might be called ‘touchpoints’ or ‘events’, and then assigning a percentage value to each that matches the influence each had on the buying decision.

Attribution vs. Analytics

Attribution is just as important as analytics, but in different ways. Analytics tell you (roughly) what your customers are doing. Attribution tells you what parts of your marketing strategy are making them do it. Both will work independently, but you’ll get a synergy boost if you use both.

Attribution Plus Analytics

There are many methodologies to pairing attribution and analytics, and each will give you slightly different results. There is no one ‘correct’ methodology though, as some work better for one industry or niche and worse for another.

Worse, it seems like there is a new marketing channel born every minute. That can make your ‘tried and true’ methodology outdated overnight. When you add so called ‘dark traffic’ to the mix, it can be harder than ever to attribute those sales to individual touchpoints accurately.

What Exactly Are You Trying To Achieve?

Like the little green muppet said “You must unlearn what you have learned” about attribution. Ignore the model types you may be using and everything you know about accuracy for a moment.

Proper attribution starts with knowing your goals. When you do select an attribution model, what should the outputs be? How will the information be used?

Let’s say that the model you select says that your point of sale displays add nothing to your sales. Would you stop using them? If your answer is that you’d keep using them no matter what the results because you believe they add intangible but important value, then maybe you shouldn’t have selected a methodology that gave you numbers on your point if sake displays at all.

You’re not asking “does this model bring me joy”, but you should be asking “does this model give me numbers I can really use?”

Another important question is “How complex should my attribution model be?” If you only have a handful of marketing channels in total, you might not need anything more complex than basic analytics and ‘last click’ attribution. If you have a Byzantine nightmare of media platform, offline and online marketing channels, social media accounts and ways to buy, you’ll need a much more robust solution!

Meeting The Attribution Challenge

Attribution is not just a technical challenge, it is a management issue as well. It isn’t the type of thing that was taught in most business schools until quite recently, though. You’ll discover a surprising number of stakeholders who are resistant to ‘complicating’ their analytics setup. Luckily, getting everyone aligned to the new goal and overcoming pushback are classically taught business strategies.

Just be clear with your message to them – Attribution isn’t a ‘new kind of analytics’, nor is it some fancy offshoot of funnel mapping. It is a method for estimating the relative effectiveness of different touchpoints so that you can compare that to their actual costs. Nothing more. It shouldn’t ‘threaten’ anyone, and it should cheaply provide data that everyone can use to make your processes more effective.

The most dangerous attitude you might face is the idea that attribution is something the CMO will do off by themselves somewhere. Proper attribution strategies require an investment of development resources. You’ll need financial modelling to be done. You’ll need leadership support. It will not be ‘just one person’s job’.

For example, let’s say your product team brings a new on-site acquisition flow online. If they haven’t given a thought to attribution (and they won’t if they don’t have to), your LTVs could be meaningless for a good long time before someone notices the discrepancy. All that work wasted, and worse, all those decisions based on inaccurate results. For attribution to be everyone’s friend, it also has to be everyone’s job.

Is Attribution Trustworthy?

You’ll also need your boss to trust those numbers, else you’ve wasted your time.

Attribution done right is not about confirming existing marketing models or making the right people feel good. It is about giving accurate data which can be acted upon. For that to work, the numbers don’t just have to be accurate, decision makers need to have faith in them.

There are two basic types of mistrust in attribution data.

  • People distrust ‘last click attribution’ because it is both easy to understand and demonstrably inaccurate in most situations.
  • People distrust complex ‘black box’ attribution algorithms because they are both almost impossible to understand and it is almost impossible for non-technical types to say whether they are accurate.

On the face of it, that seems pretty bleak, but there is actually a wide grey area between those 2 extremes.  Somewhere in there you’ll find a sweet spot where marketing teams begin seeing results they can understand but seem refined enough to inspire trust. Once they start letting that influence decisions (and get positive results), the system becomes self-reinforcing.

Note that that last step is dependent on the attribution actually giving that actionable data.

So somewhere along the line you’ll need to have a long, multi-stage conversation with everyone involved, where they can agree on a model that they all find both trustworthy and just mystifying enough.

In The End, What Does All Of That Mean? It Means It’s Hard!

Everyone deserves a good executive summary, right? In short, marketing attribution is difficult. There are many different models, and no ‘clear choice’ that is ‘right for every application’. The key is in choosing the right method for your purposes.

The trick to that is knowing what information you want to spill out of the ‘black box’, choosing a black box with the right outputs, paying attention to those outputs and ignoring all of the other outputs.

The rest is just ‘refinement’.

John will be speaking at our Digital Marketing Summit on Thursday 25 April 2019. For more information regarding the event you can view the full agenda here.