Field Notes #53: The Best Way to Measure Content Program Success 🔥

Attribution models can be a contentious topic amongst marketers and analysts.

It makes sense because beneath all the debate and math, attribution ultimately determines how much we give “credit” to each touchpoint in the customer journey.

The “credit” determines how much impact a marketer has on the business and ultimately justifies their role. So of course marketers will argue about how much credit they should get.

Of all the available attribution models, there’s a common belief that last-touch attribution is good for direct response marketing channels like paid marketing.

Others say it’s unfair because it gives all credit to the last engagement without considering all the prior touchpoints. These conversations end up discussing the merits of multi-touch attribution models (linear, U-shaped, time decay, W-shaped, custom models, oh my…).

Then there are folks who say trying to do any attribution modeling is a waste of time. To be fair, there’s some merit to this (data is never perfect), but it isn’t helpful. The conclusion may be to rely on self-reported attribution which, based on decades of qualitative research, has its own flaws.

My take? There’s no perfect solution. Just pick one that works for you and stick with it. I know, not the most satisfying answer.

(Okay the more nuanced answer is that models like incrementality and Markov chain models can be more reliable, but they require highly sophisticated tooling, data infrastructure, and data scientists to make sure they’re done correctly. 1) This is overkill and 2) a data-driven approach still won’t end the debates. 😅)

For our work with clients, we use last-touch attribution. Here’s why.

It’s simple to understand when communicating with stakeholders.

Let’s say you’re at a small company and have stakeholders across multiple departments. They aren’t as privy to the various types of attribution models. Trying to explain a W-shaped attribution model to your head of sales or CEO isn’t going to be a great use of time, because they don’t care to know and it’s not the most important thing to debate for a small company. 

What the head of sales and CEO care about is whether you’re generating leads for the business. It doesn’t matter if a lead visited the marketing site seven times before responding to a sales rep’s cold email. That sales rep converted that contact into an opportunity, and that’s that. They get the credit.

Now, if you can say to your CEO, “100 people visited this blog post and of those people, 10 filled out a form to contact sales,” that’s easy to understand and hard to argue with. You don’t need to understand W-shaped attribution models to understand that.

Once you’ve built trust and demonstrated last-touch attributable revenue to content, you can get fancier with the attribution models.

Our take, say f@%# the fancy attribution models. Keep using the simple model and spend that time on more creative brand experiments.

Easier and cheaper to implement.

You’ll have customer journey data across the board from website sessions (Google Analytics), paid social engagements (Facebook, Instagram, LinkedIn), paid search engagements (Google Ads), webinar registrations (Zoom), various content downloads (HubSpot), website chats (Qualified), and more.

You need to unify all that data needs to get a view of the customer journey. That’s a lot of time either spent with data engineering (the scope of which can expand really quickly) or setting up a platform to do this (Census, Customer.io, Segment).

Instead, just use the default last-touch attribution with Google Analytics and call it a day. Save yourself the thousands of dollars that you would’ve spent on data engineering or new software.

If you’re a small team trying to move fast, building these data pipelines shouldn’t be a priority.

It’s the most challenging model for content so results are harder to argue against.

Blog content isn’t the first medium of content that comes to mind when you think of “direct response conversion.” You’d normally think about paid marketing, which is what people tend to say is better suited for last-touch attribution.

Content tends to be somewhere along the customer journey, but it’s not often believed to be the “last touch.” 

By choosing last-touch attribution to determine the impact of the content program, we stack the deck against ourselves and make it more difficult to give credit to content. So any outcome (leads, pipeline, revenue) that is attributed to content is harder to dispute.

When I say we’re “stacking the deck against ourselves” I’m referring to the fact that these companies are still continuing all their other marketing efforts (paid, social, webinars, outbound sales, etc.).

Despite that “stacking of the deck,” we’ve proven that content programs can compete with all those other marketing programs for credit, and still demonstrate meaningful attributable results:

Easier to tie to revenue.

If you use last-touch attribution, you know what page the person visited before they converted. 

If you’re lucky, your ops team already has this reporting set up in their CRM or BI tool, and you can find out how many people who became a lead from content turned into paying customers.

This is still easier said than done, but it’s significantly easier than setting up a time decay model and then writing formulas to attribute a certain percentage of revenue to each touchpoint. (Bleh.)

If the reporting doesn’t exist, it will require some manual work to understand the impact of content on revenue.

That manual work is worth it to prove the impact. That means you should build the reporting yourself if your ops team won’t build it for you.

This type of reporting that ties activities to revenue will demonstrate the value of the content program and build trust and confidence with leadership and stakeholders.

How about the flaws of last-touch attribution?

There are plenty of flaws, and I’m not here to argue that last-touch attribution is the best attribution model. Merely that it’s a reasonable attribution model to build trust and show impact.

Here are some of the trade-offs of using last-touch attribution:

• It doesn’t capture the full customer journey that led to conversion. (My [flawed] counterargument is perhaps all those earlier touchpoints weren’t convincing enough to convert them into a customer.)
• It overweighs the impact of the last touch which may result in investing too heavily in that last-touch channel. 
• It doesn’t serve multi-channel, multi-device marketing efforts.

One solution is to use self-reported attribution. In the form used for conversions, have an open text field asking how they found out about your product or business. It may be flawed (maybe they don’t remember or misremembered), but it at least gives you an idea of what part of your marketing had the biggest impression on them.

All of this is to say, all models are flawed. All models are wrong, but some are useful. 

The goal isn’t 100% accuracy, but to have a model that teams agree on and work within the flaws, instead of constantly debating who gets credit.

Using a simple, understandable attribution model that allows you to report on the revenue impact of the content program is how you communicate impact and get more budget.

Want more insights like this? Connect with me on LinkedIn.

Recommended Reads

1. Don’t Confuse Marketing’s Outcomes From Its Outputs by Mayur Gupta
2. Attribution is Bullshit by Jeff Sauer (video)
3. Why Marketers Run the World, Adaptability for Survival, and the Generative AI Revolution with Chris Toy (podcast)