How to set up marketing attribution to build a predictable growth engine

While we agree (and respect) that Don Draper wouldn’t have done marketing attribution, he could have never imagined the level of marketing attribution we have today. He'd also laugh at how many companies don't do it correctly—or at all.

Most of our “marketing stack” doesn’t help, either. We’ve been sold tools that attribute customers incorrectly or in silos, we’ve read blog posts about how attribution is dead or going away, and every time we try to solve it ourselves we’re told it’ll take too long and to use what you have because it “works for now”.

Marketing attribution sucks for everyone, but it’s solvable. Those who attribute customers correctly through the full funnel have an unfair advantage—they truly understand how their business grows. And when you understand that down to the growth lever, you can build a portfolio of channels to control that growth.

This post will cover:

• How to choose the right attribution model
• How to align attribution logic to growth levers
• How to maximize attribution understanding as you scale

Choosing the right attribution model

We're oversimplifying, but the basics to consider are:

First-touch attribution (recommended to start)

First-touch attribution assigns all credit to the very first interaction a customer. It's about identifying what brought the customer in the door.

We'll say what most marketers won't: a first-touch attribution model works best for most companies and is your safest bet to start. Understanding what initially acquired the customer is crucial for building a repeatable engine for growth. No following touchpoints happen if they never get in the door.

Multi-touch attribution

Multi-touch attribution distributes credit across multiple touchpoints within a conversion window and up to a conversion event (and no, there will almost never be hundreds of touchpoints if you’re setting up multi-touch correctly). We usually only use this model when your average touchpoints per user before conversion is greater than 1.25 (meaning 25% have more than one touchpoint).

Common multi-touch model options:

• U-Shaped: Assigns more credit to first and last touchpoints, pushing the importance of acquisition and conversion points.
• Equal Weights: Assigns equal credit to all touchpoints. Simple but may overvalue minor interactions.
• Time-Decay: Assigns more credit to touchpoints closer in time to the conversion, valuing recent interactions more heavily.

Last-touch attribution

Last-touch attribution assigns all credit to the final interaction before conversion. This model is rarely the right fit for any company. Last-touch interactions often influence conversions but don't acquire traffic. They can add incremental costs to acquisition campaigns. If you're running remarketing campaigns, consider a multi-touch model instead of last-touch.

Aligning end-to-end attribution logic to growth levers

End-to-end attribution refers to the full hierarchy of attribution—Traffic Sources, Channels, and Campaigns. It lets you track business growth based on the team, the platform or partner type, and the levers your teams control within those—so you can truly understand and control your revenue growth.

Maximize attribution understanding as you scale

When you set up attribution properly, the data shows you whats working, whats not, and opportunities to improve. No substitute will give you that level of clarity for day-to-day scaling and optimization of growth channels, but additional measurement tools can help fill in gaps and complete the picture.

No, there is no single method of attribution that is perfect. There are channels that have halo-effects, many offline channels that are unmeasurable through click attribution, and tracking in general has gaps for many reasons out of your control (gdpr, ccpa, ad blockers, etc). Choosing one or more of the below solutions should help round out that understanding when you reach that scale.

1. Click attribution

Click attribution is your daily guide for reporting and optimization. It shows where actual growth is coming from, but only if you align it tightly with your teams, channels, and campaigns.

2. Mixed Media Modeling (MMM)

MMM is a model that helps you understand what channels are driving revenue and conversions. Use it to audit click attribution and to factor in offline channels and halo effects. It’s not pure attribution but can guide smarter spending.

3. Lift tests & incrementality

Geo-lift tests and incrementality models are your real-world validation for halo effects and offline channels. MMM can estimate impact based on historical data, but these tests confirm those estimations in a real-world test. Pause spending in one channel or region, observe the results, and measure within your model against a synthetic control to ensure your attribution is accurate.

Remember, #2 and #3 aren’t pure attribution but a data model to guide smarter future spending. A model is only as good as its inputs. Use it as an audit against click attribution to understand which channels are adding more or less value than click attribution is reporting.

4. Bonus: Self-reported

Having a ‘How did you hear about us?’ question after sign up or after conversion is another great way to gather data to audit your source-of-truth attribution.

That said, this method is prone to more errors than data modeling options above - human error. Set inputs lead to lazy responses and open-ended inputs make it hard to pattern match for an effective audit.

There are some cool things companies are doing to normalize open-ended responses with AI to make pattern match easier, but our advice is only spend as much time here as needed to audit your source-of-truth, and consider the options above for that first.

Don’t fall into the incomplete attribution trap

Incomplete attribution refers to partially implemented attribution. That might mean:

1. Implementing attribution for the traffic source and channel, but not for campaigns (growth levers).
2. Implementing attribution for your website and product metrics, but not for your revenue or sales metrics.

It can mean a number of things—the point is—you need to attribute customers through the full funnel, with end-to-end attribution, or you’ll always have blindspots.

You have to pull the thread all the way through. Full-funnel, end-to-end attribution is the only way to stay ahead and scale effectively. Anything less, and you’re leaving money—and valuable opportunities—on the table. You’ll optimize for what you see, but what you don’t see will hurt you over time.

Keep it simple, let us know if we can help

Getting this right is extremely important for any tech company. We’ve seen hundreds of companies with unfair advantages and control over their growth that most other companies don't have, and it's night and day.

Don’t overcomplicate it. Choose your model, align your logic to growth levers, add supplemental measurement tools as you scale. I promise you, if you get this right, how to grow your business will no longer be a black box.

Reach out if we can help: tim@roadwayai.com