Skip to content 
Bart is an in-house Growth Lead for a DTC brand, and he’s crushing. He’s improved revenue and ROAS from Meta three months in a row.
But a new VP of Ecommerce just started. She asks Bart, “If we’re reporting revenue/ROAS out of Meta, how can we be certain it’s incremental? Would we have gotten some of that revenue organically anyway?”
Bart doesn’t know. His job is to drive growth through Meta. Meta’s setup was configured by the marketer before him and he never looked that closely at it. But now, he’s hearing, “we need to hold off on increasing spend until we can be confident in our attribution model.”
Bart’s mistake? Relying on a platform attribution model without clear assumptions.
How to Create an Attribution Model
An attribution model is like a window into a house.
If you have a wide, clear window, you can get a good sense of what’s going on inside. But one window alone can’t give you the full picture. So, just as you need to look in a few windows to really see inside the house, you need to balance complementary attribution models to really understand marketing performance.
I recommend every business have three ways of looking at attribution:
- Single-channel attribution
- Sales/CRM record attribution
- Multi-channel attribution
Single-channel Attribution
This is when you rely on the data provided by the channel itself. For example, reporting on Meta ads out of Meta or on SEO out of Google Analytics.
These can be reliable, but make sure you address two key considerations, conversion setup and incrementality assumptions.
Conversion setup
To rely on your on-platform data, you need to send conversion data from both the browser (client-side) and your app or CRM (server-side).
Take a purchase event in GA4, for example. If the event is fired client-side, that means the event was sent to GA4 from the website’s HTML and JavaScript. If the same event is fired server-side, that means the event was sent to GA4 directly from the payment processor (e.g. Shopify or Stripe) via API.
Both sources give valuable context for ad platforms to use in optimization. The server-side event acts as a failsafe in case the client-side one is blocked. And ad platforms know to deduplicate between sources using event/user IDs and timestamps.
This dual setup gives your ads the most optimization fuel and gives you the best chance of reporting the most accurate possible conversion number.
Incrementality Assumptions
Marketers naturally want to assume that all marketing efforts are incremental, especially paid marketing. In reality, some those conversions reported from paid may have happened anyway.
Sometimes, paid just sneaks in at the end with a retargeting or branded search ad to close the deal, and the platform will report all the credit.
I call this paid “drafting:” your paid ads are drafting on the work of your other marketing efforts, taking credit for marketing work they only partially did. It is the marketing version of stolen valor. Your ads are like that one kid in your group project who does nothing but still gets the same grade as you.
You can address this in 2 ways:
- Different attribution settings for different campaigns. For your prospecting ads, you may want to set your settings to “7 day click, 1 day view”, but for retargeting, “1-day click only”.
- Applying an incrementality assumption to reporting/target CPA. If you assume your branded search ads’ conversions are only 50% incremental, and your regular target CPA is $50, you can set your target CPA to $25 instead and in the spreadsheet you report out of, apply a 50% modifier.
Great platform attribution starts with great assumptions.
Sales/CRM Record Attribution
In 2024, you need a data source that ties channel performance specifically to sales or CRM records. There’s just no way around it.
This typically means tracking and reporting data in your CRM or email platform. Examples of these platforms are Hubspot, Klaviyo, Shopify, or Salesforce. This can be intimidating at first, as marketers do not usually build reports there, and the platforms often have less robust reporting features than marketing tools.
Example of sales attribution reporting in Shopify.
But once you can point to the customers and leads who came from your marketing efforts—right down to their email addresses—you’ll develop an intoxicating confidence and clarity. Leadership will love you. You’ll never go back.
Attribution in these platforms will be less nuanced than your platform attribution. They’ll use a simple last-touch or first-touch model. But they will provide a valuable gut-check on the numbers you see in platform.
Multi-channel Attribution
Multi-channel attribution is the practice of divvying up all your sales across different sources of traffic. It’s an exercise in distributing credit.
Google Analytics’ traffic reports are a version of multi-channel attribution. This is a great place to build your model. But heads up: the out-of-the-box session report is ‘last non-direct click’—this heavily biases to bottom of the funnel. Here’s how I rate GA4’s multi-channel reporting:
- Okay: Traffic acquisition: Session default channel group
- Better: User acquisition: First user default channel group
- Best: Data-driven report (under Advertising > Attribution > Attribution models)
Use one of these to attribute conversion value across channels.
Bringing It All Together: The Certainty Gap
You’ve dialed your in-platform data, validated sources with sales or CRM records, and picked a multi-channel model in GA4. Nicely done.
Now it’s time to report on monthly results. How do you frame the data? Which attribution source do you choose?
Here’s a quick framework:
You need a single source of truth. Otherwise, people will go insane, and you won’t have alignment.
If your sales/CRM records (e.g., Shopify, Hubspot, Amplitude, etc.) have strong attribution features, use them. You’re going to look better when you can tie conversion data to the real customer records.
If your sales/CRM records are too basic, use multi-channel or single-channel—depending on how much paid is in your mix. If your marketing mix is focused on ads, you should focus on your single-channel attribution model.
Last step, in your reports, compare your source of truth to the other two attribution models. The difference between them is your certainty gap.
Multi-channel attribution (649 conversions from paid search)
Single-channel attribution (649 conversions from paid search)
CRM attribution (619 conversions from paid search)
Example: Attribution in 3 sources (some data redacted). Each source reports slightly different numbers (642, 649, 619). This is normal. This business uses their CRM attribution as the source of truth, but could reasonably say, “we believe Paid Search drove at least 619 conversions based on our attribution model, since other models report higher conversions.”
You will always have a certainty gap.
The 3 methods are based on different foundations. They are different windows into the house. Your job is to define the gap between them, hold its tension, and make decisions informed by it.
If all three models give Meta the same amount of credit, that’s easy. You can be extremely certain that the ROAS you’re reporting in Meta is correct and spend accordingly.
If your CRM is your source of truth and your in-platform data reports 20% higher conversions than the CRM, that’s helpful information. You know there’s a good chance your CRM reporting is conservative. You don’t necessarily need to up your spend, but when you present your numbers, you can say, “this is a conservative estimate.”
Alternatively, if your multi-channel model attributes 20% fewer conversions to Meta than Meta does, you may want to spend more conservatively on Meta and expectation-set around that number internally.
Budget based on your source of truth, but make decisions with the certainty gap in mind. If you do this, you are ahead of 95% of businesses.
Want to go deeper on attribution and CAC modelling? Get my free guide, including what’s changed in 2024, tracking checklists, and industry benchmarks.
Looking to grow your business and connect with the right marketing partners?
We have vetted over 1000 marketing partners and can match you with the right partner the first time.
