Here is a situation you will meet in your first year. You add up the sales that Google Ads says it drove, the sales Meta says it drove and the sales your email tool says it drove. The total is bigger than the number of sales the business actually made.
Nobody is lying. Each platform is answering the same question, “did I help?”, in its own favour. Attribution is the set of rules that decides who gets credit, and understanding it will save you from a lot of confident-sounding nonsense.
▸ Definition — Attribution: the method used to assign credit for a sale (or lead) to the marketing touchpoints a customer saw or clicked before buying.
One customer, four touchpoints
Let’s follow one person. Sara sees your Instagram ad on Monday and doesn’t click. On Wednesday she searches your brand name and clicks a Google ad. She browses, leaves, and gets your abandoned-cart email on Thursday. On Saturday she types your website address directly and buys.
▸ Example — Sara’s path: Instagram ad seen → Google ad clicked → email opened → direct visit and purchase. Diagram: four boxes left to right, the purchase box in orange.
Which of those made the sale? The honest answer is “probably several of them, in different ways”. But reports need a number, so we use a model.
The common models
Think of a football goal. Who gets credit: the striker who scored, the player who made the final pass, or the defender who started the move? Each answer is defensible. Attribution models are just different answers to that question.
| Model | Who gets credit for Sara’s sale | What it favours |
|---|---|---|
| Last click | The direct visit (or the email, if direct is ignored) | Channels that close: brand search, email, retargeting |
| First click | The Instagram ad (if a view counts) or Google ad | Channels that introduce: social, display, video |
| Linear | Each touchpoint gets 25% | Nobody in particular; it spreads credit evenly |
| Time decay | More credit to touches closer to the purchase | Closing channels, with some credit for earlier ones |
| Data-driven | Credit based on patterns across many customers’ paths | Whatever the model learns, within one platform’s data |
▸ Margin note — Google Analytics 4 uses data-driven attribution by default. It’s smarter than last click, but it can only see what GA4 can track. An ad someone saw but never clicked is usually invisible to it.
Why the platforms disagree
There are three main reasons the numbers never add up:
- Each platform only sees itself. Meta can see that Sara viewed an ad; it can’t see her Google click. So Meta credits itself for the sale, and so does Google.
- Different windows. A platform might count a sale within 7 days of a click or 1 day of a view. Another uses 30 days. Longer windows claim more sales.
- Views versus clicks. Some platforms count “view-through” conversions, sales from people who saw an ad but didn’t click. That’s sometimes fair, and often generous.
None of this is a scandal. It’s just what happens when every referee also plays for one of the teams.
How to report honestly anyway
You can’t make attribution perfect. You can make your reporting honest. Four habits help:
- Pick one source of truth for totals. Use your shop or CRM’s actual orders for “how many sales did we make”, not the sum of platform claims.
- Use platform numbers for comparisons within the platform. They’re useful for deciding which ad or campaign is better than another, even if the absolute numbers are inflated.
- Say which model and window you used. “Google Ads, data-driven, 30-day click” takes five seconds to write and prevents a lot of arguments.
- Test when the stakes are high. If a channel takes a large budget, turn it off in one region or for a few weeks and see what happens to total sales. That’s the beginning of incrementality, which is a topic of its own.
▸ Definition — Incrementality: the sales that would not have happened without the marketing. Attribution shares out credit; incrementality asks whether there was anything to credit at all.
Quick check · optional
Take it to work
Principles
- 1Every attribution model is a simplification. Choose one knowingly, not by default.
- 2Platform-reported sales will add up to more than real sales. That’s normal.
- 3Use real orders for totals and platform data for comparisons.
- 4Label the model and window on every report.
Questions to ask on Monday
- Which attribution model and window do our reports use?
- Do our platform-reported sales add up to more than our actual orders? By how much?
- Have we ever tested what happens to total sales when we pause a channel?
If attribution feels slippery, that’s because it is. Everyone working in marketing is dealing with the same uncertainty. The marketers people trust are the ones who say so clearly.