Email attribution assigns credit for a conversion to an email. It sounds like measurement and it is mostly a set of decisions — which touch counts, over what window, and what to do when several campaigns could claim the same sale.
Different reasonable choices produce numbers that differ by several times for the same underlying reality. So the useful goal is not the correct number. It is a documented, consistent number that everyone using it understands.
The two windows
Click-through attribution. Someone clicked a link in your email and converted within a window. Conservative, defensible, and it undercounts — it misses everyone who read the email, did not click, and bought later.
View-through attribution. Someone opened the email and converted within a window, without clicking. This captures real influence and it is now unreliable, because Apple Mail Privacy Protection registers opens that never happened. View-through attribution built on proxy opens credits email for conversions from people who never saw the message.
For most senders in 2026, click-through with a fixed window is the defensible default, with the understanding that it undercounts.
The window matters more than the model
A 1-day window and a 30-day window on the same campaign can differ by several times. Neither is wrong; they answer different questions.
Seven days is a reasonable default for most businesses. For considered purchases with long cycles, longer. For impulse purchases, shorter. The only real rule is that you pick one, write it on every report, and never change it in the middle of a comparison — because a window change swamps any real performance difference you might have been trying to observe.
The deduplication problem
Someone gets your newsletter on Monday, clicks. Gets an abandoned cart email on Wednesday, clicks. Buys on Wednesday.
Which campaign gets the sale? If both claim it, your total attributed revenue exceeds your actual revenue, which is how email programmes end up reporting more revenue than the business made.
Pick a rule and apply it everywhere:
- Last click within the window. Simplest and most common. Favours triggered flows over broadcasts.
- First click within the window. Favours the campaign that started the journey. Rarely used.
- Split credit. Fairer in principle, harder to explain, and the fractional numbers invite argument.
Last click is the pragmatic answer for most teams, with the caveat that it systematically undervalues the newsletter that keeps the list warm enough for the abandoned cart email to work.
Why your platform and your analytics disagree
They always do, and the gap is usually large. The reasons:
- Different windows. Your platform's default and your analytics tool's default are rarely the same.
- Different identity. Your platform knows the subscriber; your analytics knows the browser. Someone who clicks on their phone and buys on a laptop is one person to your platform and two to your analytics.
- Cross-device and cookie loss. Browser restrictions have made site-side attribution progressively worse at connecting sessions.
- Different models. Platform-side attribution is usually generous; last-non-direct-click in analytics is usually stingier about email.
Neither is right. Pick one as the source of truth for decisions, note the other for context, and stop trying to reconcile them.
The only method that answers the real question
All of the above measures correlation. The question you actually care about is causation: how much of this revenue would not have happened without the email?
A holdout answers it. Exclude a random 5–10% of the eligible audience from a campaign or a flow, then compare revenue per person between the holdout and the treated group. The difference is incremental. Everything above it was revenue you would have earned anyway.
Holdouts feel wasteful, which is why almost nobody runs them. They are also the only thing that tells you whether an automation is doing anything. Run one permanently on your highest-volume flow — the abandoned cart is the usual candidate and often the most surprising, because a meaningful share of those people were going to complete the purchase regardless.
The cost is a few percent of that flow's revenue. The return is knowing whether the flow is worth maintaining. That trade is why holdouts matter more than any attribution model, and it is what makes an ROI figure defensible.
What to write on the report
One line, every time: "Revenue attributed to email within 7 days of a click, last-click, deduplicated across campaigns and flows."
It looks pedantic and it prevents the two most common failures — comparing numbers built on different rules, and quietly widening the window in a disappointing month. It is also what makes client reporting survive scrutiny.
Frequently asked questions
What attribution window should I use for email?
Seven days from click is a reasonable default. Longer for considered purchases, shorter for impulse ones. The specific choice matters far less than keeping it fixed, since changing it moves the number more than most campaign improvements do.
Why does my email platform report more revenue than my analytics?
Different windows, different models, and different identity resolution. Your platform knows the subscriber across devices; your analytics knows browser sessions and loses people who click on a phone and buy on a laptop.
Should I use view-through attribution for email?
It is hard to justify now, because open tracking is contaminated by privacy-protection proxy loading. Crediting conversions to opens means crediting some conversions to messages nobody read.
How do I prove email caused a sale?
Run a holdout. Exclude a random slice of the eligible audience and compare revenue per person against the treated group. The difference is incremental effect. No attribution model can produce this, because models can only allocate credit among touches that occurred.
