Email takes the credit for customers you already had

A home furnishings retailer closed £890,000 last quarter. That is the figure in the accounts, and it is not in dispute. Their email platform reported £312,000 of attributed revenue over the same period. Google Ads claimed £280,000. Meta claimed £190,000.

Those three add to £782,000, which leaves £108,000 for everything else the business does: all of organic search, all of direct traffic, the affiliate programme, the showroom, the phone, and every customer who came back because a friend recommended them. Nobody believes that, including the people who pulled the reports.

Most of the argument about this lands on the paid channels, because that is where the money goes out. It is the wrong place to start. The largest single claim on that list is the email platform's, and it is the one nobody interrogates, because email is cheap and the number is flattering and the whole thing feels like a rounding error on a budget nobody is defending.

It is also, structurally, the claim most likely to be counting sales that were going to happen anyway.

What attributed revenue means when an email platform says it

Attributed revenue is the value of orders placed by people who received, opened or clicked a message within a configured window before ordering. It records a sequence, not a cause. The platform can see that a message preceded a purchase; it cannot see whether the purchase needed the message.

This is not a criticism of any particular vendor. It is what the word attribution means everywhere it appears in marketing software, and email platforms are more explicit about their mechanics than most. The problem is the gap between what the mechanism does and what the number is read as meaning by the time it reaches a quarterly review.

Every one of those is a reasonable question. Only the first is the one being answered, and it is the least interesting of the three.

Attributed revenue answers which orders followed a message. It is read as which orders the message produced. Those are different numbers and nobody labels the difference.

The window is a setting, not a fact about your customers

An attribution window is the length of time after a message during which an order is credited to it. It is chosen by whoever configured the account, it can be changed in an afternoon, and changing it changes reported revenue without anything changing in the business.

This is the part that surprises people who have not been in the settings. The share of revenue a channel appears to drive is partly a property of a form field. Two businesses with identical sends, identical customers and identical sales can report materially different email revenue because somebody picked a different number on a configuration page.

Five days is a defensible default. It is short enough to exclude most coincidence and long enough to catch a considered purchase. The issue is not the number; it is that a reader of the resulting report has no idea a number was chosen at all.

And note what the window is measured from. A conversion period that begins on receipt rather than on engagement means the clock starts whether or not the recipient ever did anything. Send to a list of forty thousand people on a Monday and you have, by construction, placed a claim on every order any of those forty thousand place before Saturday.

An open stopped being evidence of a person in 2021

Open tracking works by embedding a remote image that loads when a message is displayed. Apple Mail Privacy Protection downloads that content in the background regardless of engagement, so an open event no longer establishes that a human looked at anything.

This is documented by Apple in plain language, and the wording matters because it describes a machine doing the thing the metric assumes a person did.

The immediate consequence is well known: open rates inflated overnight and stopped being comparable with anything before them. Many teams responded sensibly, demoted the metric, and moved on to clicks.

The consequence that gets less attention is what it does to revenue attribution, because an open is not only a metric. In a platform that attributes on opens as well as clicks, an open is an attributable event — and that means an automated background fetch, performed by Apple's infrastructure on behalf of a person who may never have seen the message, can be the event that hands email the credit for a sale.

That the setting exists is to the platform's credit. Almost nobody has looked at it, which is not.

If a machine can generate the event that credits a channel with revenue, the resulting figure is a measure of your settings as much as of your marketing.

Email reaches the people who were already going to buy

Email's deliverable audience is overwhelmingly made up of existing customers and people who have already identified themselves as interested. That is the channel's great commercial strength and the precise reason its attribution overstates its influence.

Consider who is actually on the list. People who bought something. People who created an account. People who abandoned a basket, which means they had already chosen an item and reached the checkout. People who asked to be told when something came back into stock. The list is a roster of demonstrated intent, assembled specifically because demonstrated intent is the best predictor of a future purchase.

Now run any window-based attribution over that population. A channel that emails its existing customers weekly will be present in the five days before a very large share of their orders, whatever those emails said, and whether or not they were read. Presence is not influence, but presence is what gets measured.

The flows make this sharper still, because the best-performing ones are triggered by the buying signal rather than the cause of it:

• The abandoned basket message goes to somebody who selected an item and reached a checkout. A proportion of them always come back on their own.
• The back-in-stock alert goes to somebody who asked to be told when they could buy the thing they had already decided they wanted.
• The post-purchase sequence goes to somebody who has just demonstrated they buy from you, immediately before the window in which repeat purchases are most likely.
• The browse abandonment message goes to somebody who was, minutes earlier, looking at the product page.

Every one of those is worth sending. Several are among the highest-return things a retailer can automate. But each is triggered by an intention that existed before the message, and each will be credited with whatever follows it inside the window.

This is the counterfactual problem, and it is not solvable by better tracking, because there is nothing left to track. The information required — what this person would have done if you had sent nothing — does not exist in any log. It can only be produced by withholding the message from a comparable group and measuring the difference.

Why the channel reports add up to more than the business

Each platform reports the conversions it can see within its own window and subtracts none of the ones another platform is also claiming. Overlap is guaranteed, nobody nets it off, and the totals routinely exceed recognised revenue.

Return to the retailer at the top. A customer searches for a product, clicks a Google ad, browses, leaves, receives that evening's campaign email, clicks it, and orders. That is one order and one amount of money. Google Ads counts it. The email platform counts it. Both are behaving correctly according to their own documentation, and both figures land in the same spreadsheet as though they were additive.

Email sits in an unusual position in that pile-up. It is frequently the last touch, because it is the channel that arrives while somebody is deciding, and last-touch logic rewards whatever was nearest the finish line. It also has the longest reach into your existing customer base, so it appears in paths that paid media never touched at all.

The result is a channel that looks enormous under last-touch, respectable under most multi-touch models, and considerably smaller under a holdout test. Which of those is the truth depends on a question nobody in the review meeting has asked out loud: compared with what?

Reconcile against the sales you actually made

Reconciliation means starting from the orders your finance system recognises and working backwards to the marketing that preceded them, rather than starting from each platform's claim and hoping the claims are compatible. It is the only method that cannot double-count, because it begins from a number that cannot be counted twice.

The procedure is unglamorous and it is mostly clerical:

1. Export the closed sales for a period from the system finance trusts — the order, the date, the value, and an identifier for the customer such as an email address or a phone number.
2. Normalise those identifiers, because a sales file and a marketing list will not agree about capitalisation, whitespace, plus-addressing or the format of a phone number.
3. Match each sale to the marketing touches that preceded it, from every channel, not only the one whose report you are checking.
4. Classify what you find: sales with a single clear source, sales where several channels were present, and sales where nothing confident can be said.
5. Put the third group in an honest bucket and leave it there. It is usually larger than anybody expects, and shrinking it by guessing is how the original problem was created.

What this produces is not a tidier attribution model. It is a smaller set of claims you can defend, plus an explicit measure of how much of the business you cannot explain. Both of those are more useful to a finance director than a set of channel reports that sum to 114% of revenue.

CloseRev exists to do exactly this one job: take a closed-sales file and a file of calls or lead sources, normalise the identifiers, match them, and report what each channel was present for — with the unmatched remainder shown as Direct or Unknown rather than distributed among the channels that happened to be nearby. It will not tell you what email caused. Nothing will, short of a holdout. It will tell you which sales email was anywhere near, which is the honest version of the same question.

For the causal question there is one method, and it has been available the whole time: hold the message back from a randomly chosen slice of the list and compare. It costs you the revenue that slice would have produced, which is precisely why it is worth doing occasionally rather than never. A quarterly holdout on one large flow will tell you more about what email is worth than any attribution report you will ever run.

None of this means email is not working

Email is cheap, owned, and genuinely effective at bringing existing customers back. The argument is not that the channel underperforms. It is that the number reported next to it is not a measurement of its contribution, and treating it as one leads to bad decisions in both directions.

Overstating email is not a harmless flattery. It shows up as real decisions: budget moved out of acquisition and into retention on the strength of a figure that was always going to look good; an agency judged against a paid number that is being quietly out-competed by a channel measuring itself on a friendlier basis; a forecast built on the assumption that sending more will produce proportionally more revenue, when much of the attributed total was never elastic to send volume in the first place.

It cuts the other way too. A team that discovers its email revenue was overstated sometimes overcorrects into treating the channel as worthless, which is equally unsupported by the evidence and considerably more expensive, because the flows they switch off are the ones with the genuine incremental effect buried among the coincidences.

The useful posture is narrower and duller than either. Keep sending. Stop reading the attributed revenue figure as an amount of money the channel produced. Find out what the window is set to and who set it. Check whether machine opens are counted. And once a quarter, withhold something from a random slice and find out what actually changes.

The question is never how much revenue email is attributed. It is how much of that revenue would have arrived anyway — and no report you already have is capable of answering it.

Questions people actually ask

What does attributed revenue mean in an email platform?

It means revenue from orders placed by people who received, opened or clicked a message within a configured window before the order. It does not mean the message caused the order. The platform observes a sequence and reports it; whether the sale would have happened anyway is a question the data cannot answer.

Why is email's attributed revenue share usually so high?

Because email's audience is mostly people who have already bought from you or already intended to. A channel that reaches existing customers will sit in the path of a large share of orders regardless of its influence, and any window-based attribution will credit it for them.

Does Apple Mail Privacy Protection break open rates?

It makes them unusable as a measure of human attention. Apple states that Protect Mail Activity downloads remote content in the background by default, regardless of whether you engage with the email, so the tracking pixel fires whether or not a person ever looked at the message.

Can an automatic open still credit email with a sale?

It depends on settings you control. Klaviyo attributes on opens as well as clicks, and offers an option that removes Apple Mail Privacy Protection opens from attribution — which means whether machine opens count toward your reported email revenue is a configuration choice, not a fact about your business.

Do email and paid channels double-count the same sale?

Routinely. Each platform reports the conversions it can see within its own window, and none of them subtracts the ones another platform is also claiming. Add the channel reports together and the total frequently exceeds what finance recorded for the same period.

How do I find out what email is really worth?

Reconcile against closed sales. Take the orders your finance system recognises, match them back to the marketing touches that preceded them, and look at what each channel was present for rather than what each platform claims. The honest version leaves a large bucket of sales with no confident source, which is a finding rather than a failure.

Found this useful?

• Share on X
• Share on LinkedIn
• Share on Reddit
• Share on Email