Email consistently reports the best return of any channel we run, by a wide margin, and it receives the least attention internally. Both of those facts deserve examination, because part of the performance is real and part is an artefact.
Why it genuinely performs well
The audience opted in. Everybody on the list took an action to be there, which is a filter no paid channel applies.
The cost is close to zero at the margin. Sending to fifty thousand people costs slightly more than sending to five thousand, so any revenue attributable produces a very high ratio.
You control the channel. No ranking algorithm decides whether it is delivered, subject to deliverability, and no platform can change the terms.
And it reaches people repeatedly over time, which for anything with a long consideration period matters enormously.
Why the numbers overstate it
The part that gets less discussion and that we had to confront.
The list is composed of people who already know you, many of whom are existing customers. A proportion of what email is credited with would have happened anyway.
This is the same incrementality problem that affects branded search, and email is the channel where it is largest, because the audience is by definition already engaged.
Attribution also favours email structurally. A click from an email is a clean, trackable, first-party interaction, so it gets credited reliably while less trackable channels do not.
We ran a holdout — a randomly selected portion of the list receiving nothing for a period — and the incremental revenue was meaningfully lower than the attributed revenue. Still good, and not as good as the dashboard said.
What actually moves the numbers
Having tested a fair amount, the things that mattered were not the things we spent time on.
Segmentation, by a distance. Sending relevant things to smaller groups outperformed sending one thing to everybody, consistently and substantially.
Triggered messages based on behaviour outperformed scheduled campaigns by a large margin. Abandoned baskets, post-purchase sequences, re-engagement after inactivity.
Frequency, which is a genuine variable with an optimum rather than a maximum. Increasing frequency increases short-term revenue and increases unsubscribes, and the crossover is findable by testing.
Subject lines mattered less than expected once we tested properly at adequate sample sizes. The differences were real and smaller than the folklore suggests.
And design mattered hardly at all. Plain text messages performed comparably to designed ones in most of our tests, and better in some.
Deliverability is the thing that can end it
The risk nobody thinks about until it happens.
Messages that reach a spam folder do not exist. Deliverability depends on sender reputation, which depends on engagement, complaints, bounces and authentication.
The practical measures are unglamorous. Authenticate properly. Remove people who have not engaged in a long period, which feels like destroying an asset and protects the rest of the list. Make unsubscribing easy, because the alternative is complaints, which are far more damaging. And never buy a list, which will damage reputation quickly and durably.
The list-cleaning point is the one most people resist. A list of a hundred thousand of whom sixty thousand never open is worse than a list of forty thousand, because the sixty thousand are actively harming delivery to the rest.
The regulatory position
Worth stating because it varies and the penalties are real.
Consent requirements differ substantially between jurisdictions, and the stricter regimes require genuine opt-in rather than pre-ticked boxes or inferred consent.
Records of when and how consent was obtained matter if anybody asks.
And rules about what counts as a marketing message versus a transactional one are narrower than marketers would like.
This is an area where getting proper advice is worth it rather than reading a summary like this one.
Why it stays neglected
My theory, having watched it in three organisations.
It is not new, so it does not attract the attention that new channels do. Nobody builds a career on being excellent at email.
It has no external vendor pushing it, unlike every media channel, so nobody is presenting quarterly to your leadership about it.
And it is genuinely tedious work — list hygiene, segmentation logic, template testing across clients — with none of the visible output of a campaign.
Which is why it remains the most reliable underinvested channel in most organisations, and why the answer to what to do next is so often to fix the email programme first.
The one-list problem
A structural point that took us a while to fix.
We had a single list and sent everything to all of it, with segmentation applied inconsistently on top.
Moving to preference-based subscription, where people choose what they hear about, reduced volume and improved every other metric.
It also reduced unsubscribes substantially, because people who were only interested in one thing had previously had a binary choice between everything and nothing, and chose nothing.
Giving people a way to receive less is counterintuitive and it keeps them on the list.