How to Calculate Email Marketing ROI
A calculation that includes labour, and the reason the famous return-per-dollar figure is almost never your figure.
On this page
The short version
- The famous return-per-pound figures exclude labour, which is usually the largest cost in the programme.
- They are also survivor-weighted: they come from senders using paid platforms with enough volume to be surveyed, which is not the average sender.
- Calculate yours from your own figures, including time. It will be lower than the headline claim and it will be a number you can defend.
Email marketing is widely quoted as returning some large multiple for every unit spent. The figures circulate without their methodology attached, and the methodology is where all the interesting parts are.
Two things are usually true of them. The cost side counts platform fees and not the people writing the email. And the sample is drawn from senders large enough to be surveyed by an industry body, which excludes most programmes.
None of that makes email a bad channel — it is a good one. It makes the published number the wrong input to your own decision.
The calculation, with the missing line put back
Return is attributed revenue minus total cost, over total cost. The attributed part depends on your attribution model, which is a choice you should already have made. The cost part is where the honest work happens.
What belongs in the cost side
| Cost | Often counted? | Note |
|---|---|---|
| Sending platform | Yes | The only line most people include |
| Email verification | Sometimes | Tracks acquisition, not list size |
| Deliverability tooling | Rarely | Small, and cheap against an incident |
| Writing and production | Rarely | Usually the largest single line |
| Design and template work | Sometimes | Front-loaded, then near zero |
| Automation build and upkeep | Rarely | One-off per flow, plus maintenance |
| Analytics and reporting time | Almost never | Real hours, every month |
| Management and approvals | Almost never | Larger than anyone expects |
If the writing is absorbed into somebody's existing role it is still a cost — it is being paid out of whatever that person is not doing instead.
Working it out
Fix the attribution model first
Every revenue figure downstream depends on it, and changing it later invalidates the comparison you were building.
Count the hours honestly, for one month
Writing, building, reviewing, approving, segmenting, testing, reporting. Multiply by a loaded hourly cost. Most programmes find this is more than the platform fee, often several times more.
Add the tooling
Platform, verification, monitoring, testing. Annual costs divided by twelve.
Take attributed revenue for the same month
From the system you named as the number of record, not from whichever reports the higher figure.
Compare against the alternative use of that time
This is the step that makes the number decision-useful. A return of four to one is excellent unless the same hours in another channel would have returned six.
What a defensible number gets you
A figure calculated this way is lower than the published claim and is the only version that survives a finance conversation. It also tells you something actionable, which the headline figure never does: where the cost actually is.
For most programmes the answer is that production dominates, which points at templates, reusable structures and flows rather than at platform pricing — the thing people usually try to optimise first.
The figure that is more useful than return
Return on investment answers whether the programme is worth running, which is a question most people have already decided. Cost per outcome answers where to spend the next hour, which nobody has decided.
Divide total monthly cost by whatever the programme produces — orders, enquiries, activations — and then do it separately for flows and campaigns. The gap between those two numbers is usually large, and it points directly at what to build next.
It also survives an argument about attribution better than a revenue figure does, because a count of outcomes is less sensitive to the model than the money attached to them.
ROI check
- Labour is in the cost side
- The attribution model is stated alongside the number
- Flows and campaigns are calculated separately
- The period is long enough to include a slow month
- The comparison is against an alternative use of the same resources
- The number is not a published industry figure
Frequently asked questions
How do I value a subscriber who has not bought yet?
Only with a cohort measurement: take subscribers who joined in one month and track what they spend over the following year. Anything faster is a guess wearing a decimal point.
Should list growth cost count against email?
Acquisition cost belongs with acquisition. Keep them separate or you cannot tell whether the problem is that the list is expensive to build or that the programme is not monetising it.
Our return looks poor. What first?
Check the cost side is not being inflated by production for sends nobody reads, and check the revenue side is not being suppressed by a last-click model. Those two account for most surprisingly low figures.