Email marketing software ROI: a worked model in contribution, not gross turnover

What to take away

The figures below are illustrative inputs for a small online shop, not market data. Replace them with your own finance and order records. The method is what carries over: the same six steps work for Mailchimp, Klaviyo, Brevo or any other platform.

Worked example: the ROI model step by step

Break-even sits at £1,960 of monthly contribution, which is £23,520 divided by 12. At a contribution of £61.20 per order (£7,344 divided by 120), that is about 32 incremental orders a month. The model therefore breaks even well below the 120 orders assumed.

The 20% fall test and a range

Now cut incremental orders by 20% to 96 a month. Net revenue becomes £7,344 after refunds, contribution £5,875, and the annual figure £70,502. ROI is (£70,502 - £23,520) / £23,520, or 2.00 (200%). The case survives, with a lower return.

Report a range, not a single point. With 60 incremental orders a month the same arithmetic gives an ROI of about 87%; with 150 it gives about 368%. The 120-order central case is 275%. Choose the low and high orders from your holdout result, not from the platform report.

Which costs belong in the economic denominator

The denominator is not the invoice. It is every cost the decision commits, through to exit. Platforms such as Mailchimp, Klaviyo and Brevo publish tiered plans that rise with contact numbers, so take the subscription line from the live pricing page or your quote, and see Klaviyo vs Mailchimp for a UK comparison.

Under the Privacy and Electronic Communications Regulations 2003 (PECR), the ICO expects consent to be specific, informed and recorded, so the cost of keeping that evidence belongs in the denominator. The ICO publishes direct marketing guidance on those rules.

A positive modelled return does not solve a cash shortfall, and an affordable invoice does not prove the investment creates value.

On VAT and cash timing, cash leaves on the invoice date plus your payment terms, not in the period the cost belongs to. If you are VAT-registered you can usually reclaim VAT on the subscription, so model costs excluding VAT. Ask finance which basis your organisation uses.

Contact-tier pricing moves the subscription line faster than most buyers expect, so work through the pricing models worksheet before fixing that figure.

Keep the counterfactual honest

State the decision, population, intervention, comparison, outcome, time period and geography before you write the equation.

If there is no credible counterfactual, describe an observed association rather than a causal return. HM Treasury's Magenta Book separates monitoring from impact designs that test whether an intervention caused change.

A practical version for email is a holdout: keep a random share of eligible contacts, for example 10%, out of the campaign for several weeks, and compare their orders with the mailed group. A shop with a small list needs a longer period to get a usable difference.

Platform models often assign an order to a recent email touch. Keep that figure, labelled platform-reported attribution with the vendor's definition. Our note on email measurement mistakes explains why the two figures cannot share a numerator.

Pre-specify seasonal periods, concurrent promotions, channel overlap, customer eligibility and data-quality exclusions before you look at any result.

Where to source observed cost and margin data

Pull each line from a system, not from memory.

Keep the platform report for platform-attributed revenue only.

Set a decision rule and stop conditions

Name the model owner, the evidence expiry, the review date and the maximum exposure.

Pause spending when cost exceeds its ceiling, evidence quality falls, or a privacy, PECR, accessibility, security or advertising gate fails; the campaign governance guide shows where those gates sit. A positive dashboard does not override those gates.

Look only forward. HM Treasury's Green Book treats sunk cost as unrecoverable and asks appraisals to test sensitivity and optimism bias. Borrow that reasoning for a retailer's email budget, not any public-sector result.

After the test period, replace the illustrative inputs with observed costs and decide to continue, modify or stop.

Common questions

Is platform-attributed revenue ever usable?

Yes, as an operational signal, provided the vendor's definition travels with the number. It is not evidence of incremental revenue, because it assigns credit without a counterfactual.

What if finance has not defined a contribution margin?

Ask finance to name the costs deducted from revenue for the decision period. Until then, model the margin as a range and label it an assumption.

Does the denominator include spend already committed?

No. Sunk expenditure is reported separately from future decision costs, because only the latter can change the choice you are making now.