"CRO delivered a big ROI." "Retention is our best channel." Both sentences can be true and still tell finance nothing. This piece shows how to turn a conversion-rate change or a retention programme into margin, payback and cash. It is the method side of ROAS looks good, profit doesn't.
Two definitions first. CRO (conversion rate optimisation) is the practice of getting more of a site's existing visitors to take a desired action, using analytics, qualitative research and controlled tests. Retention is the share of customers who buy again within a given period, usually tracked by cohort.
Why don't vendor ROI claims transfer to your P&L?
Because the number comes from somebody else's business. ROI (return on investment) compares profit with cost. Many claims quote it without saying which profit, which cost, against what baseline.
Where do these claims appear? In our experience, mostly in sales material, including "top agencies" lists that agencies publish about their own market. Read such pages as marketing, not as independent rankings.
Five questions break most ROI claims:
- Is it revenue or margin?
- Which costs are included: tools, discounts, development time, the work itself?
- Compared with what? A control, or the period before?
- Was the effect measured on a test, or assumed from a rate?
- Does the claimant's volume, basket size and margin look like yours?
If the claim cannot answer them, it is a story. Pleasant, perhaps. Still a story.
A vendor ROI figure is only useful with its method: baseline, costs and control.
How do you put CRO in the P&L?
Start with contribution margin. Contribution margin, in our convention, is revenue minus the variable costs of an order (cost of goods, shipping, payment fees, returns, discounts) before marketing. It is our convention, not an industry standard.
For CRO the chain is short:
- Extra orders = visitors × the change in conversion rate.
- Extra contribution margin = extra orders × contribution margin per order.
- Net effect = extra contribution margin minus the full cost of the CRO work.
Why margin before marketing? Our view: the visitors were already paid for. The extra orders cost no extra ad spend, so the margin they bring is not reduced by marketing. If the change also brings in more traffic, count that spend separately.
Count the cost properly: the people doing the work, tools, development time and the tests that did not win. Losing tests are part of the price of the winning ones.
Watch the lines that sit inside margin. A change that lifts orders by leaning on discounts can raise revenue and lower margin. Returns do the same.
Example (illustrative): A shop has 100,000 sessions a month. Conversion moves from 2.0% to 2.2%. That is 200 extra orders. Contribution margin is 40 per order. Extra margin: 8,000 a month. The work costs 3,000 a month, all in. Net effect: 5,000 a month. Now suppose the extra orders came with a discount that cut margin to 25 per order. Extra margin falls to 5,000, and the net effect to 2,000. Same conversion lift. Different P&L.
A test result describes the test period. Before you book it as a run-rate, check it against what happens after launch. Is the lift still there?
How do you put retention in the P&L?
Retention in the P&L means incremental repeat margin, minus discounts, tools and the work. The word that matters is incremental. Repeat orders are not all caused by the programme. Some customers were coming back anyway.
An analytics vendor's guide defines customer lifetime value as average order value × purchase frequency × customer lifespan1. That formula runs on revenue. Our convention is to run it on contribution margin for decisions, so the discounts and returns are already inside it. A cohort view, where you follow each group of first-time buyers, shows whether repeat behaviour changes over time.
The cost side needs care:
- Discounts. A repeat-purchase offer is a cost per order, taken from contribution margin.
- Tools. Subscriptions and set-up, in the period you use them.
- The work. Strategy, build, content and maintenance.
Then comes the question that vendor figures skip. How many of those repeat orders would have happened without the programme? That is incrementality: the extra orders caused by the activity, over and above what would have happened anyway. A tool that credits every order after an open or a click counts customers who were coming back regardless. A holdout test answers it. Keep a random group of customers out of the programme and compare repeat margin per customer. Who may be emailed is a consent matter, and your lawyer decides that, not the test design.
Example (illustrative): A programme group of 9,000 customers makes 450 repeat orders in a period. A held-out group of 1,000 makes 30. The held-out rate predicts 270 for the programme group. The programme added about 180 orders. At 40 contribution margin each, that is 7,200. If the programme gave away 1,500 in discounts and cost 2,000 to run, the net effect is 3,700. The tool might have claimed all 450 orders.
How do payback and cash fit in?
Net effect per month is half the story. The other half is when the cash returns.
Payback period is the number of months it takes for margin, not revenue, to repay the cost. For a one-off build, divide the cost by the monthly net margin. For a recurring programme, look at how long the margin from a cohort takes to repay what you spent to win it back.
Example (illustrative): A one-off CRO build costs 12,000. It adds 2,000 a month in net margin. Payback is 6 months. If the effect fades after 4 months, it never pays back.
Cash adds timing. Discounts leave your margin on the day of the order. The repeat orders they are meant to produce come later. Stock, returns and payment terms can widen the gap.
The same discipline sits behind break-even ROAS. An article from a firm that sells fractional-CFO services defines break-even ROAS as 1 divided by contribution margin, and calls it algebra, not an estimate or a benchmark2. Start from margin. Then everything else is arithmetic.
What should you show the CFO?
One page, the same shape for CRO and retention:
| Line | CRO | Retention |
|---|---|---|
| Baseline | Conversion rate before the change | Repeat rate or margin per customer without the programme |
| Effect | Extra orders, from a test | Extra orders, from a holdout |
| Margin per order | Contribution margin before marketing, with returns and discounts | The same, minus programme discounts |
| Costs | Work, tools, development, losing tests | Work, tools, discounts |
| Control | Test and control groups | Held-out customers |
| Payback and cash | Months to repay the build | Months to repay the programme |
| Assumptions | What you assumed, written down | The same |
Show a range, not a point. Say which numbers are measured and which are assumed. A CFO forgives a wide range. A narrow range that falls apart in the next meeting is harder to forgive.
For ZFD: We show CRO and retention work in contribution margin, with the control and the costs on the same page. That is our convention and our view of how to read it. ZFD runs CRO audits and implements the changes, and covers retention and CRM end to end: audit, strategy, tool set-up and the email, CRM and lifecycle work itself. We do not promise a lift in advance. We show how the result was measured.