Picture the renewal meeting. The marketing manager who hired you is happy, and the finance director beside them asks the only question they came to ask: "what did we get for this?" Rankings don't answer it, and neither does a traffic chart. Proving the return honestly means showing which links in the chain you measured, which you estimated, and which you had to assume.
The chain from a ranking to a sale
A ranking earns nothing on its own. It has to turn into an impression in search results, then a click, then a session on the site, then an enquiry or an order, then a customer, then revenue. Each link is measured by a different system. ROI conversations go wrong when someone quotes a number from one link as if it proved the whole chain.
Rankings live in your rank tracker. Impressions and clicks live in Google Search Console's Performance report. Sessions and enquiries live in the client's analytics, usually GA4. Customers and revenue live in the client's CRM or sales system, which you may never have seen. GA4 for SEO and Google Search Console essentials cover the middle of the chain. The end of it belongs to the client.
Start with a table that has one row per link: what it is, where it's measured, and who owns the number. A blank in the revenue row is the honest state of most engagements. Writing it down is what lets you fix it.
Why you can't prove causation, and what you can show
Organic traffic rises and falls for reasons that have nothing to do with you. Seasonality, a Google update, the client's own advertising and a competitor's site going dark all move it. You can't run a controlled experiment on a live business, so you can never say with certainty that your work caused a given rise.
What you can show is specific. The pages you changed moved, and the pages you didn't change didn't. The timing lines up with the work, not with a known Google update. The keywords you targeted are the ones that climbed. The enquiries came through the pages you built, not the homepage. Each point is evidence, and a page of it persuades a finance director more than a causal claim they can see through.
The language follows from the evidence. "Organic enquiries rose after the service pages went live, and the new pages account for most of the rise" is a sentence you can defend. "Our SEO generated these enquiries" isn't, and a sharp client will test it by asking what else changed that month. Reading rank movements honestly goes deeper on separating your effect from everything else.
Build the model with the client's own numbers
The model turns traffic into money using two numbers only the client has: what share of enquiries become customers, and what a customer is worth. Never guess them. Ask the client, write down who gave you each figure and when, and use their number even if you suspect it's conservative. A model built on their inputs is one they can't dismiss.
Here is an invented worked example for a fictional store selling trail shoes, with figures chosen to show the arithmetic, not as a benchmark for anything. Organic sessions are up by 400 on the same month last year, and the store's analytics show that 2 in every 100 sessions end in an order, so the rise is worth 8 extra orders. The average order is £90, so those orders are worth £720 against a monthly fee of £600. It's a positive month, by a small margin.
Now run the arithmetic again on a deliberately low reading of the traffic rise, since some of it is seasonal, and present both results as a range. The low end is the number you stand behind. The high end is what's plausible. A month where the model comes out negative is a real result too, and early months often will, because SEO's costs arrive before its traffic does.
Surviving the question in the room
When the finance director asks what they got, answer in three parts, in order. First, what is measured: the movement in rankings and clicks, then enquiries, each with its source. Second, what is estimated: the revenue model, with the client's own inputs and the range. Third, what is missing and what it would take to close the gap. Usually it's a source field in the CRM so each customer records where they came from, or call tracking for a business that sells by phone (tracking calls, directions and clicks covers the local version).
Sometimes the honest answer is "not yet". Early in an engagement the technical fixes and new pages are in place but the traffic hasn't arrived, so the measured return is negative. Say so. Show the leading indicators, such as pages getting indexed and positions climbing, and name the month you expect the chain to complete. A client who was told the truth early tends to renew. One shown an inflated number early finds out, and the finding out is what ends the contract.
Frequently asked questions
How do you grow an SEO business?
Keep the clients you win. Agencies lose accounts when the client can't see what the fees bought, so a report that names the source of every link, shows a range and admits the blanks does more for retention than a bigger traffic chart. Clients who were told the truth early tend to renew.
What does ROI mean in social media?
ROI is return on investment: what came back compared with what you spent. In SEO the spend is usually your fee, and the return is the revenue from organic customers. The same caution applies on any channel, because a click or a like isn't revenue until it's tied to a customer through the chain.
Is SEO replaced by AI?
Clients now ask a newer version of the ROI question about AI assistants, so the need to measure hasn't gone away. The rule is the same: report what you can measure, label what you're estimating, and don't claim causation you can't show. Reporting AI visibility covers how to do it with measured data only.
What to take away
- SEO's return is a chain from ranking to revenue, each link measured by a different system. An honest ROI report names the source of every link and admits the blanks.
- You can't prove your work caused a rise, but you can show that the pages you changed moved, the timing fits, and the enquiries came through those pages.
- The revenue model uses the client's own close rate and order value, presented as a range with the low end as the number you defend.
- When the chain isn't complete yet, say "not yet", show the leading indicators, and name what it would take to close the gap.
Next
Clients now ask a newer version of the same question about AI assistants. Reporting AI visibility shows how to answer it with measured data only.