How Do I Forecast SEO ROI?

The only model that survives scrutiny

Four inputs, multiplied. Anything more elaborate is usually decoration over the same arithmetic:

1. Reachable demand. How many searches you could realistically appear for, sourced from your own impressions where possible.
2. Expected click-through rate at the position you expect to reach. This is the input that decides whether the forecast is honest.
3. Your organic conversion rate, from GA4, for the page types involved. Not a benchmark, yours.
4. Value per conversion: average order value, or pipeline value per lead multiplied by your close rate.

Demand times CTR gives clicks. Clicks times conversion rate gives conversions. Conversions times value gives revenue. Subtract the cost of the work and you have a return.

Getting each input honestly

• Demand: use your Search Console impressions rather than tool volume where you already appear. Impressions are a record; volume is an estimate.
• CTR: use your own CTR at each position band, which Search Console gives you directly. Published CTR curves describe averages across the whole web and are frequently wrong for a specific site.
• Conversion rate: segment GA4 to organic and to the relevant page type. A site-wide rate flatters informational pages and understates commercial ones.
• Value: use a figure finance already recognises. Inventing a lead value for the forecast is the fastest way to lose the room.

Forecast ranges, not numbers

A single number is a hostage. Present three scenarios and the assumption behind each:

State plainly which input is least certain. It is always the position assumption, and saying so before someone asks is what makes the rest credible.

Tracking the return once work starts

Forecasting is the easy half. The measurement fails more often, and always for the same reason: no baseline was recorded.

Record these on day one, because they cannot be reconstructed later:

• Organic conversions and revenue for the last 12 months, by month.
• Impressions, clicks and average position for the target pages.
• The full cost of the work: hours, tools, content production.

And start a history pipeline, because Search Console keeps only 16 months, which allows exactly one year-over-year comparison unless you capture it yourself.

A worked example, with the arithmetic visible

Numbers make the model easier to argue with, which is the point. Suppose a set of pages currently earns 40,000 impressions a month at an average position of 12, with a 1.2% click-through rate.

• Today: 40,000 impressions at 1.2% is roughly 480 clicks a month.
• Conservative case: those pages improve to around position 8, where your own data shows CTR nearer 3%. That is roughly 1,200 clicks, an increase of about 720.
• Applying conversion: at an organic conversion rate of 2%, the additional 720 clicks produce about 14 additional conversions a month.
• Applying value: at £400 per conversion, that is roughly £5,600 a month, or £67,000 annualised once fully ramped.

Two things make that credible rather than promotional. The CTR figures come from your own Search Console data at each position band rather than a published curve, and the position assumption is deliberately modest. Show both, and the forecast survives scrutiny even if it turns out low.

The ramp is the part that breaks trust

An annualised figure implies twelve equal months, and SEO does not deliver that. The first months produce little, the return accumulates, and a forecast presented as a flat annual number sets up a conversation about underperformance in month two.

Present the ramp explicitly: near-nothing for the first two to three months, partial return through months four to six, and the full run rate from month seven or later. Then the same total tells a story that matches what will actually happen, and month three stops being a crisis.

Revisiting the forecast

Update it at month three and month six against actuals, and say so when you present the original. A forecast revised in public builds more credibility than one that quietly stops being mentioned, and it turns an uncomfortable conversation into a scheduled one.

When you revise, change one input at a time and say which. Most forecasts miss on the position assumption rather than on conversion rate or value, so naming that explicitly keeps the revision from reading as a general retreat. It also improves the next forecast, because you now have real evidence about how far your pages actually move.

After two or three cycles you stop guessing at that input entirely, which is when forecasting becomes genuinely useful rather than a budgeting formality.

The four traps that make a forecast worthless

1. Forecasting from search volume. Volume is an estimate of what exists; impressions are a record of what you receive. Use the second wherever you have it.
2. Assuming position one. The CTR curve falls steeply, so a forecast built on reaching the top spot collapses when you land at eight.
3. Using a benchmark conversion rate. Yours is knowable and different.
4. Ignoring the timeline. A twelve-month return presented as an annual figure implies month-one revenue that will not arrive, which is how forecasts destroy trust even when they are eventually correct.

For what to report once the work is running, see which SEO metrics actually matter, and for how long to expect it to take, what the studies actually say.