SEO forecasting: how to forecast traffic and revenue

A forecast is a set of assumptions written down as numbers: what the site earns from search today, what the work you expect to ship could add, from when, and what a visit is worth. Its value is that the assumptions become visible, so that each one can be checked when the months arrive.

Four things are often confused with it, and keeping them apart is most of what makes a forecast honest.

Rankings belong in the first row, never in the third. Google says that no one can guarantee a first place in its results, and that an SEO audit should give realistic estimates of improvement and an estimate of the work involved^[1]. A forecast is that realistic estimate, made explicit: the work can be committed, the result can only be estimated.

How a report sets the actual beside the forecast and the target, and why the original forecast is kept unchanged, is set out in the guide to SEO reporting in revenue; this guide is about making the forecast itself.

The baseline is what the site would earn from search over the forecast’s months if nothing changed. It is the most important line of the forecast, because every piece of work is measured against it, and it should come from your own data rather than a tool’s estimate.

1. 1In the Search Console Performance report, set the date range to the last 16 months, so that the drop or rise of each season can be seen against the same month a year earlier^[2].
2. 2For a longer history, pull the data with the Search Analytics API or the bulk data export and keep it yourself^[2]; two or three years show whether a season repeats or was a one-off.
3. 3Take organic clicks by month for the last 12 full months, leaving out branded queries if the forecast is meant to show what search work adds.
4. 4Mark months that will not repeat: a migration, a tracking fault, a one-off spike from news or a sale. Replace them with an estimate and say so.
5. 5Decide whether the baseline is flat, rising or falling. A site that lost clicks every month last year will not hold its level by itself; carry the trend forward, or say that you chose not to.

Where your own history is short, Google Trends can show the shape of the year for the searches that matter. Read it as a shape, not as volumes: Trends divides each data point by the total searches of its place and time and scales the result from 0 to 100, it uses a sample of searches, and terms with low volume appear as zero^[3]. Google describes it as one data point among others before drawing conclusions^[3].

The SEO Forecast Template takes the baseline in either form: twelve monthly values typed oldest first, each repeated in the same month of the forecast, or one monthly value shaped by a seasonality index of twelve numbers, scaled so that they average one. A month at 110, against an average of 100, gets a tenth more than the average month.

The forecast rises above the baseline only through work you expect to ship: fixes, new pages, rewrites, merges. Give each piece of work its own line, with three numbers: how much more traffic it could bring once it has taken full effect, as a share of the baseline, the month it starts to count, and how long it takes to reach full effect.

Estimate the first number from evidence you have. The fairest guide is a group of pages of the same kind that already perform on your own site: how many clicks a category page, a product guide or a location page earns once it is established. A keyword tool’s search volume is a weaker guide, because a page rarely earns most of a query’s searches and the volume says nothing about your position. Google’s own Keyword Planner forecasts are for advertising: they show the conversions, clicks or impressions your keywords are likely to get based on your spend, refreshed daily from the last 7 to 10 days and adjusted for seasonality^[4]. They are not a forecast of organic traffic.

Then give the work time. Google says that some changes take effect in a few hours and others can take several months, and that in general it is worth waiting a few weeks to assess whether the work had a beneficial effect^[5]. A new section of the site needs to be crawled, indexed and linked before it can rank, so its start month is later than its publish date and its ramp is longer than a fix to a page that already ranks.

In the template, each work item adds its uplift from its start month, growing in a straight line over the ramp, and the uplifts of several items are added together rather than multiplied. Adding them is the cautious choice: two pieces of work on the same pages rarely each deliver their full effect on top of the other.

A single forecast line hides how unsure it is. Give three cases instead. The base case is the work delivering what you expect; the low case is the work delivering only part of it; the high case is the work doing better than expected. Keep the baseline the same in all three, so that the cases differ for one reason only: how much of the expected effect arrives.

In the template, the base case reaches all of the expected uplift and the low and high cases reach a share of it that you set. Read the spread, not just the middle: a wide spread is not a weak forecast, it is an honest one about work whose effect is uncertain.

The example values are round numbers chosen to show the method, not typical figures. In the example the work adds 22,771 clicks above the baseline in the low case, 45,542 in the base case and 68,313 in the high case, over the year. By the last month the three items have reached their full effect together, which is why the gap between the cases widens through the year rather than starting wide.

A forecast of clicks becomes a forecast of revenue through two of your own numbers: the share of organic visits that convert, and what a conversion is worth. Use your own history for both, by page type if you can, because a visit to a guide converts differently from a visit to a product page.

Three rules keep that figure honest. It is modeled revenue, an estimate from your inputs, and it is labelled so wherever it is quoted, never mixed with known, influenced, incremental or GA4-reported revenue, as the guide to SEO reporting in revenue sets out. It is revenue, not profit: to judge whether the work pays, multiply it by your margin and set it against the cost, as SEO ROI: how to measure it explains. And the conversion rate carries the attribution rules of the analytics that produced it, so the same clicks can be worth more or less under another model.

A forecast is arithmetic on its assumptions. Several things decide whether it holds, and none of them can be computed from the inputs.

• Search engine updates. Several times a year, Google makes significant, broad changes to its search algorithms and systems, which it calls core updates^[6]. An update can move the baseline itself, whatever work you ship.
• Changes to the results page. A results page that gives more room to answers, ads or other features can lower the clicks a position earns, so that rankings hold and traffic falls. A rise in zero-click search moves the baseline, not the work.
• Seasons that move. A sale moved to another month, a launch, or a change in what people search for can shift the shape of the year, and one past year may hold spikes that will not repeat.
• The work itself. Pages ship later than expected, fixes are only partly deployed, and a migration in the middle of the year can reset positions for weeks.
• Measurement. A change to tracking, consent or the attribution model changes the actual figures the forecast is compared with.

Write the main risks beside the forecast, with what each would do to it. A reader who knows that the base case assumes the new guides are published by the third month can see at once why the actual falls short when they are not.

A forecast nobody checks teaches nothing. Set a review date when you make it, and compare it with the actual figures each quarter.

1. 1Save the forecast with its inputs and the date, and do not edit that version again. The template’s CSV file holds the inputs, one row a month with the baseline and the three cases, and a total row.
2. 2Each month, add the clicks you actually got beside the baseline and the three cases.
3. 3At each quarterly review, find where the actual sits: below the low case, inside the range, or above the high case.
4. 4When it falls outside the range, find the assumption that was wrong before changing anything: the baseline, a start month, a ramp, an uplift, or something the forecast could not see.
5. 5Make a new version with the corrected assumption and a new date, and keep both. The record of past forecasts against their actuals is what makes the next one believable.

Revenue is reviewed the same way, against the order record rather than analytics where you can, and only once the clicks have been read: a revenue miss with clicks on forecast points at conversion or value, not at search.

• Forecasting from a keyword tool’s search volume instead of what pages of the same kind earn on your own site.
• Leaving out the seasons, so that the first quiet month reads as a failure of the work.
• Starting every piece of work in the first month, as if pages ranked the day they were published.
• Giving one line where three cases belong, then treating the base case as a commitment.
• Rewriting the forecast as the months arrive, so that it always matches the actual and can never be judged.

Avoid these and a forecast becomes what it should be: a clear statement of what the work is expected to do, which the months that follow can confirm or correct.