SEO forecasting: how to estimate the clicks your work will earn

An SEO forecast estimates how many extra organic clicks a set of changes is likely to earn, and over what period. The most reliable way to build one is bottom-up: start from the queries you already rank for in Google Search Console, estimate what a better position or a better snippet would add for each one, weight each estimate by how likely you are to achieve it, then add them up. Forecasts built from keyword-tool volumes alone tend to overshoot, because they ignore your real click-through rate.

Why a bottom-up forecast beats a traffic projection#

There are two common ways to forecast SEO. The first extends the trend line of last year's organic traffic into next year. It answers “where are we heading if nothing changes”, which is useful as a baseline, but says nothing about what your work adds. The second multiplies keyword volume by an assumed CTR for a target ranking. It answers “what is this keyword worth”, but it starts from third-party volume estimates and an industry CTR that may be nothing like your own.

The bottom-up method combines the strengths of both. Your Search Console data already contains the demand (impressions), your current position and your real CTR for every query. You only have to estimate one thing per query: where it could realistically move.

The method, step by step#

1. 1

   Set the baseline

   Export at least 12 months of clicks by month from Search Console (Performance → Date → Export). Compare each month with the same month last year so seasonality is visible. The baseline is what you expect next quarter with no new work.

2. 2

   Export queries for the last three months

   Performance → Queries → Export gives you up to 1,000 rows with clicks, impressions, CTR and average position. Three months is long enough to smooth out weekly noise and short enough to reflect today's rankings.

3. 3

   Remove brand queries

   People searching your name click at rates no curve predicts, and better SEO rarely moves them. Filter them out with a regex before you estimate anything. See branded vs non-branded keywords.

4. 4

   Pick a realistic target for each query

   For a query ranking 4 to 10, the top 3 is a realistic target. For one ranking 11 to 20, page one (around position 8) is. Beyond 20, treat the query as new content, not an uplift.

5. 5

   Estimate the uplift with a CTR curve

   Multiply impressions by the difference between the CTR at the target position and your current CTR. Scale the curve to the query's own CTR, so a query that under-clicks today is assumed to keep under-clicking. CTR by position explains the curve.

6. 6

   Weight by confidence

   Not every push works. Multiply each estimate by a probability: higher for queries already on page one, lower for page two, lower still where competitors are much stronger.

7. 7

   Add a time profile

   Changes take weeks to be recrawled and re-ranked. Spread the uplift over the quarter instead of counting it from day one.

8. 8

   Present three scenarios

   Conservative (only the top fixes you will actually ship, at their confidence), expected (all fixes at their confidence) and ceiling (all fixes fully achieved). The ceiling is there to show the size of the prize, not as a target.

A worked example#

Take a hypothetical query, “invoice software for freelancers”, with 6,000 impressions a month at an average position of 7.2 and a CTR of 2.4%. On the baseline curve, position 7.2 earns about 3.0% and position 3 about 9.2%. The query clicks at roughly 80% of the curve, so we assume it would also earn about 80% of 9.2%, which is 7.4%, at position 3.

Repeat that for every query worth pushing, add them up, and you have a forecast where every number can be traced back to a row in Search Console. That traceability is what makes the forecast defensible in a review meeting.

Run it on your own export in a minute#

Our free Search Console opportunity finder does steps 2 to 8 in your browser. Upload the Queries.csv from a Search Console export, add your brand names, and it returns the queries worth pushing, the snippets worth rewriting and a conservative, expected and ceiling forecast. The file never leaves your computer and there is no sign-up.

Forecasting new content#

For topics you do not rank for yet, Search Console has no impressions to work from, so you fall back on keyword volume. Keep these forecasts separate from the uplift forecast and discount them harder: you do not know your future position, and the volume itself is an estimate. A reasonable approach is volume × the CTR of a mid page-one position × a low probability, ramped over six months.

Mistakes that make forecasts wrong#

• Counting brand queries. They inflate the baseline and nothing you do changes them much.
• Assuming position 1. Forecast to the next realistic step, not the top.
• Using one CTR curve for everything. Your own data beats any benchmark; blend the two where your data is thin.
• Ignoring results pages that changed. If an AI Overview now sits above the results, the same position earns fewer clicks than it did a year ago.
• Reporting the ceiling as the plan. Clients remember the biggest number you said.
• Not checking afterwards. Compare each shipped fix with the same number of days before and after, against the rest of the site as a control.

Go deeper#