SEO Forecasting: Build a Forecast That Isn't Fiction

An agency handed you a spreadsheet. Month 6: 4,000 visits. Month 12: 12,000 visits and $180,000 in pipeline. One line, going up and to the right, with no error bars and no stated assumption about where you rank.

That is not a forecast. It is a sales document with a chart on it. This post shows you how an honest one is built, so you can either build it yourself or audit the one you were given.

Can SEO be forecast honestly at all?

Yes — but only as a range, and only with the assumptions written down next to it. Ahrefs, which sells the data most of these decks are built from, describes SEO forecasts as “educated guesses based on data” and warns that “forecasts are always going to have some amount of uncertainty because changes are continually happening” (Ahrefs, SEO forecasting guide).

Its own forecasting model does not output a single number. It outputs a shaded band, and Ahrefs states there is “an 80% probability of our forecasted organic traffic being within that range.” If the people who built the model won't give you one number, you should be suspicious of any agency that does.

So the question is never “can you forecast SEO.” It is “what is the range, what ranking outcome does it assume, and what kills it.” An answer to those three is a forecast. Anything else is a wish with formatting.

Why does the standard volume x CTR x conversion model fail for a new site?

Because every version of it on this SERP starts by assuming you get to page one — usually position 3 — and page one is the part you have not earned yet. Ahrefs' own guide gives the basic model as “search volume x average click-through rate (CTR),” then notes it “only accounts for the keywords you have entered and is usually focused on one keyword per page, when pages rank and get traffic from many terms.”

That model was built for an in-house SEO on an established domain who needs a CFO to approve headcount. On a site with authority, assuming position 3 is a reasonable bet. On a two-month-old domain in a competitive trade, it is fantasy dressed as arithmetic.

The failure is structural, not arithmetic. Multiply 1,000 searches by a position-3 CTR and you get a number. Nothing in the calculation ever asks whether you will reach position 3, how long that takes, or what happens if you don't. The ranking assumption is the entire forecast, and it is the one input nobody prints.

What base rate should a fresh domain actually plan against?

1.74%. Ahrefs tracked 1 million URLs first seen by its crawler in September 2023 and found that only 1.74% of them ranked in Google's top 10 within a year (Ahrefs, published May 2025). Filtered to non-empty English content across 2 million URLs, 6.11% made it — the number Ahrefs calls fairer. Either way, that is your prior. Every forecast you are handed is implicitly claiming to beat it, and almost none of them say why.

The same study found 72.9% of pages currently in Google's top 10 are more than three years old, and the average #1-ranking page is five years old. You are not competing against content. You are competing against pages that have been accumulating links and engagement since before you registered the domain.

One number in that study cuts the other way and is worth planning around: of the pages that did reach the top 10, 40.82% got there within one month. Ranking, when it happens, often happens fast. That is why a 90-day read on a keyword cluster is a real signal, not an excuse to wait a year — and why we run a 90-day kill switch on any channel that produces no qualified leads.

Plan against the base rate, not the ceiling. Then let the long tail — the low-volume, low-difficulty terms where a new domain can actually place — carry the first two quarters. That is also the honest reason SEO takes as long as it does.

How do you forecast booked jobs instead of sessions?

You forecast the last step first and work backwards, because the only number the owner actually buys is booked jobs. Sessions are an input the agency controls the narrative on. Booked jobs are an output your calendar can audit.

Build the chain in this order, using your numbers for the bottom three — you already know them and they are the least uncertain things in the model:

Notice what this does. It puts the certain numbers (your economics) at the front and the wildly uncertain ones (your future rank) at the back — the exact opposite of the standard model, which anchors the entire projection on the input nobody can control. It also converts the forecast into the currency you already think in, which is the same reason we push clients to measure SEO ROI in closed revenue, not sessions.

And use a current CTR curve, not a 2019 one. Ahrefs' Google Search Console data across 300,000 keywords puts the average position-1 CTR for informational keywords at 3.9% in December 2025 — down from 7.6% in December 2023. On keywords where an AI Overview appears, position-1 CTR fell from 7.3% to 1.6% over the same window (Ahrefs, AI Overviews and clicks). If your forecast used the widely circulated “position 1 gets 39.8% of clicks” figure, it is off by an order of magnitude on this query type.

What does a forecast with a confidence band look like?

It looks like three scenarios with explicit ranking assumptions, and the conservative one is the one you budget against. Not a mid-case dressed up as a promise.

The verdict: budget against the conservative column and treat everything above it as bonus. An agency that shows you only the optimistic column has not made a forecasting error. It has made a sales decision.

The band also gives you a kill criterion. If month-6 actuals are tracking below the conservative line, the program is not “still ramping” — it is failing, and you should be cutting or changing it. A forecast with no downside case cannot ever be wrong, which is exactly why it is useless.

Which assumptions would invalidate the whole forecast?

Six, and every honest forecast lists them on the same page as the chart. If they are not written down, the forecast cannot be falsified — and an unfalsifiable projection is a marketing asset, not a plan.

• The ranking assumption. If the model assumes top-10 placement, it is betting against a 1.74% base rate for new pages. Say so.
• The CTR curve. Position-1 CTR on AI Overview keywords was 1.6% in December 2025 versus 7.3% two years earlier (Ahrefs, GSC data). A forecast built on a pre-2024 curve is built on a number that no longer exists.
• Search volume accuracy. Third-party volume is an estimate, and on 100-500/mo local terms the error bars are wide enough to swallow the forecast.
• Seasonality. An HVAC forecast that ignores July and January is not a forecast. Pull two years of Search Console or Trends data before you smooth anything.
• Conversion rate on a site that hasn't converted anyone yet. If the site is new, your organic lead rate is a guess. Carry it as a range and re-baseline at day 90.
• Publishing cadence actually happening. Most forecasts assume 8 pages a month ship. Then legal review takes three weeks. Model the cadence you will actually hit.

How do you spot a forecast that is really a sales document?

One tell beats all the others: no confidence interval and no stated ranking assumption. If the projection is a single number and never says what position it assumes you reach, it was built to close you, not to be checked against reality later.

• A single number instead of a range. Even Ahrefs' own forecasting model ships an 80% probability band.
• No stated ranking assumption anywhere in the deck.
• Traffic forecast, no revenue forecast. Sessions are safe to promise because nobody gets fired over them.
• No downside case. The scenario where it doesn't work is missing entirely.
• A hockey stick that starts at month 6, with nothing explaining what changes at month 6.
• The word 'guaranteed' anywhere near a ranking. Nobody can guarantee a ranking, and anyone who says otherwise is telling you something about themselves.
• It arrives before the audit. A forecast written before anyone looked at your Search Console, your backlink profile, or the SERP is a template with your logo on it.

That last one matters most. A forecast is downstream of a diagnosis. That is why we run the free audit first and the numbers second — and why our pricing is published rather than reverse-engineered from what your forecast made you feel.

What should you ask before you sign off on any projection?

Five questions. Any agency that can't answer all five in a live call does not have a model — it has a spreadsheet it filled in the night before the pitch.

• What ranking position does this assume, for which keywords, by which month?
• What is the conservative case, and what would you tell me to do if we're tracking below it at month 6?
• Which CTR curve did you use, from what year, and does it account for AI Overviews on these queries?
• Is this forecasting sessions or booked jobs — and if sessions, what conversion rate turns them into jobs, and where did that rate come from?
• What has to be true for this to be wrong, and how will we know by day 90?

A good answer sounds like: “We assume top-10 on 6 of these 20 long-tail terms by month 9, top-20 on the rest, using our own GSC CTR curve. Conservative case is 30% of that. If we're under it at month 6, we cut the cluster.” That is a person who intends to be measured. Compare it to “you'll see results in 3-6 months,” which is a person who intends to invoice.

If you want a projection you can actually hold someone to, start with the diagnosis, not the spreadsheet. Our free audit tells you which terms a domain your age can realistically place for and what the conservative case looks like in booked jobs — no lock-in, no ranking guarantee, no hockey stick. Get my free audit.