How to Forecast an SEO Campaign Without Lying to Yourself

Every SEO proposal comes with a forecast, and most of them are made up. Someone multiplies search volume by a made-up click rate, times a made-up conversion rate, times an average order value, and hands you a number with two decimal places and zero humility. It looks precise. It is fiction.

I forecast differently, because I learned the hard way. I run SEO for my own store, Simple Sheets, and the first time I asked an AI to forecast its revenue it gave me a beautiful, confident, wrong number. The more real data I fed it, the less certain and the more useful it got. One thing to say up front: what follows is the model I landed on for my situation. It is not the only model, and the last two sections are about how to find the one that fits yours.

Why most SEO forecasts are fiction

The standard forecast stacks four guesses: search volume from a tool, a generic click-through-rate curve, a conversion rate someone half-remembers from a case study, and an average order value. Each guess has an error bar, and multiplying them multiplies the error. By the time you reach a dollar figure the range is so wide the point estimate means nothing. Worse, the inputs are borrowed. A generic CTR curve is not your CTR curve. A case-study conversion rate is not yours.

A forecast is only useful if it is built from evidence you can point to, and if it tells you where it is likely to be wrong. Here is how I build one.

Start from your own data

The single biggest upgrade to any SEO forecast is to stop trusting third-party traffic estimates and start using Google Search Console. GSC reports your actual clicks, impressions, and average position, straight from Google. Tools estimate; Search Console measures.

On my own store this mattered. Ahrefs estimated the site’s organic traffic had roughly tripled over a year. Search Console said clicks had fallen about 70% in the same period. Both cannot be true, and GSC is the one Google actually sends traffic on. The tool was counting rankings that never got clicked, plus impressions the site was losing to AI Overviews. Forecast off the tool and I would have promised growth into a decline.

Pick the model that fits your business

Once you trust your click data, you need a way to turn clicks into dollars. There are several ways, and the right one depends on how cleanly you can attribute revenue and how your business actually makes money. Here is the model I used, why it fit my store, and how to choose your own.

My store has poor attribution. Simple Sheets sells spreadsheet templates on Kajabi, its best revenue comes from its most expensive product, and buyers convert through a mix of landing pages and email, often days after the first visit. That makes any single click-to-purchase conversion rate a fiction. So I used the crudest honest anchor available: revenue per session. Take gross revenue over a period, divide by organic sessions. On my store that was about ten cents per organic session, and the useful part was that it held steady through a roughly 70% traffic collapse. The conversion machinery did not break; the traffic left. So the forecast could say, honestly, that restoring sessions should restore revenue at about the same rate, with no invented conversion rate anywhere in it.

I want to be clear that I chose this because it was the best available for a business with messy attribution, not because it is the best model. Revenue per session works when your per-session value is stable and you cannot trust anything more granular. It breaks the moment your traffic mix or pricing shifts, and it tells you nothing about which pages or which intents actually drive the money.

If your attribution is cleaner, use something sharper. A direct-response ecommerce site with solid tracking can forecast buyer-intent clicks to specific landing pages, times a real conversion rate, times average order value, page by page. A lead-gen business forecasts qualified leads and has to carry a much wider error band, because a long sales cycle and lifetime value sit between the click and the revenue. A SaaS company has to model trials, activation, and churn on top of that, which makes the forecast far less certain and far more valuable to get roughly right. The further your money sits from the click, the more assumptions you stack, and the more honest your range has to be.

So the method is not a formula to copy. It is a sequence: measure your own clicks, understand how your business actually converts, pick the simplest model your data can support, and write down where it will break. My ten-cents-per-session anchor is the right answer for my store, precisely because its economics are stable and its attribution is not. Yours will look different, and it should.

Model click upside with your real CTR curve

To forecast the clicks a ranking change earns, you need a click-through-rate curve: how often people click position 1 versus 5 versus 10. Every tool ships a generic one. Do not use it as gospel. Build yours from Search Console, because your curve depends on your brand strength, the SERP features on your queries, and your query mix. Here is the actual curve from my store, pulled from Ahrefs’ GSC insights over the last 90 days.

Notice how fast it falls. Moving a page from position 8 to position 3 is not a small bump on that curve, it is more than a four-fold jump in click rate. Moving from the middle of page two to the bottom of page one barely registers. That shape tells you which ranking gains are worth forecasting and which are rounding error.

That curve is an average, though, and averages hide a lot. Your real click rate at a given position swings with what else is on the results page. A result sitting under an AI Overview earns fewer clicks than the same position with no AI Overview above it. Add paid results on top and it drops again. Informational queries behave differently from commercial ones, where buyers will scroll and click further down. If you want a forecast you actually trust, segment your CTR by whether an AI Overview shows, by whether ads show, and by intent, instead of leaning on one blended number.

This is also where a tool earns its place, as long as you use it to sharpen your own data rather than replace it. Earlier I said not to trust a tool’s traffic estimate over your Search Console clicks, and I stand by that. But Ahrefs has a GSC insights feature that pulls in your real Search Console data and layers its own keyword, difficulty, and SERP-feature data on top, including whether an AI Overview is present on a query. That combination, your actual clicks enhanced with SERP context, is genuinely useful for exactly this segmentation, and it is what I use on my own store. The rule still holds: your data is the ground truth, the tool just makes it richer.

Find the trapped opportunity

Now you can find the pages worth working. I call it the page-two-to-five problem: keywords where you already rank between position 11 and 50, so Google already trusts the page, but nobody sees it because it is stranded below the first page of results.

Pull every query ranking in that band with real search volume and low difficulty, then score each by volume divided by position, so a high-volume term stuck at position 40 floats to the top. Cluster the winners by the URL that ranks, because fixing one page usually lifts a dozen keywords at once. Then use your CTR curve to estimate the click swing as that page climbs from, say, position 25 toward position 5. That swing, times your revenue per session, is your forecast, page by page, grounded in rankings you already hold.

When I ran this on my own store, the highest-value targets were not the ones I would have guessed. Joining the ranking data to real revenue per page surfaced two product pages that were quietly worth the most, and neither was in my original plan. Forecasting off gut would have missed them both.

Turn it into a range, not a promise

A single number is a lie by omission. Real forecasts are ranges. I build three: conservative, base, and optimistic, each tied to a different assumption about how many trapped pages actually move and how far.

On my store that landed somewhere between about 6,700 and 12,800 additional organic sessions a month if the recovery lands, which at ten cents a session is roughly 470 to 1,540 dollars a month in incremental revenue. The gap between low and high is not sloppiness. It is the honest width of what I do not control. Present the range with its assumptions written down: how many pages, how far they move, what has to be true. A forecast you can audit later is a forecast worth making.

Where the forecast breaks

I have watched forecasts fail in the same handful of ways, including my own.

• Single-keyword math. A page ranks for dozens of queries; forecasting off its top keyword alone undercounts it badly. Model the page, not the phrase.
• Trusting tool traffic over Search Console. The roughly five-fold gap between what a tool estimated and what GSC measured, on my own store, is the cleanest reminder I have: estimates are directional, clicks are real.
• Broken measurement underneath. For a long stretch my store’s analytics recorded zero revenue, because a purchase event was never wired up. Every forecast is only as good as the plumbing under it. Confirm the numbers you build on are actually being recorded.
• Correlation mistaken for cause. A revenue drop that lined up with an email-list decline looked like the cause, until the dates showed the list fell after the revenue did. Forecast a recovery by fixing the wrong lever and you miss.
• False confidence. My first forecast for the store was a wide, cheerful revenue range with nothing under it. It was wrong. The more real history I fed the model, the more it hedged, and the more useful it got. Certainty is a warning sign, not a feature.

Find your lever (mine was links)

One more honest note, because a forecast that ignores it is incomplete. A forecast tells you the prize; it does not tell you which lever earns it. On my store, every path pointed at the same one: links. But the reasoning is specific to my situation, and yours will be different.

Here is why links, for me. My pages are already well optimized. The technical SEO is clean. And I cannot meaningfully build the site out further, because it lives on Kajabi. On top of that, AI Overviews have crushed the store’s organic traffic, and AI has made the product itself less essential than it used to be. When the on-page work is done and the platform is capped, more authority is the only lever left that moves rankings, which means backlinks and mentions. For weeks, two AIs, mine and my business partner’s, kept suggesting on-page tweaks, meta rewrites, and new informational content, passing a giant document back and forth by email. They were not wrong in general. They were wrong for this store, because that work was already finished.

So my realistic options narrow to three: add other services and revenue lines, which is in progress; earn authority through links and mentions, also in progress; or open new marketing channels, which I do not have the bandwidth for right now. Links are the lever for me. That is not a happy story, it is an honest one, and it is exactly why a live lab is more useful to write about than a polished case study.

Your lever will probably be different. A complex ecommerce site drowning in crawl and indexation problems might unlock a pile of sales just by fixing its technical SEO. A thin site might need content before anything else. A site with strong pages but weak authority needs what I need. Run the same diagnosis: what is already handled, what is capped, and what is the one constraint that, if removed, actually moves the number. Forecast the prize, then be honest about the lever.

If you want the two things that tend to move a forecast in practice, a trapped-page audit and the link and mention building that lets those pages climb, that is the work I do. Start with a two-minute fit check, or read how I think about getting cited by AI. I am also building a free forecasting tool, so you will be able to run a version of this on your own Search Console data without hiring anyone. It is not live yet; when it is, I will link it right here. The Field Notes are free in the meantime, while I keep testing all of this on my own store.

Frequently asked questions

Can you forecast SEO for a brand-new site with no historical data?

Yes, but honestly and with wider error bars. Without your own Search Console history you borrow reference points, keyword search volumes, competitor click curves, and category benchmarks, and you present a range rather than a single number. As your own impression and click data accrues, you swap the borrowed assumptions for real ones and the forecast tightens.

How accurate is an SEO forecast?

Directional at best, and that is the point. A forecast is a planning tool for setting expectations and priorities, not a promise. The useful version gives a range with stated assumptions you can revisit, so when reality diverges you can see which assumption was wrong instead of just missing a number.

What data do you actually need to forecast SEO?

At minimum: the keywords you are targeting and their search volumes, a click-through-rate curve by position (your own from Search Console if you have it, a reference curve if you do not), and a conversion rate plus value per conversion for the pages involved. The more of those that come from your own data instead of industry averages, the less fictional the forecast.