SEO Forecasting: Methods, a Worked Example and How to Check It

SEO forecasting is projecting how much organic search traffic, how many conversions and how much revenue your site will get over a set period, usually the next twelve months. You build it from two things: your own history (Search Console and analytics) and a keyword plan (search volume × the click-through rate of the position you expect to reach). A useful SEO forecast ends in conversions and dollars, and it is a range, not a promise.

This guide covers the methods, which one to use when, how AI Overviews change the click math, a seven-step build, and a full worked example from clicks to dollars. It also covers what most forecasts skip: how to avoid double counting, how to forecast by landing page instead of by keyword, and how to check whether your model is any good before you show it to anyone.

What Is SEO Forecasting?

An SEO forecast answers one question: if we keep doing what we do, or if we carry out this plan, what will organic search deliver, and when? Reporting looks backward at what happened. A forecast looks forward and puts a number on the result of a decision you have not made yet.

There are three layers you can forecast, and each one depends on the one before:

• Traffic: impressions and clicks from Google and Bing, by month.
• Conversions: leads, sign-ups, calls or orders those visits produce.
• Value: revenue or the dollar value of those conversions, which is what finance and clients actually approve budgets on.

Most forecasts stop at traffic. That is the least useful layer to a decision-maker, because a page that brings 5,000 visits and no leads is worth less than one that brings 300 visits and 20 leads.

Why Forecast SEO at All

• Budget and buy-in. A projection with a cost and an expected return gets approved more often than “SEO takes time, trust us.” If the company target is higher than what your forecast says current resources can deliver, the gap is your argument for more content, engineering or link building budget.
• Realistic expectations. Agreeing on a slow first quarter up front protects you when the first quarter is slow.
• Prioritization. Forecasting each project (a new page cluster, a refresh, a technical fix) shows which one returns the most for the effort.
• Channel comparison and opportunity cost. Money spent on SEO is money not spent on ads, email or events. A forecast in conversions and dollars lets you compare them on the same scale. Our SEO ROI guide covers the cost side.
• Competitive planning. Projecting your trend next to a competitor’s shows whether you are closing the gap at current effort.

Forecasting has limits. It cannot predict algorithm updates, a competitor’s big launch or a sudden new trend with no search history. If a client or boss wants a guaranteed number, a forecast is the wrong tool, and saying so early is better than missing a number you never controlled.

The Data You Need

Use first-party data for everything you already rank for and third-party data only for what you do not: new keywords, new sites and competitors. Two habits make the inputs trustworthy:

• At least twelve months of monthly history, so the model sees a full seasonal cycle. With less, a normal summer dip looks like a decline. Search Console only keeps a little over a year, so export it every month into a sheet or a warehouse.
• Audit conversion tracking first. If a form stopped firing in March, every rate you compute from that month is wrong before you start. Our SEO conversion tracking guide explains how to check what organic search actually converts.

The Four Types of SEO Forecasting Methods

General forecasting textbooks split methods into qualitative (judgment) and quantitative (data), and the quantitative side into time series and causal models. Translated to SEO, that gives four practical types:

Method 1: Keyword-based forecasting

The core formula is a chain. Each step multiplies by a rate you either measured or assumed:

Three refinements separate a defensible keyword forecast from a wishful one:

• Forecast per page, not per keyword. One page ranks for many related queries. Group keywords by the page that will rank for them, and use the page’s total demand, not one head term. Some tools sum the traffic the current top page gets from all its queries (Ahrefs calls this Traffic Potential), which is a reasonable ceiling for a page.
• Use your own CTR curve. Export queries from Search Console with clicks, impressions and position, bucket them by position (1, 2 to 3, 4 to 5, 6 to 10, 11 to 20) and compute clicks ÷ impressions per bucket. Do it separately for branded and non-branded queries, because branded CTR is much higher.
• Be honest about target positions. Position 1 on every keyword is not a forecast. Pick positions based on keyword difficulty, your site’s authority and how far the competitors are ahead.

Method 2: Statistical forecasting from your history

This answers “where are we heading if nothing changes?” From simplest to most robust:

• Straight-line growth: apply last year’s month-over-month or year-over-year growth rate. Quick, but it ignores seasonality and assumes growth never slows.
• Moving average: smooths noise, good for spotting the direction, weak at predicting peaks.
• Linear regression: fits a trend line (FORECAST.LINEAR or TREND in Excel, FORECAST or TREND in Google Sheets). Fine for steady sites with little seasonality.
• Exponential smoothing with seasonality: Excel’s FORECAST.ETS uses exponential triple smoothing, detects the seasonal cycle, and FORECAST.ETS.CONFINT returns a confidence interval for each month, all without code (see Microsoft’s forecasting functions reference). If you work in Python, libraries such as Prophet or SARIMA models do the same with more control, and several free Colab notebooks wrap them.

Before you run any of these, mark the months distorted by a migration, a tracking outage or a core update. Either exclude them or treat them as known events, otherwise the model learns that a broken month is normal.

Method 3: Competitor benchmarking

A new domain has no history to extend. Borrow one: find a competitor that ranks for your target keywords and started from a similar position, pull its estimated organic traffic curve from a keyword tool, and use its growth path as a template. Pick a peer, not the category leader. The leader’s traffic is the result of years of links and content you do not have yet. Replace the borrowed curve with your own data as soon as you have a few months of it.

Which Forecasting Method Is Best?

The best method is the one that matches your data and the decision you need to make. Use this rule:

AI Overviews, AI Assistants and the Click Math

The keyword method assumes a ranking earns a stable share of clicks. That assumption broke on queries where Google shows an AI Overview. In Pew Research Center’s study of 900 US adults’ real Google searches, users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when it did not, and clicked a link inside the summary in only 1% of visits.

What to do with that in a forecast:

• Flag every keyword that shows an AI Overview. Most keyword tools mark it. Keep two lists with two CTR curves.
• Measure your own discount. In Search Console, compare CTR at similar positions for your flagged and unflagged queries. Published estimates of the drop vary a lot by study, query type and period, so your own ratio beats any single benchmark. Pew’s gap, roughly half the clicks, is a reasonable starting assumption until you have yours.
• Expect the gap to be largest on informational queries. Commercial and navigational searches still get clicked because the searcher needs a page to buy, book or sign in.
• Forecast AI assistants separately. ChatGPT, Perplexity, Gemini, Claude and Copilot publish no search volume or rankings, so the keyword method does not apply. Treat referrals from them as a trend line from your own analytics, and remember that visits with no referrer land in Direct. Our AI conversion tracking guide explains how to measure that channel.

How to Build an SEO Forecast in 7 Steps

The classic “seven steps of forecasting” (define the use, choose what to forecast, set the horizon, pick a model, gather data, run it, validate) map cleanly onto SEO:

Step 1: Set the goal, metric and horizon

Write down the decision the forecast supports (next year’s budget, a content plan, a pitch) and the metric that decides it. For most businesses that is conversions or revenue, not clicks. Default to twelve months, by month.

Step 2: Pull and clean the baseline

Export monthly Search Console clicks and impressions, and organic conversions by landing page from analytics, for at least twelve months. Annotate migrations, outages and algorithm updates.

Step 3: Split branded and non-branded

Branded searches follow brand awareness, PR and ads, not SEO work. Mixing them in makes SEO look better or worse than it is. Search Console now does the split for you: the branded queries filter classifies queries with an AI-assisted system that catches misspellings and product names. It only appears on top-level properties with enough query volume, so smaller sites still need a regex filter on their brand terms.

Step 4: Project the baseline trend

Run FORECAST.ETS (or your model of choice) on monthly non-branded clicks. This is the “do nothing new” line. Forecast branded clicks separately, usually flat or tied to brand plans.

Step 5: Model the incremental opportunities

For each planned keyword group: volume × your CTR at the target position × the AI Overview adjustment, minus current clicks. Then spread it over time with a ramp. New pages take months to reach their target position, and harder keywords take longer. If you do not have your own ramp data, assume nothing in the first quarter and a gradual climb after that, and say so in the assumptions.

Step 6: Convert clicks into conversions and value

Multiply each line by the conversion rate of the landing page type it will land on, then by the value per conversion. The next section explains why this beats one site-wide rate.

Step 7: Build the range, backtest and track

Produce conservative, likely and optimistic scenarios, test the model on months you already know, then compare forecast to actuals every month and re-forecast every quarter.

Forecast Conversions by Landing Page, Not by Keyword

Here is the problem every keyword forecast runs into at step 6: Search Console has queries but no conversions, and your analytics has conversions but no queries. There is no such thing as a measured conversion rate for a keyword. Applying one site-wide conversion rate to every keyword hides the biggest difference in your forecast: a blog post and a pricing page convert at very different rates.

The fix is to forecast through the landing page:

1. Map each keyword group to the page (or page type) that will rank for it.
2. Pull the organic conversion rate of that page or page type from your analytics: blog, product, category, pricing, service, tool.
3. Assign a value per conversion. For a purchase that is revenue. For a lead, value = close rate × average deal value (our guide on calculating conversion value walks through it).
4. Multiply forecast clicks by that rate and that value, per group.

This also makes the forecast checkable later: you compare forecast and actual conversions per landing page, which is exactly the level at which organic conversions can be measured. SEOConversion reports conversions and their value from Google, Bing and AI assistants by landing page, which gives you that actuals column without stitching exports together.

Worked Example: From Clicks to Dollars

All numbers in this example are illustrative. A B2B software site wants a twelve-month SEO forecast to support a content budget.

Part 1: Baseline

Non-branded organic clicks over the last twelve months: 96,000 (8,000 a month on average). The seasonal trend model projects 10% growth, so the baseline for the next twelve months is 105,600 clicks. Branded clicks are forecast separately at a flat 3,000 a month and left out of the SEO case.

Part 2: Incremental opportunities

Three planned projects, using the site’s own Search Console CTR curve and a 0.5 multiplier for the cluster that shows AI Overviews:

The ramp assumption: 0% of the increment in months 1 to 3, 25% in months 4 to 6, 60% in months 7 to 9 and 100% in months 10 to 12. Over the year that is 3 × 0 + 3 × 0.25 + 3 × 0.6 + 3 × 1 = 5.55 months’ worth of the full increment.

Part 3: Conversions and value by landing page

Organic conversion rates from analytics: pricing page 3%, blog 0.4%, integration pages 1.5%, non-branded organic overall 0.8%. Value per lead: 10% close rate × $3,000 first-year deal = $300.

Look at the how-to cluster: it brings almost as many clicks as the comparison cluster but about an eighth of the conversions. A traffic-only forecast would have ranked the two projects as roughly equal.

Also notice what the content budget should be judged on: the $29,250 increment, not the $282,690 total. The baseline mostly happens whether or not you fund the new work.

Part 4: The range

Conservative: baseline growth 5% and half the planned increment (rankings land lower or later). Optimistic: baseline growth 15% and 1.3× the increment.

How to Check Whether Your Forecast Is Any Good

A forecast you never test is an opinion with a chart. Two checks, one before you present it and one after:

Before: backtest on a holdout

Hide your last three months of data, build the forecast from the months before, and compare its prediction for the hidden months with what actually happened. Score it with the mean absolute percentage error (MAPE): the average of |actual − forecast| ÷ actual.

Report that error next to the forecast. If the backtest misses by far more than your planned uplift, the model cannot tell your plan apart from noise, and you need a better baseline before you argue about increments.

After: forecast vs actual every month

Put the forecast and actual clicks, conversions and value side by side, per line. When a line misses, find which assumption broke: the ranking (check position), the click-through rate (position held but clicks fell, often an AI Overview), the conversion rate (traffic arrived but did not convert) or the tracking. Then update the assumption and re-forecast, rather than defending the old number.

Failure Modes and How to Debug Them

SEO Forecasting Tools and a Template

You do not need paid software, but tools save time at scale:

• Spreadsheets (Excel or Google Sheets): full control of every assumption and easy to show your working in a pitch.
• Keyword platforms (Semrush, Ahrefs, SE Ranking): volumes, difficulty, AI Overview flags, competitor traffic and built-in traffic estimates.
• Dedicated forecasting tools such as SEOmonitor model keyword clusters and scenarios for agencies.
• Free calculators (for example the Animalz SEO forecasting tool) estimate traffic from your domain authority, publishing pace and a keyword list. Good for a first look; check their CTR and ramp assumptions against your own data.
• Python notebooks with Prophet or SARIMA for statistical forecasts on large sites.

Copy these columns into a sheet, one row per keyword group:

Presenting the Forecast

• Lead with value and the range, then show traffic as the supporting evidence.
• List the assumptions in plain words: target positions, CTR curve source, AI Overview multiplier, ramp, conversion rates. Anyone should be able to disagree with one line instead of the whole forecast.
• Pair the forecast with the investment it needs: content, development, links, tools. A forecast without a cost is not a business case.
• Show the backtest error so nobody mistakes the likely case for a guarantee.
• Adapt for local businesses: local volumes are small and the map pack behaves differently from standard results, so rely more on your own Search Console and call or form data than on national CTR curves.

If you need to decide whether SEO deserves budget at all, our guide on whether SEO is worth it uses the same numbers to answer that.

FAQ