An honest SEO forecast is a range, not a promise: a low, base and high case, each with its assumptions, a date for the first fair reading and a rule for what happens if a Google update lands in the window. No one can guarantee a number-one ranking, so promise the work and the reporting, not a position.
- Google suggests asking an SEO what results to expect, in what timeframe and how success is measured (Google, Search Central, read 30 October 2026).
- From 11 December 2025 to 30 October 2026 the Search Status Dashboard listed seven core and spam updates, and no gap between them could hold a 90-day window (our arithmetic from Google, Search Status Dashboard, read 30 October 2026).
- In Penlid's simulations a real +50% lift on a page with 4 clicks a day was found 61.23% of the time with 6 untouched pages and no drift between pages (Penlid, methodology, read 30 October 2026).
Why can't an SEO forecast be a guarantee?
An SEO forecast cannot be a guarantee, and Google's own advice on hiring an SEO is blunt about ranking promises: No one can guarantee a #1 ranking on Google
. The same advice lists useful questions to ask an SEO, including what results to expect, in what timeframe and how success is measured. Our reasoning: a forecast is the written answer to those questions, and you can honestly commit to the work, the reporting and the dates you will check. Penlid's SEO test calculator keeps to the same limit: It cannot say that a change will improve a ranking, and nothing on this page does
.
How long until a change shows, and when can you check?
Google gives a range rather than a date, and Penlid measures one page change 38 days after it goes live. Google's SEO Starter Guide puts the delay between a few hours and several months and suggests waiting a few weeks before assessing the work.
The 38 days is a reading rule, not a promise about when results arrive: a 28-day after window, a 7-day allowance for the newest days to settle and 3 days of reporting lag (28 + 7 + 3 = 38). The 7 is Penlid's own allowance, not a Google figure. Our reasoning: under this rule, a promise of a lift in 30 days falls due before the first fair reading.
What sets the width of a forecast range?
Two kinds of noise set the narrowest honest range: counting noise, which shrinks as a page gets more clicks, and drift between pages, which does not. For events that arrive independently at a steady rate, the standard deviation of a count is the square root of its mean (Wikipedia's Poisson distribution article), so a page averaging 25 clicks varies by about 5, or 20%, and one averaging 400 by about 20, or 5%.
The calculator calls drift tau: how far apart untouched pages move in a month beyond counting noise. Its presets are 0.05, 0.15 and 0.35, and with them two untouched pages in three end a month within about 5%, 15% or 35% of where the group's trend says they should be. It is the noise a real change has to stand out from, and it does not shrink as a page gets bigger
.
Our arithmetic treats the last 28 days as the page's true level, as the calculator does, and adds the two noises as variances: spread = √(1/n + tau²) for n clicks.
| Clicks in the month | Counting noise only | Tau 0.05 | Tau 0.15 | Tau 0.35 |
|---|---|---|---|---|
| 25 | 20.0% | 20.6% | 25.0% | 40.3% |
| 100 | 10.0% | 11.2% | 18.0% | 36.4% |
| 400 | 5.0% | 7.1% | 15.8% | 35.4% |
| 1,600 | 2.5% | 5.6% | 15.2% | 35.1% |
At typical drift, quadrupling a page from 400 to 1,600 clicks narrows the spread only from 15.8% to 15.2%: extra traffic cuts counting noise and leaves drift alone.
What else can move the whole range?
Seasons, Google updates and a bad recent stretch can shift the whole range, not only widen it, so a forecast should say which of them it allows for.
- Season. Penlid's calculator does not simulate seasonality, and in a full 16 months, the most Search Console keeps, four calendar months appear twice and eight once (our arithmetic), so most calendar months have only one year behind them. Google's debugging guide suggests the Last 16 months range to see whether a drop recurs yearly.
- Updates. The Search Status Dashboard listed three core and four spam updates from 11 December 2025 to 30 October 2026, plus a February 2026 Discover update that we did not count. The longest stretch without a core or spam update, from the end of one to the start of the next, was about 85 days (29 December 2025 to 24 March 2026; our arithmetic adds each listed duration to its start date), so every 90-day window in that period overlapped at least one. Google also makes smaller core updates without announcing them. Overlap is timing evidence, not proof of cause.
- A bad recent stretch. Penlid's methodology says
A page that has just had a bad four weeks is likely to have a better next four whatever you do to it
. Left unchanged, such pages still came back "improved" in its simulations 5.50% to 9.20% of the time, against at most 3.10% otherwise. Our reasoning: a baseline taken from the dip builds a recovery into the base case.
Which forecasts can you check afterwards?
You can check a forecast only when the page has enough clicks, and enough untouched pages, for the measurement to see a change of the size you promised; Penlid's methodology measured that on simulated sites.
| Clicks a day | Real lift | Untouched pages | Found |
|---|---|---|---|
| 4 | +50% | 6 | 61.23% |
| 4 | +80% | 6 | 92.01% |
| 40 | +30% | 6 | 99.50% |
| 400 | +10% | 6 | 99.95% |
Drift costs a lot: with 15% drift and a +80% lift on six untouched pages, the lift was found 36.95%, 59.65%, 79.55% and 82.20% of the time at 1.4, 4, 40 and 400 clicks a day. A page with no clicks in the 28 days before the change cannot be measured, and one with 1 to 9 gets a direction and no size. Our reasoning: the smaller the lift you promise, the larger the page and the calmer the site must be to check it. The calculator prints the change that gets an 80% chance of a clear verdict for your own counts; the guide to the minimum detectable effect explains how.
What should the written forecast say?
A written forecast should state a baseline, three cases, an update rule, a first fair reading and a revision trigger; this structure is our suggestion, not a Google standard.
- Baseline. The last 28 days, checked against a longer average so a dip is not taken for the normal level.
- Low, base and high cases, each with its assumed lift, taken from changes you have already measured against untouched pages. None of the sources below gives a typical lift, and Penlid's tables show how often a known lift is found, not how big lifts usually are.
- Update rule. If a core update starts inside the window, wait at least a full week after it completes, as Google advises, then compare that week with the week before it began.
- First fair reading. The go-live date plus 38 days; before then, report activity, not results.
- Revision trigger. The reading outside the low or high case, and what you will do then.
Report against it as in what a monthly SEO report should contain, prove results as in how to prove SEO results to a client, and if traffic falls see what to check first when traffic drops.
Example (illustrative numbers)
These numbers are invented; they are not data from any site or calculator output. A service page earned 625 clicks in its last 28 days, about 22 a day, on an otherwise flat site, and we use the typical drift preset, tau 0.15.
Counting noise is √625 = 25 clicks, so 25 ÷ 625 = 0.04. The combined spread is √(0.04² + 0.15²) = √0.0241 = 0.155, and 0.155 × 625 = 97 clicks. If nothing changes, and the spread is roughly bell-shaped, about two months in three land between 625 − 97 = 528 and 625 + 97 = 722, and about 19 in 20 between 625 − 194 = 431 and 625 + 194 = 819.
| Case | Assumed change | Clicks | Where it sits |
|---|---|---|---|
| Low | 0% | 625 | Middle of the range |
| Base | +8% | 625 × 1.08 = 675 | Inside 528 to 722 |
| High | +20% | 625 × 1.20 = 750 | Above 722, inside 431 to 819 |
The base case sits 50 above the level, about half a spread, and the high case 125 above, about 1.3 spreads. Both are inside the range that covers 19 no-change months in 20, so this page's clicks alone cannot confirm either: promise the work, not +8% or +20%.
If the refresh goes live on Monday 11 January 2027, the first fair reading is Thursday 18 February 2027 (11 January + 38 days = 18 February). At 22 clicks a day the page sits between the 4-a-day and 40-a-day rows of the earlier table, so run 625 and a drift preset through the calculator before promising anything.
What this cannot tell you
- The ranges are floors, not worst cases. They assume a flat site, Poisson counts and independent noise; Penlid's calculator says real daily counts are somewhat more variable and that it simulates neither seasonality nor regression to the mean.
- The calculator does not forecast your clicks. It shows what a test on one page could see; this guide gives no way to size a forecast for a group of pages or a whole site.
- No keyword-and-position method. None of the Google pages below gives a click-through-by-position figure, and Google calls position a complex metric that can be misleading if its subtleties are not understood, so we do not build forecasts from them.
- No AI-feature click series. Google reports AI Overviews and AI Mode traffic within the Web search type, and its Generative AI performance report describes impressions without mentioning clicks, so those pages give no AI-feature click series to forecast from.
Questions people also ask
How do you predict SEO results?
Start from the page's recent clicks, allow at least the spread that counting noise and page drift imply, then lay low, base and high assumptions on top, taking each lift from changes you have already measured. The result is a prediction interval, a range meant to contain one future month, and it behaves differently from an estimate of an average. Wikipedia's dice example shows the difference: more rolls pin down the average face value, but the next roll still lands anywhere from 1 to 6.
Why can't SEO results be guaranteed?
Google's own advice for site owners is that nobody can promise the top spot, and its Search Essentials adds that a page can meet all of its requirements and best practices without Google crawling, indexing or serving it. What an agency can commit to is what it controls: the pages it will change, the dates they go live, the date of the first fair reading and the report that follows it.
Why does SEO not produce instant results?
Google says every change takes some time to be reflected on its end, with crawling alone taking anywhere from a few days to a few weeks. Its overall range runs from a few hours for some changes to several months for others, and it suggests allowing a few weeks before judging the work. A client expecting results within days is working from the wrong clock; the guide to how long to wait after a content update covers the windows.
How do you know if SEO is working?
Read the result against the forecast on the date you set. Compare the page with untouched pages over the same dates, then see whether it landed inside the low-to-high range, above it or below it. A rising total alone proves little, because a season or an update can lift many pages together; if the answer is "we cannot tell", say so and report the smallest lift the measurement could have seen, as the guide to how much traffic a page needs explains.
Sources
- Do you need an SEO?, Google, Search Central, read 30 October 2026.
No one can guarantee a #1 ranking on Google
- Google Search Essentials, Google, Search Central, read 30 October 2026.
- Search Engine Optimization (SEO) Starter Guide, Google, Search Central, read 30 October 2026.
- Ask Google to recrawl your URLs, Google, Search Central, read 30 October 2026.
- Google Search's core updates and your website, Google, Search Central, read 30 October 2026.
- Google Search Status Dashboard: summary of ranking updates, Google, Search Status Dashboard, read 30 October 2026.
- Debugging drops in Google Search traffic, Google, Search Central, read 30 October 2026.
- Connect Search Console to Google Analytics, Google, Analytics Help, read 30 October 2026.
- What are impressions, position, and clicks?, Google, Search Console Help, read 30 October 2026.
- AI features and your website, Google, Search Central, read 30 October 2026.
- Generative AI performance report (Search), Google, Search Console Help, read 30 October 2026.
- Poisson distribution, Wikipedia, read 30 October 2026.
- Prediction interval, Wikipedia, read 30 October 2026.
- Methodology: how we measure, and what we can't, Penlid, read 30 October 2026.
A page that has just had a bad four weeks is likely to have a better next four whatever you do to it
- SEO test calculator, Penlid, read 30 October 2026.
It cannot say that a change will improve a ranking, and nothing on this page does
the noise a real change has to stand out from, and it does not shrink as a page gets bigger