The core forecasting problem is that standard SEO traffic models assume a stable relationship between ranking position and click-through rate, and that assumption breaks down precisely in markets prone to SERP layout volatility. If an enterprise has observed large CTR swings tied to layout changes rather than ranking movement, any forecast built on a static position-to-CTR curve will be systematically wrong during and after those shifts, and the size of the error compounds the more the forecast horizon extends. This is a description of a real, well-understood forecasting failure mode, and the enterprise’s own internal observation of large historical CTR swings is the premise to work from here, not a number to independently re-verify against a public study.
Why position-based forecasting breaks
Traditional SEO forecasting models work by taking a target keyword’s search volume, applying an expected CTR for a projected ranking position (usually from an industry-average CTR-by-position curve), and multiplying through to a projected click and conversion volume. This works reasonably well when the SERP layout for that query type is stable over the forecast period, because the CTR curve was built from historical data that reflected that stable layout.
The moment Google changes what else appears on the page for that query type, the CTR curve itself is invalidated for that query, independent of anything the site did. An AI Overview appearing above organic results, an expanded local pack, a larger shopping carousel, additional People Also Ask entries, or a new knowledge panel can all compress the click share available to organic position 1 without organic rankings moving at all. Google’s own Search Liaison account has repeatedly acknowledged, in public statements about SERP features and ranking, that the search results page evolves continuously and that features are added, removed, and resized on an ongoing basis as part of normal product development. That acknowledgment is the honest, checkable grounding for the mechanism; it does not amount to Google confirming any specific percentage swing, and no forecast should present a specific figure as an official Google-published statistic. Whatever historical CTR shift an enterprise has observed in its own data for its own market is a legitimate internal data point to plan around, but it should be described as the enterprise’s own observed history, not attributed to Google as a documented industry-wide figure.
The specific complications this creates for forecasting
Point forecasts become misleading. A single-number forecast (“this keyword will drive X clicks next quarter”) implicitly assumes the current CTR curve holds for the full forecast period. In a market with a documented history of layout-driven CTR shifts, that assumption is known to be fragile, which means presenting a single point estimate overstates the forecast’s real precision.
Historical CTR data has a shrinking useful shelf life. The CTR-by-position relationship an enterprise measured last year (or even last quarter) may no longer describe the current SERP for the same query, if a layout change happened in between. Forecasts built on trailing CTR data need an explicit check for whether the underlying SERP layout for the relevant query set has changed since that data was collected, not just an assumption that history repeats.
Position gains can be forecast correctly while traffic gains are not, and vice versa. Because CTR shifts here are explicitly independent of position, a team can execute a successful ranking campaign, hit the position target in the forecast, and still miss the traffic target because a layout change compressed the available clicks at that position in the meantime. This decouples two numbers (rank achieved vs. traffic achieved) that most forecasting models treat as tightly coupled, which is a direct source of forecast miss even when the SEO work performed exactly as planned.
Hypothetically, imagine a hypothetical enterprise site we’ll call “Site S” operating in an informational query space prone to AI Overviews. If Site S’s SEO team successfully moved a target keyword from position 4 to position 1 as forecast, but an AI Overview appeared above the results for that query during the same period, hypothetically the traffic gain could come in far below the forecast despite the ranking work performing exactly as planned, simply because the layout change compressed the clicks available at position 1.
Aggregation across many keywords hides the volatility. At the portfolio level, layout changes rarely hit every tracked keyword simultaneously; they tend to hit specific query types (informational queries prone to AI Overviews, transactional queries prone to shopping features, local-intent queries prone to map pack expansion) at different times. A blended forecast across a large keyword portfolio can look stable in aggregate while individual query segments are experiencing large, uncorrelated CTR swings that cancel out in the average and reappear unpredictably as the mix shifts.
What a more robust forecasting approach looks like
Rather than relying on a static, industry-average CTR curve, the more defensible approach is to build the forecast from the enterprise’s own actual CTR-by-position trends, pulled directly from Search Console over a rolling window, segmented by query type or SERP feature exposure where possible. This turns the forecast input from “what CTR should this position get, generally” into “what CTR has this position actually been getting for this specific query cluster, recently,” which is far more resistant to being blindsided by a layout change, because it’s already reflecting whatever layout currently exists.
The second practical adjustment is forecasting in ranges rather than single points, explicitly building a downside scenario that accounts for plausible further SERP feature encroachment and an upside scenario that assumes layout stability. For a market with a documented history of large layout-driven CTR swings, presenting stakeholders with a range and the specific risk driving the range’s width is more honest, and ultimately more useful for planning, than presenting a single confident number that a future layout change can invalidate without any change in actual rankings.
Finally, treating SERP-feature monitoring as an ongoing input to the forecast, not a one-time assumption, matters here: checking periodically whether new features have appeared for the enterprise’s core query set, and re-baselining the CTR assumptions when they have, keeps the forecast responsive to the exact failure mode this market has apparently already experienced historically.
