How long before a change shows up in your own numbers
A forecast describes when discovery conditions change. Your analytics describe when enough people had already moved for the change to become visible. Those are weeks apart, and the gap is where most wrong conclusions get drawn.
The lag is not one delay, it is four stacked
A change lands in an assistant's behaviour. Then it has to reach the people who ask that kind of question, which depends on how often they ask. Then enough of them have to act for the movement to clear the noise in a weekly figure. Then your reporting window has to close.
For a business with steady traffic that is usually two to six weeks between the change and the point where anyone could reasonably say it happened. For a business with low volume it can be a quarter, and for a seasonal one the effect may not become legible until the season repeats.
What the lag causes people to conclude
That the forecast was wrong. Checked a week later, nothing moved, so the call is dismissed. The record shows this is the single most common reason a correct forecast is written off, and it is why grading windows are stated in advance rather than chosen afterwards.
That the fix worked. Work done in response to a change frequently completes just as the underlying movement was going to show up anyway. Attribution here is genuinely hard, and anyone selling certainty about it is selling something.
That nothing is happening. The most expensive one. A business watching a flat weekly number through a real change concludes it is unaffected, and finds out two quarters later.
What to watch instead while you wait
Leading indicators move before volume does. Whether you are named in answers to your own category questions at all. Whether the description attached to your name is still right. Whether a competitor has started appearing where you used to. None of those require enough traffic to be statistically visible, and all of them change first.
That is the argument for measuring position rather than only outcomes: outcomes are the last thing to move, and by the time they do, the cause is a month behind and hard to name.
What we hold, and what we do not
Every forecast states its window before the fact, and grading uses that window rather than a convenient one. A call graded early would flatter the record and a call graded late would hide the misses.
What we do not do is estimate your specific lag. It depends on your volume, your seasonality and your reporting cadence, none of which we hold, and a general number applied to a particular business would be wrong in a direction nobody could see. Where a change is expected to be slow to surface, the forecast says so instead of implying it will be visible next week.
When to ignore a forecast covers whether to act at all, why referral share is a floor covers what your numbers can and cannot see, and both sit under AI Visibility Forecasting.