Forecast

How the Forecast screen projects clicks: the site model, the keyword baseline, the ranking scenario, the 80% range and the limits of the method.

Forecast answers one question: what would clicks to your tracked keywords look like if their rankings moved? It draws a baseline that follows your site's own trend and seasonality, and a scenario on top of it that you control.

Use it to size a goal or a proposal. It is a what-if calculation, not a promise.

Inputs

The two period choices are kept in the URL, so a forecast link opens with the same windows.

Step 1: the site model

The model learns from the property's daily clicks over the Learn from window, using final days only. Whole-site totals are stored for 16 months on every plan.

• A straight trend, fitted by least squares.
• A weekday pattern (for example, weaker weekends).
• A yearly pattern from the calendar months in the window. This needs a 12- or 16-month window with about a year of data: at least 10 calendar months with 14 or more days each.
• Going forward the trend flattens out. It adds at most half the window's length of slope, so a short window with a steep rise does not run away.
• With fewer than 14 days of data no trend is fitted and the forecast stays at the recent level.

Step 2: baseline and scenario

• Baseline. Each tracked keyword in scope starts from its own last 30 days: clicks, impressions and average position. The keywords then follow the shape of the site model.
• Scenario. Keywords at an average position above 3 and up to 15 move up by the number of positions you set, reached in a straight line over the months you set, never past position 1. Other keywords keep their position.
• Extra clicks. For a moved keyword: monthly impressions × (expected CTR at the new position − expected CTR at the current position). Impressions stand in for search volume, which Search Console does not have.
• Risk. With the risk option on, keywords at position 3 or better slip one rank over the same months, and their loss is counted.

The three CTR assumptions

The expected CTR comes from a generic CTR-by-position curve. The assumption scales the clicks gained from it:

Losses from the risk option are always counted in full.

What the range means

The shaded band around the baseline is an 80% range. It is built from how far actual clicks strayed from the fitted line inside the training window, and it widens the further ahead it looks, faster for short windows. The Site clicks card shows the same range for the whole horizon.

The band covers normal ups and downs. It does not cover algorithm updates, site changes or anything else that was not in the training window. When the window is too short to measure the scatter, no band is drawn.

Reading the screen

• Three cards: Tracked keywords over the horizon against the baseline, Site clicks over the horizon against the same number of days before, and Keywords in the top 3 at the end of the horizon.
• The keyword chart draws Actual, Baseline and Scenario. The site chart draws Actual and Model. Long spans are drawn per week or per 30 days.
• Largest contributors lists up to 7 keywords with Impr. / month, Rank now → then and Extra clicks.
• How this forecast is made under the site chart states the trend per 30 days, which patterns were used and the typical scatter.
• Export has the two chart series and every keyword in scope with its rank and extra clicks at the end of the horizon.

Limits of the method

Keyword history follows your plan, so the Actual line of the keyword chart can be shorter than the site chart. See Plans and billing.