SEO Traffic Forecast: project clicks, growth scenarios and revenue

The SEO Traffic Forecast projects your organic traffic forward from its own history. It separates seasonality from the underlying trend, fits five kinds of trendline, picks the one that explains your history best without running away, and projects 3, 6 or 12 months ahead with a 95% confidence band. On top of that baseline you can model a growth scenario and turn both into revenue, to set a target or make a case for investment.

Load your history

Under Step 1: Load Historical Data, choose one source:

• GSC Data: press Load 16 months of data to fetch daily clicks for the property over the last 16 months, the most Search Console keeps (16 months of data), whatever the date picker says. Filters at the top apply, so you can forecast one country or one section.
• Upload CSV or Paste Data: one date, value pair per line, with dates as YYYY-MM-DD or YYYY-MM. A monthly value is spread evenly across that month's days. Use this for sessions exported from Google Analytics, or for more than 16 months of history.

The note on the page is worth repeating: Search Console data is clicks, not sessions. The two usually move together but are not the same number, so label the forecast with the metric you loaded.

Configure the forecast

Under Step 2: Configure forecast:

• Granularity: Monthly or Weekly periods.
• Forecast Months: 3, 6 (default) or 12.
• Trendline type: Auto-select (best R²), or force Linear, Logarithmic, Exponential, Power or Polynomial (degree 2).
• Growth rate (% per year) and Added sessions / month for the growth scenario.
• Conversion rate (%) (2 by default), Average order value (50 by default) and Currency (USD, EUR, GBP, INR, AUD, CAD or JPY) for revenue.

Everything recalculates as soon as you change an input.

How the forecast is calculated

1. Periods. Daily values are summed into calendar months or weeks. A partial first or last period is dropped, so the current, unfinished month never drags the trend down. At least 4 complete months or 8 complete weeks are needed.
2. Seasonality. A straight line is fitted through the periods, and for each calendar month the average ratio of actual to that line becomes its seasonal index (1.0 is an average month). Indices are normalised to average 1, kept between 0.4 and 2.5, and months without data get 1. Weeks use the month they start in.
3. Trend. The history is divided by its seasonal indices, and the five trendlines are fitted to what remains by least squares (exponential and power in log space).
4. Selection. Auto-select takes the highest R² among trendlines whose projection stays between zero and four times your historical peak; when two are within 0.02 R², the one with the lower MAPE (mean absolute percentage error) wins. The Model table shows every trendline's R², MAPE and whether it was eligible.
5. Projection. The chosen trend is extended and the seasonal index put back on each future period. The 95% band is ±1.96 residual standard deviations, widening the further ahead it goes.
6. Growth scenario. The yearly growth rate is compounded per period from the first forecast period, and added sessions are a fixed monthly uplift (prorated for weeks).
7. Revenue. Sessions × conversion rate × average order value, for the baseline and the growth scenario.

Read the results

The cards show the Current Month (the last complete period), the Baseline Forecast and Growth Forecast totals over the horizon, and the Growth Revenue. Baseline vs growth charts history, baseline and growth with the confidence band dashed, and Seasonal index shows your strong and weak months. The Forecast table lists every period with actual, fitted trend, seasonal index, baseline, the 95% bounds, the growth scenario and both revenue figures, and exports as CSV.

Use it with care

A forecast assumes the future behaves like the past. Core updates, migrations, new sections and seasonal events that only happen some years are not in the model. A wide band means the history is noisy, and a forecast with 16 months of data has seen each calendar month at most twice, so treat its seasonality as a sketch. Check the selected trendline: a polynomial that bends sharply at the end is a warning, even when its R² is high.

How AutoPSEO helps

The Added sessions / month input is the place to model new content. If the content engine is adding articles at a steady weekly pace, estimate their monthly clicks from the articles already live and enter that as the uplift. A content group for the engine's section then shows whether the new pages keep up with the estimate.