Use case

Churn forecasting when the number that matters is a rate

Forecast monthly churn from a CSV, keep the denominator stable, and use the range to plan retention work instead of reacting to one bad month.

Run the sample Put this in production

Churn is a rate, and a rate is easy to miscalculate. If the denominator changes because of a pricing change or a cohort shift, a stable churn process can look like a crisis. Forecast a consistently defined rate, then apply it to the customer base you actually expect.

The sample is monthly logo churn in percent. If you need revenue churn, forecast that separately; logo churn and revenue churn can move in opposite directions when small and large customers behave differently.

The sample series

The studio below is preloaded with this monthly logo churn (%) series and a 6-step horizon. It is an illustrative example, not a real company’s data. Replace it with your own CSV to run the same workflow.

Show the CSV

month,churn_pct
2024-07-01,3.8
2024-08-01,3.6
2024-09-01,3.9
2024-10-01,4.1
2024-11-01,3.7
2024-12-01,3.5
2025-01-01,4.4
2025-02-01,4.0
2025-03-01,3.8
2025-04-01,3.9
2025-05-01,3.6
2025-06-01,3.4
2025-07-01,3.7
2025-08-01,3.5
2025-09-01,3.8
2025-10-01,3.9
2025-11-01,3.6
2025-12-01,3.3

CSV forecast studio

Forecast your churn forecasting series.

This studio is preloaded with the monthly logo churn (%) example above. Replace it with a CSV of your own history, choose a horizon, and run the forecast. The public demo is free while the Hugging Face Space is available; the API is the paid product.

1 · Paste or upload a CSV2 · Pick a horizon3 · Download the 21 quantiles

Steps

1 credit per API call · the web demo is free while the public Space is available

What to watch in the output

Keep the denominator definition fixed

Churn from a growing base is not comparable with churn from a shrinking one. Use the same cohort and the same denominator for every month in the series.

Logo churn and revenue churn are different

A small customer leaving and a large customer downgrading can produce the same logo churn but very different revenue impact. Forecast the one your decision uses.

Small denominators make rates noisy

If you have 40 customers, one churn is 2.5 percentage points. The band should be wide; do not overreact to a single month inside it.

Common mistakes

—Forecasting churn counts instead of rates. The count is driven by growth, not retention.
—Mixing voluntary and involuntary churn. Failed payments and cancellations have different levers.
—Ignoring a pricing change. A price increase can create a temporary spike that is not a new baseline.

How to put it to work

1. Start with the sample. Run the preloaded series once so you see the output shape before you debug your own CSV.
2. Replace it with your history. Keep one timestamp column and one or more numeric columns. Mark outages and stockouts as empty cells, not zeros.
3. Backtest before you plan. Hold out the last few periods, compare the forecast with what happened, and check whether the actuals land inside the 80% band.
4. Move it into production. Use the REST API for a nightly job or an internal tool. One credit per successful call; failed calls are refunded.

FAQ

Can I forecast revenue churn?

Yes, provide monthly revenue churn as the series. Do not mix it with logo churn in one column.

How many months of history do I need?

At least 12 months is a reasonable starting point for a monthly rate. Eighteen to twenty-four months is better if the business has seasonality.

Does the model explain why churn changed?

No. It forecasts the pattern in the series. A causal explanation needs cohort analysis, interviews or experiments, not a univariate forecast.

Need this as an API?

The same forecast is available as a metered REST endpoint with JSON and a downloadable quantile CSV.

See the API