How a moving average forecast works
A moving average forecast projects the next period by averaging the actual results of a fixed number of recent periods, then rolling that window forward as each new period closes. It is the simplest of the revenue forecasting models a team can build, and it needs nothing more than a history of closed revenue or bookings.Two versions cover most use:
- Simple moving average (SMA): every period in the window counts equally. A three-month SMA forecasts next month as the average of the last three months. - Weighted moving average (WMA): recent periods count more than older ones, so the forecast reacts faster to a fresh trend while still smoothing out noise.
If the last three months of new bookings were $420K, $460K, and $500K, a three-month SMA forecasts $460K for next month. A weighted average that favors the most recent month lands higher, closer to $480K.
Why it is the bar predictive models have to clear
A moving average is the baseline you measure every fancier model against. Before you trust a regression or a machine learning model, you check one thing: does it beat the moving average on out-of-sample forecast accuracy? If a model that took weeks to build cannot outperform a three-month average, the model is noise dressed up as insight.
This is why the moving average earns its keep even at a company that will never ship it to the board. It sets the floor. ORM takes four to six weeks to fully train a model on a company's historical sales performance, then targets 95% forecast accuracy on new and expansion business and updates it as the quarter progresses. The moving average is the number that target has to beat to justify the model over a spreadsheet.
Where a moving average forecast breaks
The method assumes the near future looks like the recent past, so it fails exactly when that assumption fails.
- It ignores seasonality. ORM's data shows Q2 and Q4 run stronger than Q1 and Q3, and the third month of a quarter closes more than the first two. A flat average smears those patterns together and under-calls the strong periods. - It reacts late to change. When a new competitor pressures pricing or a rate change slows buyers, deal sizes and win rates shift before the average catches up. The forecast keeps quoting the old world for another window or two. - It sees only what already closed. A moving average knows nothing about the pipeline in front of you or the in-quarter deals not yet created, so it cannot explain how a quarter is going to happen.
Treat it as a benchmark and a sanity check, not as the number you commit to. The value of a real model is that it clears this bar on day one and keeps clearing it as conditions move.
Frequently Asked Questions
What is a moving average forecast?
A moving average forecast predicts the next period by averaging the actual results of the last few periods, then rolling that window forward as each new period closes. It is the simplest quantitative forecast a revenue team can build and needs only a history of bookings or MRR.
How do you calculate a simple moving average forecast?
Add the values of the periods in your window and divide by the number of periods. If the last three months of new bookings were $420K, $460K, and $500K, a three-month simple moving average forecasts $460K for next month. A weighted version gives recent months more influence, so the same data lands closer to $480K.
Is a moving average forecast accurate enough for SaaS revenue?
On its own, no. A moving average assumes the near future looks like the recent past, so it misses seasonality and reacts late when pricing pressure or win rates shift. Use it as a benchmark. ORM takes four to six weeks to fully train a model on a company's historical sales performance and targets 95% forecast accuracy on new and expansion business, updated as the quarter progresses.
When should you use a moving average forecast?
Use it as the baseline you measure every other model against, and as a sanity check on stable, high-volume series like steady renewals. If a more complex model cannot beat a three-month average on out-of-sample accuracy, the added complexity is not paying for itself.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like moving average forecast into prescriptive action for your team.
Schedule a Demo