Moving Average Calculator

A moving average is the simplest way to forecast next period's demand: take the last few periods, average them, and call that your forecast. No trend adjustment, no seasonality correction, no statistics degree required — just recent history, smoothed out.

The reason it's still used, decades after fancier methods came along, is that it works fine for a lot of products. Demand isn't always trending or spiking — plenty of SKUs just sell at a roughly steady clip with some week-to-week noise. For those, a moving average smooths the noise without overcomplicating the forecast. It's usually the first thing a planner tries before reaching for exponential smoothing or anything more involved.

Where it falls short: it treats every period in the window identically, old and recent alike, so it lags behind real shifts in demand. If sales are climbing, a moving average will always trail slightly behind reality. That tradeoff — simplicity for responsiveness — is the whole story of this method.

Pick a window size — 3 periods, 6 periods, 12 periods — and average just that many of the most recent data points, ignoring everything older. A shorter window reacts faster to recent changes but jumps around more; a longer window is smoother but slower to notice a real shift.

Historical demand: 100, 120, 110, 130, 125 (oldest to newest)
Window (N): 3

The forecast for the next period is the average of the 3 most recent values.

Try a 5-period window on the same data instead (which, with only 5 points, means averaging everything):

The wider window pulls the forecast down slightly, because it's now including the older, lower demand periods that the 3-period window left out. Neither answer is "correct" — the right window size depends on how quickly this product's demand actually changes.

This method assumes demand is roughly stable, without a strong upward or downward trend and without meaningful seasonality — it has no mechanism to account for either. It also assumes older periods within the window are just as relevant as recent ones, which is rarely quite true.

If demand is trending, a moving average will consistently under-forecast during growth and over-forecast during decline — it always lags. If demand is seasonal, a moving average will smooth right over the seasonal pattern instead of capturing it, unless the window happens to span a full season. In either case, Weighted Moving Average or Exponential Smoothing usually does better.

Using a window that's too wide. Averaging 12 months when demand shifts meaningfully month to month buries real signal under a pile of stale history. The forecast barely reacts to anything recent.

Using a window that's too narrow. Swing the other way — a 2-period average on noisy demand — and the forecast jumps around as much as the raw data did, defeating the point of smoothing it in the first place.

Applying it to trending demand without adjustment.A plain moving average has no way to see a trend coming; it just reports where demand has been. On a fast-growing SKU, that means under-forecasting every single period.

Ignoring seasonality entirely. A window that doesn't span a full seasonal cycle will average summer and winter demand together into a number that describes neither well.

Never checking forecast accuracy. Picking a window size once and never validating it against actual outcomes means you might be carrying a bad forecast for years. Run the results through MAD or MAPE periodically to see how far off you actually are.

There's no universal "right" window size — it depends on how volatile the demand is and how far ahead you need to see. As a starting point, retail planners often use a 3-4 period window for fast-moving SKUs (weekly or monthly granularity) and a 6-12 period window for slower, steadier items where short-term noise matters less than the underlying level. Whatever you start with, check the forecast's actual error with MAD or MAPE and adjust the window based on what you see, not on a rule of thumb alone.

Once you have a baseline forecast, refine and use it:

1. Try weighting recent periods more — Weighted Moving Average gives recent demand more influence without fully committing to Exponential Smoothing.
2. Check accuracy — MAD and MAPE tell you how far off this forecast has actually been running.
3. Feed it into ordering decisions — a demand forecast is only useful once it's driving an EOQ or reorder point calculation.

• Weighted Moving AverageGive recent periods more say than older ones.
• Exponential SmoothingA more responsive alternative that weights recent demand automatically.
• Mean Absolute % ErrorCheck how accurate this forecast actually turned out to be.
• Economic Order QuantityFeed the forecast into an order-size decision.

Books:

• Forecasting: Principles and Practice by Rob J Hyndman and George Athanasopoulos (free online)

Standards & curricula:

• APICS (ASCM) CSCP certification curriculum (demand planning module)

General references for further study, not endorsements — verify course availability and content directly with the provider.