Weighted Moving Average Calculator

A plain moving average treats every period in its window as equally important — a sale from six weeks ago counts exactly as much as yesterday's. That's often wrong. A weighted moving average fixes it by letting you assign more influence to recent periods and less to older ones, while still averaging over a defined window.

This sits in the middle of the forecasting toolkit — more responsive than a plain moving average, but still simple enough that you're choosing the weights by hand rather than letting a formula derive them automatically the way exponential smoothing does. It's a reasonable next step once a plain moving average starts lagging noticeably behind a trending product.

Each historical value gets paired with a weight at the same position — the first weight applies to the first (oldest) value, and so on, with the last value and last weight being the most recent pair. Weights need to sum to 1.0, and the more recent periods should generally carry more weight if you're trying to track a moving trend.

Values: 100, 120, 140 (oldest to newest — a clear upward trend)

First, weight every period equally:

Now weight the most recent period much more heavily:

Equal weighting produces a forecast of 120 — right at the midpoint, ignoring the trend. Weighting toward the most recent period pulls the forecast up to 126, closer to where demand actually seems to be heading.

Neither is automatically "right" — the equal-weight version is a safer bet if last period's jump might be a fluke; the recency-weighted version is the better bet if the trend is real and likely to continue.

This method assumes you can pick reasonable weights, which is harder than it sounds — there's no formula that tells you the "right" weights, just judgment and testing. It still averages over a fixed window, so it shares the plain moving average's blind spot for anything happening outside that window, like a seasonal pattern longer than the window itself.

Picking weights arbitrarily. "50/30/20 sounds about right" isn't a method — test a few weight sets against actual outcomes with MAD or MAPE before committing to one.

Forgetting weights must sum to 1.0. Weights that don't sum to 1.0 don't represent a true average and will skew the result in ways that aren't obvious just from looking at the numbers.

Overweighting a single outlier period. If the most recent period had an unusual spike or dip, a weight scheme that leans heavily on it will chase noise instead of tracking a real trend.

Using the same weights for every product. A fast-changing SKU might justify heavy recency weighting; a stable one might not need it at all. One weight scheme for the whole catalog is rarely optimal for all of it.

Not comparing against a plain moving average. If the weighted version isn't actually beating a simple average on forecast accuracy, the added complexity isn't earning its keep — check with MAD or MAPE.

There's no standard weight scheme — it depends entirely on how much recent periods should matter for a given product. As a starting point, many planners try something like 50/30/20 (most recent to oldest across a 3-period window) for moderately trending items, and adjust from there based on measured forecast accuracy rather than intuition alone.

Once you have a forecast, validate and use it:

1. Check accuracy — run the forecast through MAD or MAPE against actual outcomes.
2. Consider Exponential Smoothing if picking weights by hand feels arbitrary — it derives the weighting automatically from a single alpha parameter.
3. Feed the forecast into ordering decisions — EOQ and Reorder Point both need a demand number to work from.

• Moving AverageThe simpler version, where every period counts equally.
• Exponential SmoothingWeights recent demand automatically instead of picking weights by hand.
• Mean Absolute % ErrorCheck whether your chosen weights actually improve accuracy.

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.