Home » Glossary Terms » Weighted Moving Average

Weighted Moving Average

Weighted Moving Average - Featured Image
Note
A weighted moving average assigns higher importance to recent sales periods when forecasting demand, letting sellers react faster to trend changes than a simple moving average allows. It bridges the gap between equal-weight averaging and exponential smoothing.

How the weighted moving average works

A weighted moving average assigns different importance levels to each period in the averaging window rather than treating all periods equally. Recent months get higher weights, older months get lower weights. This makes the weighted moving average more responsive to demand shifts than a simple moving average while remaining more transparent than exponential smoothing.

For FBA sellers running 50 to 200 SKUs, the weighted moving average works well when you want more control over how much influence each period has on your demand forecast. You explicitly choose the weights, so the logic is auditable. If your category buyer or operations manager asks why you ordered 600 units instead of 400, you can show exactly how the weighted moving average calculation arrived at that number.

The trade-off is maintenance. With a weighted moving average, you need to decide on a weighting scheme for each SKU or product group, and those weights may need to change as demand patterns evolve. That manual tuning overhead is why many sellers eventually graduate to exponential smoothing, which achieves a similar effect with a single alpha parameter.

Weighted moving average formula

WMA = (w1×D1 + w2×D2 + ... + wn×Dn) / (w1 + w2 + ... + wn)
VariableMeaning
w1...wnWeights assigned to each period (higher = more influence). Common scheme: 3-2-1 for 3 periods.
D1...DnActual demand for each period (D1 = most recent)
nNumber of periods in the window

Example: a $45 fitness product

You sell a resistance band set at $45 ASP with a 55-day lead time from Guangzhou. Your last 3 months of sales: Month 1 (oldest) = 350, Month 2 = 420, Month 3 (most recent) = 480. Let’s compare equal weights vs. a 3-2-1 weighted moving average:

MethodCalculationForecast
Simple MA (equal)(350 + 420 + 480) / 3417
WMA (3-2-1)(3×480 + 2×420 + 1×350) / 6438

The weighted moving average gives 438 vs. the simple average of 417. That 21-unit difference reflects the upward trend in recent months. With a 55-day lead time (about 1.8 months), the WMA forecasts lead time demand of 788 units vs. 751 from the simple MA. At $45 ASP, the WMA orders an extra $1,665 in inventory to cover the growth pattern.

If you used a 4-3-2-1 scheme over 4 months (even stronger recency bias), the forecast would shift further toward recent actuals. The right weighting scheme depends on how quickly your product’s demand changes.

FBA-specific considerations

The weighted moving average is particularly useful for FBA sellers heading into Q4. From October through December, demand patterns shift fast, and a simple equal-weight average lags behind. Heavier weights on recent periods (3-2-1 or 4-3-2-1) help your forecast catch up to holiday demand without waiting for an entire averaging window to roll over.

Post-Prime-Day cleanup is another strong use case. If Prime Day produced a 3x sales spike, that single inflated period will distort an equal-weight moving average for months. With a weighted moving average, you can either set the Prime Day month’s weight to zero (excluding the spike entirely) or weight it lower than the trailing months that better represent baseline demand.

One FBA-specific gotcha: Amazon’s storage tier rules change quarterly based on your IPI score. A weighted moving average that aggressively up-weights last month can recommend orders too large for your current storage allocation. Always cross-check WMA-driven reorder quantities against your storage limits before placing the PO.

Common mistakes

  1. Using arbitrary weights without testing. Sellers often pick 3-2-1 because it sounds reasonable, then never validate the weights against actual forecast accuracy. Backtest your weighting scheme by comparing forecasts to actuals over the past 6 months. If 4-3-2-1 produces lower MAPE than 3-2-1, switch.
  2. Over-weighting recent data during anomalies. If last month was a Lightning Deal month with 2x normal sales, a heavy 3-weight on that period inflates your forecast. The next month’s reorder quantity will be too high, leading to excess inventory after the promotion fades. Either exclude promotional months or temporarily reduce their weight.
  3. Not adjusting weights seasonally. The optimal weighting scheme for a stable summer month is different from the optimal scheme heading into Q4. Many sellers set weights once and forget them. Review weights at least quarterly, or use exponential smoothing with adaptive alpha to handle the adjustment automatically.

Related terms

How Profit Hawk handles this
Profit Hawk automatically tunes period weights based on each SKU's demand pattern, so you never have to pick a 3-2-1 vs. 4-3-2-1 scheme by hand. It also flags SKUs where a weighted moving average is being beaten by exponential smoothing on accuracy. See the forecasting engine.

Frequently asked questions

What is a weighted moving average?

A weighted moving average is a forecasting method that assigns higher importance to recent sales periods and lower importance to older ones, then averages them to predict next period's demand. Unlike a simple moving average that treats all periods equally, the weighted version reacts faster to demand changes.

How do I choose weights for a weighted moving average?

Common starting schemes are 3-2-1 for 3 periods (most recent gets weight 3) or 4-3-2-1 for 4 periods. Test schemes against historical data: backtest each weighting scheme on 6 months of past sales and pick the one with the lowest forecast error. There is no universally optimal scheme.

When should I use a weighted moving average instead of exponential smoothing?

Use a weighted moving average when you need transparent, auditable forecasts that anyone on your team can reproduce in a spreadsheet. Use exponential smoothing when you want the simplicity of a single tunable parameter and minimal maintenance overhead across hundreds of SKUs.

How many periods should I include in a weighted moving average?

3 to 6 periods works for most FBA products. More than 6 periods dilutes the recency benefit, since older periods get small weights regardless. For seasonal products, 12 periods with seasonally-tuned weights captures annual patterns better than a short window.

Does a weighted moving average handle promotional spikes?

Not automatically. If a promotional month falls in your weighting window with a high weight, it will inflate the forecast. Either exclude promotional periods from the data feed or set their weight to zero in the calculation.

Keep going

[ph_glossary_nav]

Nine free Amazon FBA calculators — plain English, no signup.