Inventory questionhow accurate are my inventory forecasts

How accurate are my inventory forecasts?

Your inventory forecasts are as accurate as the gap between predicted and actual sales at the SKU and period level, typically measured with metrics like MAPE. By comparing forecast versus actual each month and tracking these errors, you can see which products are well-modeled and which need better data or simpler assumptions.

Overview

Forecasts always contain error; the real question is whether that error is acceptable for your business. A forecast that is wrong by 5% drives very different outcomes from one that is wrong by 50%. Measuring forecast accuracy lets you decide whether to simplify, refine, or replace your methods before they create costly stockouts or dead stock.

Compare forecast versus actual at SKU level

Export your forecasted quantities and actual sales from Shopify Analytics → Sales by product, then calculate the percentage error using the MAPE formula. Doing this per SKU shows where your model tends to over- or under-predict, instead of relying on a single storewide accuracy number.

Track accuracy over multiple periods

A single month can be noisy, so calculate MAPE across several periods, such as at least three months, to see whether your accuracy is improving. Consistently high errors signal that you may be overfitting short-term trends or ignoring known drivers like BFCM’s 2–4× demand uplift.

Relate forecast accuracy to business impact

Even a small percentage error can matter if it consistently under-forecasts top sellers and leads to stockouts, hurting your ability to hit a 6–8× turnover. Tie forecast errors back to missed revenue or excess inventory so you know which categories deserve the most improvement effort.

Use accuracy feedback to refine models

Once you know MAPE and other error metrics, you can test whether simpler models or Synplex-driven forecasts reduce the error. Iterate on lookback windows, seasonality assumptions, and safety stock policies from /calculators/safety-stock-calculator using measured accuracy as the scorecard.

Formula

Mean Absolute Percentage Error (MAPE)

MAPE (%) = (sum(|Forecast - Actual| / Actual) / n) * 100

  • Forecast Forecast: Forecasted demand for a period
  • Actual Actual: Actual demand for the same period
  • n Number of periods: Number of periods included in the calculation

Worked examples

Cosmetics brand evaluating monthly forecast accuracy

  • Forecasted units for a SKU: 1,000 units
  • Actual units sold: 900 units
  • Number of months measured: 3 months
  1. 1. For one month, compute percentage error as |1,000 − 900| / 900 = 0.111.
  2. 2. Convert to a percentage by multiplying by 100 to get 11.1%.
  3. 3. Repeat this process for three months and average the results to get MAPE.

Result: If the three months’ percentage errors are 11.1%, 9.0%, and 10.0%, the MAPE is about 10.0%.

A MAPE of roughly 10% may be acceptable for many categories, but you should compare it to the business impact and cost of improving further.

How to apply this in Shopify

  • Export product-level sales from Shopify Analytics → Sales by product to gather the Actual values needed for forecast accuracy calculations.

  • Use Shopify Analytics → Reports → Inventory to see how forecast misses have affected days of supply and potential dead stock.

  • Filter Admin → Products → Inventory to focus on high-revenue SKUs when evaluating forecast performance.

  • Apply Inventory adjustments after cycle counts so your accuracy calculations start from correct on-hand numbers.

  • Track any manual overrides to forecasts in your planning tool so you can later see whether overrides improved or hurt accuracy.

Common mistakes

Looking only at total revenue accuracy

Focusing on storewide sales accuracy hides large SKU-level errors that still create stockouts and dead stock.

Fix: Calculate MAPE per key SKU or collection and fix issues where errors are biggest, even if top-line revenue looks close to plan.

Ignoring accuracy feedback

Continuing to use the same forecasting approach without measuring outcomes prevents learning and improvement.

Fix: Regularly compute MAPE and other metrics, then iterate on models, lookback windows, and seasonality assumptions based on those results.

Overreacting to a single bad month

Making large changes after one outlier month can destabilize otherwise solid forecasts.

Fix: Evaluate accuracy over multiple periods, such as at least three months, before declaring a method broken.

Frequently asked questions

Related resources

Related questions

Related guides

Related calculators