Forecast Accuracy
Forecast accuracy measures how close your revenue prediction landed to reality. Here's how to score it with MAPE and what counts as good for a DTC store.
Forecast Accuracy
How close a revenue forecast was to actual results, usually measured as the percentage error between predicted and observed sales.
Forecast accuracy is the gap between what you predicted and what actually happened, expressed as a percentage. For an online store, the forecast is usually monthly or weekly revenue, and the standard scoring metric is Mean Absolute Percentage Error (MAPE) — the average absolute difference between forecast and actual, divided by actual.
It sits inside the broader revenue intelligence stack alongside pipeline velocity, cohort retention, and channel-level attribution. A monthly MAPE under 10% is good for an established store; under 5% is excellent. Anything north of 20% means your plan, hiring, and inventory commitments are running on guesses.
Forecast accuracy matters because every downstream decision — inventory buys, paid-media budgets, hiring, cash-flow runway — is sized against the forecast. A 25% miss on a €800k month is €200k of stock you either over-ordered or sold out of two weeks early.
It's also the cleanest scoreboard for your finance and growth teams. You can't improve a forecast you don't grade. Track MAPE month over month and the bias (are you consistently over or under?) and the forecast stops being a vibe and starts being a system.
MAPE = (1/n) * Σ |Actual − Forecast| / |Actual| × 100
n
Number of periods
How many months (or weeks) you're scoring.
Actual
Actual revenue
Revenue your store actually booked in the period.
Forecast
Forecasted revenue
What you predicted at the start of the period.
An apparel store grading its last three monthly forecasts.
Jan forecast / actual: €420,000 / €455,000
Feb forecast / actual: €380,000 / €362,000
Mar forecast / actual: €510,000 / €498,000
→ MAPE = (7.69% + 4.97% + 2.41%) / 3 ≈ 5.0%
A 5% MAPE puts this store at the top end of the DTC range — tight enough that buyers can commit to 60-day stock with confidence.
Two notes on reading MAPE. First, average it over at least three periods — a single tight month can hide a systematic bias. Second, also track signed error (Actual − Forecast) so you know whether you tend to under-forecast (and stock out) or over-forecast (and sit on inventory).
Monthly revenue forecast accuracy (MAPE) benchmarks for online stores
| Forecast maturity | Typical MAPE | What it usually means |
|---|---|---|
| First 90 days, no historical model | 25–40% | Spreadsheet guesswork; missing seasonality and channel mix. |
| Basic trend + seasonality model | 12–20% | Captures the obvious cycle but misses promo lifts and stock-outs. |
| Channel-level model with promo calendar | 7–12% | Solid baseline — good enough for inventory planning. |
| Cohort + channel model, weekly tracked | 4–7% | Top-quartile DTC; finance can plan cash to the week. |
| Best-in-class with live signal correction | 2–5% | Rare — usually a dedicated analyst plus clean attribution. |
If you're new to forecasting, the fastest accuracy jump comes from importing 12–24 months of historical GA4 and order data into the model — it kills the cold-start problem and gives the seasonality curve something real to fit against. Most stores cut their MAPE in half within two cycles once they do this.
Forecast accuracy FAQ
For monthly revenue, a MAPE under 10% is good and under 5% is excellent. Weekly forecasts are harder — expect 2–3 percentage points worse than monthly because random variance has less time to average out.
Absolute values stop positive and negative misses from cancelling each other out. A forecast that's +20% in January and −20% in February has zero average error but is clearly broken — MAPE correctly scores it at 20%.
Accuracy (MAPE) tells you how far off you are on average. Bias tells you the direction — whether you systematically over- or under-forecast. You want low accuracy error AND a bias close to zero; track both.
Score every period as soon as actuals close — monthly for the headline number, weekly if you're forecasting that granular. Review the rolling 3-month and 12-month MAPE in your monthly business review.
Only if you feed back the errors. A forecast that gets re-run on the same model every month with the same blind spots will stay at the same MAPE forever. Improvement comes from analysing the misses and adjusting the model.
Revenue intelligence is the broader category — attribution, cohort behaviour, channel mix, and forecasting all sit inside it. Forecast accuracy is the scoreboard that tells you whether the rest of your revenue intelligence is actually predictive.
Aim for at least 13 months — that gives you one full seasonal cycle plus the current month for trend. With less than 6 months you're effectively extrapolating, and a MAPE under 15% is unlikely until you've lived through a peak season.
Only if you don't model them. Treat promo periods as a separate forecast with its own uplift assumption, then score promo MAPE separately from baseline MAPE. Lumping them together hides where the real error is coming from.
Forecast both. Revenue is what finance plans against, units are what the warehouse plans against, and discount-heavy months can move them in opposite directions. Score MAPE on each.
Yes — run a separate MAPE per channel. A channel where you can predict revenue within 5% is a planning asset; one with a 30% MAPE is a coin flip and shouldn't anchor your media plan, regardless of its absolute ROAS.
See Metricuno on your data
Bring your stack — Google Analytics, Stripe, a CRM, anything — and we'll walk through the metric tree that turns your funnel into one number.