Back buttonBack to all news
Retail Insights
November 20, 2026 10:05 AM

Benchmarking Forecast Accuracy: What 'Good' Looks Like for Mid-Market Fashion and Lifestyle Retailers

More articles:

"Is 78% forecast accuracy good?" is one of those questions that sounds simple until someone has to answer it in a trading meeting. Good compared to what? Measured which way? Across the whole range, or just the lines that matter? Most mid-market fashion and lifestyle retailers can quote a forecast accuracy percentage. Far fewer can defend it, benchmark it against a credible standard, or explain what it would actually take to move it.

That gap matters more than it used to. CFOs and Commercial Directors want forecast accuracy numbers that hold up under scrutiny, not a single blended figure pulled together the night before a board pack is due. This guide sets out the three metrics worth tracking, what "good" genuinely looks like at each stage of forecasting maturity, and the most common ways forecast accuracy gets measured in a way that flatters the number rather than the business.

‍

Why a Single Forecast Accuracy Score Isn't Enough

A blended, business-wide forecast accuracy figure is easy to put on a slide and almost impossible to act on. It smooths over the categories quietly running 20 points below target and the ones that only look strong because they are easy to forecast in the first place. Two retailers can report the same 75% headline number and be in completely different positions: one with tight, consistent demand forecast accuracy across the range, the other propped up by a handful of predictable, slow-moving lines while newness and promotional stock run far worse.

The fix is not a more complicated dashboard. It is choosing a small number of metrics that describe accuracy, bias and availability separately, then reading them together rather than collapsing them into one score a planner has to unpick before anyone can act on it. That discipline is also what turns a forecast accuracy figure into something a CFO can defend externally, not just report internally.

‍

The Three Metrics Worth Tracking

Three figures do most of the work when it comes to understanding whether a forecast is actually good, and each answers a different question. Together they answer the practical question of how to measure forecast accuracy without needing a data science team, showing whether a forecast is close, whether it is consistently skewed in one direction, and whether it is actually delivering stock to the shelf.

Mean Absolute Percentage Error, or MAPE

MAPE answers the most basic question: on average, how far off was the forecast? It is the most widely used forecast accuracy metric in retail because it is easy to explain and easy to compare period to period, though it treats every miss as equally bad regardless of direction.

Forecast accuracy  formula: MAPE = average of  |Actual − Forecast| ÷ Actual, expressed as a percentage

Forecast bias

Bias answers a different question: is the forecast consistently too high or too low? A retailer can carry a healthy MAPE and still have a serious bias problem if the same categories are chronically over-forecast, tying up cash in slow stock, or under-forecast, driving lost sales that never show up in a simple accuracy score. Tracking bias alongside MAPE is one of the fastest ways of improving forecast accuracy without changing the underlying model at all, because chronic bias is usually a process or assumption problem rather than a data one.

Bias formula: Bias = total (Forecast − Actual) ÷ total Actual, expressed as a percentage

Fill rate

Fill rate measures something more practical again: how often was the right stock actually available when a customer wanted it? A forecast can be statistically accurate and still fail commercially if lead times, minimum order quantities or allocation rules mean the stock never lands in the right place at the right time. Fill rate is the metric that connects forecast accuracy back to the number that ultimately matters, which is sales the business did not miss. It is also easiest to track consistently when top-down commercial targets and bottom-up forecasts sit in one reconciled place, the approach PlanIT's WSSI is built around, rather than being argued over across separate spreadsheets.

‍

What "Good" Actually Looks Like

Benchmarks only help if they are honest about the fact that "good" depends on where a retailer is starting from. A business moving off spreadsheets for the first time should not be judged against the same bar as one with several years of clean, structured forecasting data behind it. The table below sets out a simple way to read the three metrics together across four stages of maturity, so a result can be judged against a realistic next step rather than an arbitrary target. These are illustrative starting tiers for weekly, line-level forecasts rather than published industry benchmarks, so treat them as a guide to direction rather than a pass mark.

Stage of Maturity MAPE Bias Fill Rate
Reactive Above 40% Beyond ±15% Below 85%
Developing 30% to 40% ±10% to 15% 85% to 90%
Controlled 20% to 30% ±5% to 10% 90% to 95%
Forward-Looking Below 20% Within ±5% Above 95%

 

A retailer sitting in the "developing" column on all three metrics does not have a forecasting crisis. It has a clear, sequenced set of things to fix, usually starting with bias before MAPE, because a consistently skewed forecast is normally faster to correct than a noisy one.

‍

Common Mistakes That Distort the Number

Even well-run planning teams can end up reporting a forecast accuracy figure that looks better, or worse, than reality because of how it was calculated rather than how the business actually performed. Six mistakes come up again and again.

  1. Measuring accuracy only at the aggregate level. A strong category-level MAPE can hide SKU-level chaos that only shows up once a customer cannot find their size, which is exactly why an exception-based forecasting engine like PlanIT's Line Card flags outliers at SKU level instead of waiting for the aggregate number to move.
  2. ‍Ignoring the denominator problem. Small absolute unit variances on slow selling lines with near-zero actual sales can swing a MAPE score by several points without anything in the business materially changing.
  3. ‍Comparing accuracy across categories with very different demand patterns. A continuity line and a six-week fashion drop should never be benchmarked against the same MAPE target, because volatility, not planning skill, explains most of the gap between them.
  4. ‍Reviewing forecast accuracy too infrequently to catch drift. A quarterly review can let a bias problem compound for months before anyone notices it in the numbers rather than in the stockroom. PlanIT's early warning framework for merchandising teams sets out the weekly signals that catch this kind of drift earlier.
  5. ‍Measuring against the wrong forecast. It’s relatively easy to accurately reforecast next week’s sales when promotional activity, opening stock position and trends are known, but if it’s too late to amend a promotion or distribute stock, an accurate forecast produced the week before is of little use.
  6. ‍Focusing on one number only. Once a forecast is set, planning teams can pull various levers to ensure it is met, often sacrificing one target to achieve another. For example, deeper promotions can help to hit a tough category sales forecast, but this comes at the expense of margin and stock forecasts.

    ‍

Turning Forecast Accuracy Into a Repeatable Process

Improving forecast accuracy rarely comes down to a better algorithm. It usually comes down to a planning process that reconciles top-down commercial targets with bottom-up, line-level detail automatically, so the two versions of the truth are checked against each other every trading cycle rather than argued over in a meeting.

The same discipline applies at SKU level, where exception-based review replaces manually checking every line by hand. The retailers in PlanIT's customer stories run this kind of process at a scale no spreadsheet could support, from TG Jones forecasting 11 million combinations a day to Amazon Fashion previously forecasting billions of KPIs every week.

None of this requires ripping out an existing system to get started, and demand forecast accuracy tends to improve fastest when bias is tackled before the model itself.

‍

Where This Leaves Planning Teams

A forecast accuracy score only earns trust once it can be broken down, benchmarked and explained. That means tracking MAPE, bias and fill rate separately rather than blending them into one number, judging results against a realistic stage of maturity rather than an arbitrary target, and treating a low score as a sequence of fixable process issues rather than a verdict on the team. Retailers wanting a clearer picture of what improving forecast accuracy would be worth in practice can use PlanIT's benefit calculator to estimate the likely financial impact before committing to any change.

‍

See How Your Own Numbers Benchmark

If you want a second opinion on what your MAPE, bias and fill rate figures actually say about your planning maturity, PlanIT's team can talk it through against the benchmarks in this guide. Get in touch to arrange a conversation.

 

Frequently Asked Questions

How do you measure forecast accuracy in retail?

The forecast accuracy formula most retailers default to is MAPE, calculated as the average absolute difference between forecast and actual demand as a percentage of actual demand, alongside bias to check for consistent over or under-forecasting and fill rate to check stock actually reached the shelf.

What is a good forecast accuracy MAPE score?

It depends on category volatility and planning maturity, which is why a single target rarely works well. A continuity range in a mature planning process should sit well ahead of a fast-moving fashion drop, so the benchmark tiers in this guide are deliberately built around stage of maturity rather than one fixed number.

How is forecast accuracy different from stock accuracy?

Forecast accuracy measures how close a prediction was to actual demand. Stock accuracy measures whether the physical stock position matches what the system records. A retailer can have strong forecast accuracy and still lose sales to poor stock accuracy, which is why the two are tracked separately rather than assumed to move together.

How often should forecast accuracy be reviewed?

Weekly or period-level review catches drift while it is still cheap to fix. Waiting for a quarterly review is usually how a small bias problem turns into a markdown problem.

‍