# Size Confidence Scores: Making Fit Recommendations Trustworthy | NimbleTailor

Source: https://nimbletailor.com/blog/size-confidence-scores-fit-recommendations-2026-10-15/

NimbleTailor answers


A size confidence score tells a shopper how much to trust a recommendation, not just which size to pick. High confidence comes from a close match between the shopper measurements, past purchases, and how a specific garment actually runs. Showing the score, and the reason behind it, helps shoppers decide when to trust the pick and when to check the measurements themselves.

## What the score should be built from

Start with what the shopper kept, not what they ordered. A size that was bought and returned for fit is a negative signal even if it was the most popular size for that item. Layer in garment specific data: this style runs small at the waist, that fabric has no stretch, this cut suits a longer rise. Generic size charts alone produce confident sounding guesses.

For more on this, see [this related guide](https://nimbletailor.com/blog/fit-profiles-returning-shoppers-2026-10-15/) and [our earlier explainer on the same theme](https://nimbletailor.com/blog/size-aware-personalization-right-fit-first-2026-10-14/).

Brand level history matters less than category history. A shopper size in knit tops says little about their size in structured trousers, so confidence should drop when the recommendation crosses categories, and the interface should say why instead of presenting the same certainty everywhere.

## When to show confidence and when to hide it

Show the score when it changes behavior: high confidence supports a single size pick and discourages bracketing, low confidence is the honest moment to suggest checking measurements or choosing between two close sizes. A score displayed identically on every product becomes decoration shoppers ignore within one session.

Never let a low score look like a failure. Frame it as information: limited data for this style, or your history sits between sizes. Shoppers who feel warned make better choices than shoppers who feel judged, and they come back after a return that was their informed call.

## Calibrating the score so it means something

A confidence score is calibrated when, among recommendations scored at ninety percent, about nine in ten shoppers keep the size. Track that by score band, category, and new versus returning shoppers, and retrain or rescale when a band drifts. Uncalibrated scores are worse than none because they borrow trust they have not earned.

Readers also found [this companion piece from our archive](https://nimbletailor.com/blog/weather-aware-personalization-local-forecast-2026-10-13/) useful.

Watch returns by score band as the ground truth. If high confidence recommendations return at the same rate as low ones, the score is measuring the wrong thing, often purchase volume rather than fit evidence. Fix the inputs before touching the display.

## Using confidence to cut bracketing honestly

Bracketing, buying two sizes to return one, is rational when a shopper cannot tell which size fits. A credible high confidence score, backed by a free exchange promise if the pick is wrong, gives them a reason to order one. The exchange promise matters: it puts the cost of a model mistake on the store, where it belongs.

Report the effect carefully. Measure single size orders, fit related returns, and exchange rate together, because pushing shoppers to one size while exchanges quietly spike is not a win. The goal is fewer parcels and fewer returns at the same time, not a prettier returns chart.

## Frequently asked questions

### What is a good confidence threshold for a single size pick?

Many stores start by steering single size picks only above a band where historical keep rate is at least eighty five percent. The right threshold is the one your own returns data supports, reviewed by category rather than set once for the whole catalog.

### Should shoppers see a percentage or a label?

Labels like high, medium, and low, paired with a short reason, tend to work better than a precise percentage that implies false accuracy. Whatever format you choose, keep it consistent so shoppers learn what each level has meant for them.

### Does showing low confidence hurt conversion?

It can lower conversion on that item while raising kept sales and repeat purchase. Judge it on kept revenue and returns cost over thirty days, not on checkout clicks alone, or you will optimize for orders that come back.

### How much history does a shopper need for a high score?

Usually a few kept purchases in the same category, plus measurements or a fit profile, is enough. One purchase should rarely produce top confidence, because a single kept item can reflect tolerance rather than a true fit match.