Serving Match Score

What we examined

Earlier drafts treated pack format and broad assumptions as evidence of a typical portion. That was not sufficient for a product rating, so those scores have been withdrawn.

What the label shows

The label serving is the quantity used to calculate nutrition information. On a US Nutrition Facts label, serving size is not dietary advice.

What it will be compared with

The comparison quantity will be the median of valid portions reported for the same product and market. Pack size alone is not a substitute for a reported portion.

Data classification: the label serving is product or third-party database data. The benchmark is user-reported PerBiscuit data. The score is a PerBiscuit calculation.

The formula

Both quantities are converted to comparable units. Let L be the label serving and M be the median reported portion.

Serving Match Score = 100 × min(L ÷ M, M ÷ L)

An exact match scores 100. A label serving that is 20% smaller or larger than the reported median scores 80. The score measures numerical agreement only; it does not judge health, label legality or company conduct.

Evidence threshold

A product becomes eligible only after PerBiscuit has at least 30 valid reports from at least 20 distinct sessions, collected across at least 30 days. Reports must identify the same product and market, use convertible units and pass duplicate and automated-submission checks.

Until every condition is met, the public status is Pending evidence threshold. A round number based on a broad assumption is not a score.

Confidence and ranking rules

PerBiscuit will calculate a 95% bootstrap confidence interval for the median portion and translate both interval limits into score limits. The report will display that interval beside the point estimate.

Products with overlapping score intervals will be shown as tied rather than placed in a precise order. PerBiscuit will not publish a leaderboard until at least ten products independently meet the evidence threshold.

Results

Interpretation

A high score would mean the label serving and reported median are numerically close in the available sample. It would not show that either quantity is advisable for an individual.

Limitations

• Reports are self-selected and may not represent all purchasers.
• A median does not describe the range of individual portions.
• Product variants, package sizes and countries must remain separate.
• Small or coordinated samples can distort a result, even after basic checks.

Why the information may be useful

When the threshold is met, the measure may help readers compare the label calculation unit with reported use. Publishing the sample and uncertainty keeps that comparison open to scrutiny.

Method and sources

The complete specification is on the methodology page. Source categories, validation status and update notes are described on the data sources page.

Last reviewed and corrections

This method was last reviewed on July 14, 2026. It requires independent statistical review before scores are released. Please send concerns through the corrections process.