Cocktail Sugar and Calories Study (Under Revalidation)
This study's result is withheld while it is revalidated (since 2026-10-17).
Most recipes in the original cohort come from a curated dataset whose recorded quantities did not match their original sources; calories per standard drink depends on those quantities. The original statistical result, charts, examples and conclusions are not shown until the cohort has been rebuilt from source-verified recipes. The methodology and limitations below describe how the original analysis was done.
Published August 26, 2026 · Observational data study · Status: under revalidation
Methodology
This is an observational, cross-sectional analysis of recipe reference data — not a controlled experiment.
- Source: WhatDrink's governed public detail view (
vw_public_prepared_drink_detail_v1). - Eligibility gate: collection = cocktails, published candidate, not needing review, not adult-flagged, nutrition block shown, nutrition quality “good” or “partial,” and 100% of recipe quantities parsed.
- Population: the cocktails that meet the gate; the primary analysis uses those with at least 0.5 U.S. standard drinks, because a near-zero denominator inflates the ratio. The cohort is being rebuilt from source-verified recipes.
- Metric: calories per U.S. standard drink = estimated calories ÷ estimated U.S. standard drinks (14 g pure alcohol each). Both are governed WhatDrink fields.
- Statistics: Pearson (linear) and Spearman (rank) correlation. Spearman is the primary statistic because it is robust to non-linearity and outliers; ties are handled with averaged ranks.
- Sensitivity: the analysis is recomputed at 0.25 / 0.50 / 0.75 standard-drink floors.
- Same-basis check: the association was also computed with sugar and calories on the same denominator — per serving, and per standard drink — so it does not rely on the serving-vs-standard-drink pairing (see Robustness checks).
- Missing data: a missing value is never treated as zero — recipes lacking a required figure are excluded and counted, not defaulted.
- De-duplication: recipe duplicates are identified by their normalised canonical recipe-item payload — not by nutrition figures, which distinct recipes can share — and the correlation is reported over both the full and de-duplicated population (a robustness check; the official population is unchanged). Duplicate names are collapsed in the presentation tables.
The exact figures depend on the governed nutrition estimates, which are documented in how WhatDrink calculates cocktail nutrition and strength. The analysis is reproducible from the governed database (it never re-crawls public pages).
Dataset limitations
These are estimates, not measurements. Calories, sugar and ABV are estimated from standard recipe quantities and ingredient reference data. Actual values depend on brand, pour size, preparation and ice dilution, which is not modelled.
Correlation is not causation. This is an observational analysis. Sugar is one of several calorie sources, and the relationship shown is statistical association only.
Not health guidance. Nothing here indicates that any drink is healthy, unhealthy, or suitable for any diet. Do not use these figures for medical, dietary, or allergy decisions. For responsible-drinking guidance, refer to your national guidelines.
Coverage. The 999-cocktail population reflects recipes that pass WhatDrink's data-quality gate at the analysis date; it is not the entire universe of cocktails.
Common questions
Why report calories per standard drink instead of calories per cocktail?
Total calories mix together two different things — how sweet a drink is and how much alcohol it delivers. Normalising by U.S. standard drink (14 g of pure alcohol) puts every recipe on the same alcohol footing, so the comparison reflects calorie load per unit of alcohol rather than serve size. It is a fairer basis for comparing very different drinks.
Explore the data
Questions or corrections? Contact us.
