How Sthenia builds and checks a meal plan

Sthenia turns your training profile and goal into estimated targets, then into meals, quantities and a grocery list with Pro. The method combines calculation rules, food composition references and AI-assisted meal generation. Understanding these three steps helps you distinguish a documented recommendation from a consistent calculation or an actual measurement of your needs. This page describes how the product works, as checked on 23 October 2026, and its limitations.

Your profile is used to estimate targets

The engine takes into account your weight, height, age, the profile information you enter, your sport, training frequency and goal. It estimates energy expenditure, then calorie and macronutrient targets. These results are calculated starting points, not individual measurements of metabolism.

The default formula is Mifflin–St Jeor. The ten Haaf–Weijs calculation branch is selected for endurance sports when training at least five times a week. Selecting this branch is a product rule; it does not prove that the formula is the best choice for every person.

References: [1]

What the scientific references establish

The equation by Mifflin and colleagues was published in 1990 using data from healthy adults. The 2014 study by ten Haaf and Weijs concerns recreational athletes aged 18–35. The populations covered by these studies must be distinguished from the full range of people using an app.

The engine’s recommendation matrix cites Morton 2018, Helms 2014 and Thomas 2016, among others, for its nutrition recommendations. These references document methodological choices; they do not clinically validate Sthenia. The engine should not be presented as automatically incorporating all the latest studies. An article’s recent publication date does not prove that it has been implemented in the product.

References: [1][2][3]

Ingredients provide a basis for calculating macros

The generation process proposes meals and their ingredients. Where matches are available, quantities are linked to food composition values to recalculate protein, carbohydrates and fat, then round the results to the nearest gram.

At the time of verification, the 111 reference entries used in this calculation included 98 CIQUAL matches with a code, 11 label-derived values and 2 specific values. This audit does not establish which version of the CIQUAL table was imported: we do not claim here to use its latest edition. An identified source and an appropriate quantity are necessary to interpret the result.

References: [1]

When ingredient matching is incomplete

A partial match lowers the confidence level. Unmatched ingredients do not contribute to the recalculated total, which may lead to an underestimate. If none of a meal’s ingredients can be matched, the engine may retain the AI estimate with a low confidence level.

Conversions from individual items use average weights; conversions from millilitres to grams rely on an approximation. Check the ingredients, whether they are raw or cooked, quantities, oils and the actual products used. A calculated value with more decimal places would not be a more accurate measurement.

References: [1]

Planning, manual logging and photos: three different uses

A plan’s calculations start with proposed ingredients and quantities. Manual logging records what you actually consumed. A photo provides an estimate: it does not measure portions exactly. These workflows should not be treated as interchangeable.

For careful tracking, check foods and quantities before logging them. Manual calorie and macro tracking remains free and unlimited; the personalised seven-day plan with a budget and grocery list is part of Pro.

References: [1]

What was checked and what remains to be measured

On 23 October 2026, eight internal offline checks were run on the deployed calculation functions. They covered energy calculation branches, quantities, confidence levels and budget limits, among other aspects. They passed within that scope.

These checks are not a clinical trial, a representative benchmark of all plans or a validation of photo estimates. To assess a particular plan, check its meals and quantities. To understand the amounts, read how Sthenia estimates your grocery budget.

References: [1]

Frequently asked

Are the suggested calories a measurement of my metabolism?

No. They are estimates based on your profile and calculation rules. They do not replace an individual measurement of metabolism.

References: [1]

Does Sthenia use only AI-generated figures?

No. Meal macros are recalculated from quantities and available reference data. However, incomplete coverage lowers confidence; if no ingredients can be matched, an AI estimate may be retained.

References: [1]

Does a bibliography prove that every plan is accurate?

No. References inform the recommendations; food composition data and calculations determine the plan’s values. The internal checks described here do not demonstrate universal accuracy or clinical validation of Sthenia.

References: [1]

Does a plate scan measure my portions exactly?

No. A photo provides an estimate. Plan generation and manual logging are separate workflows; check the foods and quantities.

References: [1]

Sources

Also worth reading