Comparing recipe nutrient calculations across nutrition analysis software for workplace menu planning

Nutritional analysis plays a critical role in understanding the nutrient content of foods and recipes. While chemical analysis in accredited laboratories is considered the gold standard for accuracy (Church, 2015), it is impractical for routine use because of the vast variety of foods and associated costs (Machackova et al., 2018). Nutritional analysis software has emerged as a practical and widely accepted alternative, enabling rapid nutrient calculations based on food composition databases (FCDs), ingredient properties, cooking adjustments and portion sizes. These tools are pivotal across sectors such as nutrition research, dietetics, food manufacturing and public health, supporting tasks like menu labelling, therapeutic diet planning and regulatory compliance (Stumbo, 2008; Church, 2009; Brinkley et al., 2025).

In the European Union, Regulation Number 1169/2011 mandates nutrient declarations for prepackaged foods and permits calculation-based methods for most products (European Union, 2011; Traka et al., 2020). However, these methods depend heavily on the quality of underlying FCDs and the accuracy of the software’s algorithms. For example, differences in data sources, yield factors and nutrient retention calculations can introduce variability between software tools. This variability underscores the importance of using software tailored to the geographical context, as nutrient profiles in FCDs can differ significantly between countries.

Despite advancements in software technology, challenges persist in managing FCDs. The UK’s McCance and Widdowson data set, for example, requires frequent updates to accommodate changes in agricultural practices, processing methods and new research. However, financial and institutional constraints often limit the scope of these updates, resulting in data inconsistencies. Traka et al. (2020). highlight the difficulty of maintaining reliable data sets under such conditions, where limited funding and a lack of access to innovative tools contribute to data obsolescence and variability, compromising overall data quality and consistency.

Efforts like those by the European Food Information Resource (EuroFIR) aim to harmonise recipe calculation standards and improve data comparability across Europe (Reinivuo et al., 2009). A key focus has been on recipe calculation procedures, with the most commonly used method being the application of yield and nutrient retention factors, which EuroFIR now recommends (Reinivuo et al., 2009).

A growing number of nutritional analysis software tools are now available, including Nutritics™, MyFood24 and NutritionistPro, each offering varying levels of functionality and database integration. These tools provide essential features like recipe analysis, compliance with labelling regulations and identification of allergens (Kapsokefalou et al., 2019). With numerous options available, choosing the right software can be challenging, as the functionality, features and data integration vary significantly across programs. However, differences in their underlying algorithms and data sources can lead to inconsistencies in nutrient calculations (Kapsokefalou et al., 2019; , Geaney et al., 2013a). For example, discrepancies in macronutrient and micronutrient estimates between software can impact decision-making in dietetic practice and food industry applications. Additionally, compatibility with geographic-specific food composition data, such as national food databases, is crucial for ensuring the accuracy and relevance of the analysis (Brinkley et al., 2025). The software should also offer additional features like meal planning, recipe analysis and reporting tools that align with the dietitian’s goals, whether for individual counselling, food service management or public health initiatives.

In this context, custom in-house solutions, such as those used by food choice at work (Cork, Ireland) (FCW), may offer distinct advantages. FCW, a scientifically validated workplace well-being company, uses an evidence-based approach to promote employee health and well-being, underpinned by over six years of research and 23 peer-reviewed publications ( Geaney et al., 2013a ,Geaney et al., 2013b; Geaney et al., 2016; ). Tailored recipes are central to their service, with the company using an in-house nutrition analysis software to analyse menus, recipes and food items. While in-house systems can be tailored to meet specific organisational needs, the nutrient outputs generated by FCW’s software have not yet been independently compared with established commercial tools. Given its use in real-world workplace catering, demonstrating good agreement with widely used nutrition analysis software is important for building confidence in its professional application.

Therefore, this study aimed to compare FCW against Nutritics™ and MyFood24 to assess its performance in calculating the nutrient content of complete recipes. Specifically, this study examined differences in macronutrient and micronutrient estimates across the three platforms for workplace menu planning and recipe-based dietary analysis.

The FCW recipe database is an internally developed repository of standardised recipes used by FCW for workplace menu planning, nutritional analysis and menu labelling in commercial and institutional catering settings. Recipes are developed by registered nutritionists and dietitians and are designed to be prepared at scale for workplace foodservice, with clearly defined ingredients, quantities, cooking methods, yields and portion sizes. Food composition data are primarily sourced from the UK McCance and Widdowson Composition of Foods Integrated Data set (Seventh Edition) (Finglas et al., 2014), supplemented where necessary with manufacturer-provided nutrient data and internally standardised pre-analysed sub-recipes. The software calculates energy, macronutrients, water, vitamins and minerals per 100 g of each ingredient, incorporating factors for edible portions, cooking methods and nutrient retention based on the Bognár Tables (Bognár, 2002). Yield factors and nutrient retention factors are applied in accordance with EuroFIR recommendations, with yield factors applied at the recipe or cooking-stage level and nutrient retention factors applied at the ingredient level based on the specified cooking method. Where nutrient values for a given micronutrient are unavailable in the source database, the software records these as missing rather than imputing values. Specific handling is applied for salt and fat absorption during cooking processes. Ingredients are flagged for allergen content according to EU regulations, and dietary suitability (e.g. vegetarian and vegan) is also determined.

Final recipe nutrient values are adjusted for cooking yield and are calculated per 100 g and per portion. Energy and salt values are determined following EU Regulation 1169/2011 (European Union, 2011), using macronutrient concentrations and sodium content, respectively.

A total of 12 recipes were selected from the FCW recipe database with three recipes selected across key meal categories commonly used in workplace catering (breakfast, lunch, dinner and snacks) to reflect typical menu planning scenarios rather than to represent the full spectrum of home-prepared foods. At the time of analysis, the FCW database comprised several hundred recipes routinely used across multiple workplace catering sites in Ireland and the UK.

Each recipe included detailed information on ingredients, quantities, cooking method, cooking time and the number of servings. The use of a limited number of recipes (n = 12) was intentional, allowing for controlled, standardised comparisons across platforms using identical inputs. Although this does not reflect the full variety of recipes used in practice, it does support a more detailed examination of how methodological differences may influence nutrient calculation.

Nutritional analysis was performed for each recipe, using identical ingredient inputs across all software platforms to enable standardised comparisons. Macronutrients assessed included energy (kcal), protein, carbohydrates (total, sugars and fibre), fat (total and saturated) and salt. Micronutrients analysed included fat-soluble vitamins (A, D and E), water-soluble vitamins (B1, B2, B6, B9, B12 and C) and minerals (calcium, iron, potassium, magnesium and phosphorus).

To evaluate the performance of FCW, two widely recognised nutritional analysis software tools, Nutritics™ and MyFood24, were selected for comparison. Both platforms use the UK McCance and Widdowson Composition of Foods Integrated Data set as their primary food composition source, which provides nutrient profiles for several thousand generic and branded food items and is widely used in nutritional research across the UK and Ireland. Using software drawing from the same core database allowed for a controlled comparison of methodological differences, particularly in the application of yield, nutrient retention and edible portion factors.

Nutritics™ was selected because of its robust analytical capabilities and widespread use in the Irish and UK markets. A key strength of Nutritics™ is its application of yield, nutrient retention and edible portion factors, enabling detailed and accurate recipe analysis. The software applies EuroFIR-aligned procedures to account for weight changes and nutrient losses during cooking, supporting its use as a benchmark tool in this study.

Identifying a second suitable software proved challenging because of variability in key features needed for standardised recipe analysis, such as yield factors, nutrient retention factors and edible portion adjustments. Free trials of several tools, including Nutritionist Pro™, Dietplan7 and Nutricalc, were conducted to assess their features. Both Nutritionist Pro™ and Dietplan7 lacked the ability to apply all three critical adjustment factors – yield, nutrient retention and edible portions – rendering them unsuitable for the study.

MyFood24 was ultimately selected as the second software. While it adjusts for weight changes during cooking, it does not explicitly apply generic nutrient retention factors (e.g. EuroFIR). MyFood24 is supported by extensive development research, is widely used in dietary studies and, similar to Nutritics™, bases its nutrient calculations primarily on the UK McCance and Widdowson data set (Carter et al., 2015; ).

The 12 recipes were initially analysed using Nutritics™ (Dublin, Ireland, V5.74). Each recipe was given a title, and the ingredients were inputted as raw items along with their respective quantities (grams or millilitres) and cooking methods. The number of servings per recipe was also entered, and the software automatically calculated the recipe yield. Weight changes were accounted for using yield factors integrated into the Nutritics™ software.

These yield factors were applied at the recipe level, following the procedure for recipe calculation outlined by EuroFIR (Reinivuo et al., 2009). Nutrient retention factors were applied automatically based on the cooking method selected for each ingredient or sub-recipe. For composite dishes involving multiple cooking methods, the recipe was broken down into sub-recipes, each corresponding to a distinct cooking method. For example, the steak and mushroom pie was deconstructed into four sub-recipes based on different cooking methods, which were then combined into one final recipe. The nutritional analysis for each recipe was exported as a PDF file for review.

The 12 recipes were inputted into MyFood24 (University of Leeds, UK). First, a project was created, and “test” was entered into each required field. McCance and Widdowson Composition of Foods Seventh Edition ( Finglas et al., 2014) and UK branded foods database were selected. A new food diary was created for each recipe by accessing the food diary via a unique URL link. Under the “new recipe” tab, each recipe was given a title, and the ingredients were searched from the selected databases and entered as raw ingredients.

The quantity (grams or millilitres) for each ingredient was specified, followed by the number of servings per recipe. The yield factor was automatically applied by the software for selected raw ingredients (food items typically not consumed raw). After entering all necessary details, the recipe was saved and added to the food diary under the appropriate mealtime (e.g. breakfast, lunch, dinner or snack). The software then calculated the weight per serving (grams). Finally, the food diary was submitted, and the nutritional analysis was exported into a Microsoft Excel spreadsheet.

The same 12 recipes were inputted into the FCW nutrition analysis software (Cork, Ireland, V9.0). Each recipe was given a title and classified as either a food item or a drink. The ingredients were entered as raw items, with their respective quantities, measurements (grams/kilograms or millilitres/litres) and cooking methods specified at the ingredient level.

Weight changes were calculated using yield factors integrated into the FCW software, which were applied at the recipe level, in accordance with EuroFIR procedures (Reinivuo et al., 2009). For composite dishes, ingredients were grouped according to their cooking methods, and a yield factor was assigned to each group. The number of servings was entered, and the software automatically calculated the total cooked weight (grams) and weight per serving (grams). The nutritional analysis for each recipe was then exported as a PDF file.

Data was exported from each analysis software and manually recorded in Microsoft Excel (Microsoft Corporation, Redmond, WA, the USA). Statistical analyses were performed using SPSS Statistics, Version 28.0 (IBM Corp., Armonk, NY, the USA). Normality of the data was assessed through skewness and kurtosis. Nutrients unavailable in each database were recorded as missing and were not imputed. Although we did not calculate the proportion of missing values for each nutrient, these missing data are acknowledged as a potential contributor to between-software variability, particularly for micronutrients. For each recipe, the mean percentage difference, minimum difference and maximum difference were calculated. Agreement between the software outputs was evaluated using Bland–Altman plots, in which the differences between paired measurements were plotted against their means. The 95% limits of agreement were calculated as the mean difference ± 1.96 standard deviations (SDs). Pearson’s correlation coefficients were also computed to further assess the strength of the association between methods. Statistical significance was set at p ≤ 0.05.

Between April and June 2022, 12 recipes from the FCW website were analysed using both MyFood24 and Nutritics™. These recipes were selected to represent a variety of meal types, including breakfast (n = 3), lunch (n = 3), dinner (n = 3) and snacks (n = 3) (Table 1).

Overall, differences between the nutritional analysis software were generally small across most recipes, although some notable variations were observed. For example, in the overnight oats recipe, MyFood24 calculated a higher sugar content (21 g) compared to Nutritics™ (18 g) and FCW (18 g). While MyFood24 recorded the lowest average sugar values across all recipes, the differences between software were relatively minor, typically within 1.0 g on average, and appeared to be more dependent on the specific recipe rather than the software itself. For salt content, FCW reported the lowest value for the baked potato recipe (0.12 g). Across most recipes, salt content was consistent between the software types, with only small variations. However, discrepancies were evident in some cases; for instance, in the stew recipe, Nutritics™ and FCW reported similar salt values (0.23–0.24 g), whereas MyFood24 calculated a substantially higher salt content (2.4 g).

For each recipe, the mean percentage difference, along with the minimum and maximum differences, was calculated across the tools to assess variability (Table 2).

Differences were detected for fat, saturated fat, sugars, fibre and salt across at least one pairwise comparison (p ≤ 0.05). Energy content exhibited the greatest variability, with the largest range observed between FCW and Nutritics™ (−3.2–101 kcal), followed by Nutritics™ and MyFood24 (−59–23 kcal).

Nutritional profiles from FCW, MyFood24 and Nutritics™ were compared to quantify differences in nutrient estimates (Table 3). Energy estimates were similar, with FCW reporting the highest mean energy (289 kcal), 4.4% higher than MyFood24 (277 kcal) and 8.5% higher than Nutritics™ (266 kcal). Protein content was consistent, with FCW reporting the highest mean (19 g), 5% more than MyFood24 (18 g) and 2.9% more than Nutritics™ (19 g). FCW also reported lower salt content (0.46 g), 36.4% lower than MyFood24 (0.73 g) and 5.9% lower than Nutritics™ (0.49 g). Sugars and fibre estimates showed moderate variability, with differences in sugar content ranging from −2.6 to 5.6 g between Nutritics™ and MyFood24 and fibre differences ranging from −1.3 to 5.7 g between FCW and Nutritics™.

Only micronutrients showing differences greater than 10% between software platforms are presented in Table 4. Data for other micronutrients and minerals (e.g. vitamins B1, B2, B12, C, calcium, magnesium and phosphorus) are not shown for brevity, as differences were small (<10%) and not of practical or statistical relevance. For vitamin A, differences were observed between FCW and Nutritics™ (−145–0.25 µg, p ≤ 0.05) and between FCW and MyFood24 (−179–13 µg, p ≤ 0.05) (Table 4). Vitamin D values ranged from −0.37 to 0.02 µg (p < 0.001) between FCW and Nutritics™ (Table 4). Vitamin E estimates varied, particularly between FCW and MyFood24 (−3.2–0.06 mg, p ≤ 0.05). Differences were also observed for vitamin B6 (−0.92–0 mg, p < 0.001). In contrast, no significant differences were found for vitamins B1, B2 or Folate.

Agreement between the software tools for macronutrient estimates was assessed using the Bland–Altman method (Table 5). Energy (calories) were excluded from Bland–Altman analysis because of a significant difference between FCW and Nutritics™ (p ≤ 0.05), indicating potential proportional bias that violates the assumptions of the method; macronutrients were analysed independently, as per standard practice. The greatest mean differences between FCW and MyFood24 were 0.91 g for protein (95% CI: −3.5, 5.3), −0.87 g for carbohydrate (95% CI: −10, 8) and −1.03 g for fat (95% CI: −5, 2.5).

This study found considerable variation in both macronutrient and micronutrient content across the 12 recipes analysed using three different nutrition software platforms: FCW, Nutritics™ and MyFood24. Among the macronutrients, calorie content showed the greatest statistical variability across platforms, with the largest mean difference observed between FCW and Nutritics™ (−23 kcal), followed by a difference of 12 kcal between FCW and MyFood24. While macronutrient variation was relatively minor, much larger discrepancies were identified in the micronutrient analysis – particularly for potassium, vitamins A and B9. The largest mean difference in potassium was between FCW and MyFood24 (−140 mg), and in vitamin A, it was between FCW and MyFood24 (−47 µg), followed by Nutritics™ and FCW (24.6 µg). Notably, 10 of the 12 recipes showed more than a 20% difference in vitamin B9 content when comparing FCW and MyFood24, whereas only two recipes exceeded that threshold when comparing FCW and Nutritics™. These findings suggest that FCW is generally more consistent with Nutritics™ than with MyFood24, especially in the analysis of micronutrients, although the practical impact of these differences may vary depending on the intended use of the data. MyFood24 produced higher estimates for certain micronutrients – including potassium, vitamins A and B9 – when compared to FCW and Nutritics™. Although this comparison does not establish the accuracy of any platform, it demonstrates that depending on the software used, different nutrient estimates can be generated from identical recipes, particularly for micronutrients. This may have important implications for menu labelling, nutrition communication and the choice of software in workplace and other practice settings.

Previous studies assessing nutrition analysis software and mobile apps have generally shown good agreement in energy and macronutrient estimates, with reported differences in calorie values typically ranging between ±15 to ±37 kcal when compared to reference methods (Fallaize et al., 2019). The differences observed in this study – namely, a mean difference of −23 kcal between FCW and Nutritics™ and 12 kcal between FCW and MyFood24 – fall within this expected range. Although these differences were statistically detectable, their magnitude is unlikely to be nutritionally meaningful for individual recipes, reinforcing the existing literature that macronutrient differences between validated platforms are relatively minor in practical terms. The observed difference in calorie estimates likely stems from how each platform handles cooking-related weight changes. Both FCW and Nutritics™ use similar yield factors derived from Bognar (Bognár, 2002), yet minor discrepancies may occur based on how these factors are applied within each platform. In contrast, MyFood24 applies these adjustments automatically at the ingredient level, which deviates from EuroFIR guidelines recommending their application at the recipe level (Reinivuo et al., 2009). This difference in methodology likely contributed to some of the variation seen in calorie estimates, particularly when comparing MyFood24 to the other two platforms. MyFood24 supports the addition of home-cooked recipes but lacks functionality for creating sub-recipes or multi-layered composite dishes within a single workflow (Threapleton et al., 2022; Carter et al., 2015). This limitation restricts the ability to account for cooking-related changes at different preparation stages, reducing precision for complex recipes. In contrast, FCW and Nutritics™ allow users to build sub-recipes, apply different cooking methods and make adjustments at both ingredient and recipe levels. This feature enhances accuracy for composite dishes and aligns with best-practice guidelines for recipe analysis (British Dietetic Association, 2023). The absence of such capabilities in MyFood24 likely contributed to the greater variability observed in its nutrient estimates for complex recipes in this study.

The literature consistently reports much greater discrepancies for micronutrient values than for macronutrients both between different software platforms and compared to reference methods (Zhang et al., 2019; Fallaize et al., 2019). Factors contributing to this variability include differences in FCDs, the completeness of micronutrient data and how each platform handles recipe calculation, ingredient matching and cooking losses. Our finding that potassium, vitamins A and B9 showed the largest differences-sometimes exceeding 20% aligns with previous reports that micronutrient estimation is less reliable and more sensitive to database and methodological differences (Carter et al., 2015; Fallaize et al., 2019). The discrepancies observed in micronutrient content can be partially explained by the differing implementation of nutrient retention factors across platforms. Evidence from validation and methodological studies on MyFood24 shows that nutrient calculations are based on food composition tables, primarily the McCance and Widdowson data set, and do not mention the application of nutrient retention factors to account for cooking or preparation losses (Kapsokefalou et al., 2019,Threapleton et al., 2022). This is in contrast to both FCW and Nutritics™, which apply nutrient retention factors at the ingredient level based on the cooking method specified, in line with EuroFIR guidelines (Reinivuo et al., 2009). This distinction reflects the intended use of each platform: FCW and Nutritics™ are designed for detailed recipe-level and catering analyses, applying ingredient-level yield and retention factors, whereas MyFood24 is primarily intended for epidemiological dietary assessment, focusing on population-level intake estimates rather than precise recipe-level outputs. Differences in micronutrient estimates should, therefore, be interpreted within the context of each software’s purpose rather than as a methodological deficiency. These findings are consistent with previous research. (Ramirez-Silva et al., 2020) found that not applying nutrient retention factors resulted in significantly higher estimated intakes for micronutrients-leading to an underestimation of the prevalence of inadequate intake by 2%–55.5% in the Mexican population. Similarly, (Zhang et al., 2019) evaluated the nutrient calculation accuracy of 12 popular food diary apps by comparing their recipe outputs to a reference method to account for cooking losses. This study found that applying nutrient retention factors significantly reduced the reported micronutrient values in most cases, with decreases exceeding 45% for vitamins B6, B12 and folate, while calcium remained unaffected (Zhang et al., 2019). Recent work examining AI-generated dietary data has revealed significant differences in energy, micronutrient and protein estimates across tools, highlighting how differences in underlying databases and calculation methods can affect the accuracy and reliability of nutrition estimates (Bayram and Arslan, 2025; Bayram and Ozturkcan, 2024).

From a workplace catering perspective, accurate recipe-level nutrient calculation represents a foundational step in estimating the nutritional quality of meals consumed in busy high-volume, settings. Employees typically select combinations of items (mixed dishes) rather than individual recipes, so small recipe-level discrepancies can accumulate across full meals. Although this study examined individual recipes rather than full meal combinations, the observed variability – particularly for micronutrients and salt – highlights how software-specific calculation methods could influence population-level nutrient exposure, in large catering operations. Consistent and transparent recipe calculations are, therefore, essential before extending assessments to composite meals and menu cycles typical of workplace foodservice.

The limitations of this study are that we included only three software platforms, limiting the generalisability of findings to other tools. Furthermore, the small sample of 12 recipes may not fully capture variability in ingredients, preparation methods and complexity of recipes used in broader foodservice settings, where menu combinations, batch cooking and ingredient substitutions are common. The analysis relied solely on software-based nutrient calculations and did not include a reference method such as laboratory analysis. Differences observed may, therefore, reflect variation in underlying FCDs, how yield and nutrient retention factors are applied and the handling of missing nutrient data – rather than actual inaccuracies. Strengths of this study include consistent portion sizes which enabled controlled assessment of both macro- and micronutrient outputs, providing practical insights for nutrition and workplace catering applications while building confidence in FCW’s real-world nutrient calculations.

This study highlights the importance of understanding the differences in functionality between nutritional analysis software tools, particularly in workplace menu planning contexts. It should be noted that only a small sample of recipes were analysed; while this allowed for a controlled comparison, the small sample size severely limits the generalisability of the findings, and results should, therefore, be interpreted with caution when applied to broader menu planning scenarios.

While overall agreement between platforms was acceptable for most macronutrients, discrepancies were observed for energy, salt and several micronutrients. FCW demonstrated closer agreement with Nutritics™ than with MyFood24, especially for micronutrient estimates, likely reflecting similarities in FCDs and calculation approaches. The suitability of FCW for workplace nutrition analysis stems from its standardised methodology rather than agreement with any specific comparison tool.

This study evaluates agreement between software platforms rather than accuracy against measured reference values, and results should be interpreted within this context. These findings are particularly relevant for organisations relying on in-house systems for operational nutrition analysis, where unvalidated outputs may otherwise go unchecked. Understanding the methodological differences between software platforms is critical for interpreting nutritional outputs and informing menu planning decisions. Future research should incorporate larger and more diverse recipe data sets and assess combined menu offerings typical of workplace catering environments. Such work would further strengthen understanding of the performance and applicability of nutrition analysis software in real-world food service settings.