We analysed multiple commonly used healthiness measures as part of our assessment.

Based on our analysis, we recommend the introduction of an average NPM score-based target.

This measure has the optimal balance between impact and feasibility of implementation, as it is a more holistic measure of the health of food and is already established in legislation (see technical appendix for a full appraisal of health metrics). Retailers are already required to calculate NPM scores for many of their products to comply with existing HFSS legislation. This target would be applied across a retailer’s entire food product portfolio (for branded and own brand) and sales weighted (see Box 1) to ensure that products that have a higher volume (in kg) of sales contribute more to average scores than those that are purchased less frequently and in smaller volumes. 

While we are recommending an average NPM score, we believe that a calorie density measure would also work. A calorie density-based target provides a direct route to tackling obesity by incentivising a reduction in calories sold and modelling shows it can achieve the same impact as the NPM-based target. However, it is an unfamiliar metric to industry. It only considers improvements in a single element of food composition, unlike the NPM score, which captures a more holistic view of ‘healthiness’ and is the basis of existing legislation. It should be noted that if the targets were to be extended to the out-of-home sector, calorie density may be a more viable metric as large businesses are already required to calculate calorie information for their meals to comply with calorie labelling legislation. 

We ruled out implementing an HFSS-based target. A target aiming to reduce the proportion of a retailer’s HFSS sales by applying a binary classification to products as either HFSS or non-HFSS limits its effectiveness, as it only incentivises improvements in products near the HFSS classification boundary and not in the most unhealthy products. Our modelling demonstrates that retailers would have to enact more extreme sales or reformulation shifts to meet such a target, potentially making the policy either too difficult to implement or not sufficiently impactful to justify its adoption. See the technical appendix for a detailed appraisal of the healthiness metrics.

[7] We transformed raw NPM scores for the converted NPM target with a commonly used formula developed by the University of Oxford, which involves multiplying the raw NPM score by -2 and adding 70. Using this formula, a raw NPM score of 4 is equal to a converted NPM score of 62 (the threshold for a low converted NPM score, HFSS and ‘unhealthy’ classification). We have referred to this scaled NPM score as a ‘converted NPM score’ (see technical appendix for more details).