How companies should measure and set targets for the healthiness of their products
www.nesta.org.uk/blog/how-companies-should-measure-and-set-targets-for-the-healthiness-of-their-products/
www.nesta.org.uk/blog/how-companies-should-measure-and-set-targets-for-the-healthiness-of-their-products/
In recent years, some food retailers and manufacturers have begun reporting how healthy the products they sell are and setting health targets to improve the healthiness of the products they sell. However, only recently has meaningful attention been given to how companies should report and set targets on health.
In this blog, we set out why measures of the healthiness of a company’s sales should be:
And how both of these approaches are better for health outcomes and better for business than binary measures (whilst acknowledging the benefits of binary measures for communication).
In this blog, we refer to health measures as the statistics that companies report which quantify how healthy the food they sell is. These measures serve two key purposes: reporting current product healthiness and setting targets for improvement.
Binary health measures, eg the UK’s ‘high in fat, salt and sugar’ threshold, report the percentage of a company's sales that meet a defined "healthier" threshold. These measures are based on nutrient profiling model (NPM) scores, which evaluate a product's nutritional composition (N.B. all recommendations we make assume that nutrient profiling models used are government-endorsed rather than company-own versions). Products scoring above a certain cutoff are categorized as "healthier," while those below are considered "less healthy". [1]
A continuous health measure uses the same underlying NPM score range without using a cut-off to distinguish between healthier and less healthy products. Therefore, rather than categorising products as simply healthier or less healthy, an average of NPM scores is calculated across all products, which importantly should be sales-weighted, to best demonstrate impact on health. [2]
To illustrate how binary and continuous health measures are calculated, consider a company selling two products:
Product: Pasta
Product: Pizza
In this example, if a product scores higher than 62 it is considered “healthier” and if it scores less it is considered “less healthy”.
Under a binary measure, we calculate the proportion of sales from healthier products where the sales are measured in kilograms:
Healthier sales proportion = Healthier sales / total sales
= 50 / (50 + 100)
= 33%
The company would report that 33% of their sales are healthier.
The weighted average calculation gives more influence to products that make up a larger percentage of total sales. Here's how it’s calculated:
So, under a continuous measure the company reports that its sales-weighted average NPM score is 60.
This is similar to how a class grade might be calculated - if your final exam is worth 70% of your grade and your homework is worth 30%, you would:
A key takeaway here is that continuous health measures do not require any more effort than binary measures. Both continuous and binary measures use the exact same input data: product NPM scores and sales figures.
Many issues with binary health measures stem from a key limitation: they only incentivise changes in a narrow subset of products. Figure 1 below illustrates this point.
HFSS = High fat, salt, and sugar. Figure 1 represents the distribution of sales from GB retailers using a converted NPM score. It is taken from Nesta’s policy proposal for mandatory health targets for retailers (Source: Nesta analysis of Kantar’s Worldpanel 2021 Take Home Data).
Under a binary measure, there is no incentive to change products already classified as “healthy” (the red shaded area on the right side of Figure 1). For example, under the UK NPM, both Coco Pops and apples are classified as healthy. Consequently, making Coco Pops healthier – or switching sales from Coco Pops to apples – would not improve parent company Kellanova’s binary health score, which is clearly not a desirable outcome.
There is also little incentive to reformulate products which are extremely unhealthy. These are the products which are worst for our health and are shown in the red shaded part on the left of Figure 1. Reformulating extremely unhealthy products to cross the threshold would improve a company’s binary health score. However, this extreme level of reformulation will be expensive, could change the nature of the product, and will often not be technically possible. This issue also applies to shifting sales. It is much harder for companies to find close substitutes for products that are extremely unhealthy to ones classified as “healthier”.
This leaves companies with only a very narrow subset of products for which they are incentivised to reformulate or shift the sales of - those products that are just below the healthy threshold (the green shaded part of Figure 1). These are the products that stand a chance of making it across the binary threshold.
A continuous health measure incentivises improvements across a company's entire product range. This allows companies to achieve a given level of health impact by making smaller, more feasible changes to a wider range of products. In contrast, a binary measure restricts companies to improving only products near the "healthier" threshold, forcing them to make more extreme and unrealistic changes to this narrower set of products to achieve an equivalent health impact.
To illustrate this, we’ve approximated how much reformulation could be needed under both binary and continuous measures to reduce daily calorie intake by the same amount, across an indicative portfolio of Great Britain (GB) supermarket sales. [3]
Figure 2 shows:
Figure 2: Binary targets require extreme and unrealistic changes to achieve the same impact as continuous targets
The key takeaway is that under a binary target, fewer products must change by a larger amount, which will be extremely difficult to achieve. Under a continuous target, smaller improvements spread across more products can deliver the same overall impact. For example, in one scenario, to achieve an equivalent level of kcal reduction the following levels of reformulation were needed:
A company can improve its percentage of "healthy" products without actually making their overall sales healthier. This seemingly contradictory situation reveals a key limitation of using binary health measures alone.
The graph below shows how binary scores (the proportion of "healthy" vs "unhealthy" products) can improve while the actual average health score remains the same or gets worse.
A binary measure only shows what proportion of product sales cross the line between "healthy" and "unhealthy" classifications (the green arrow). Meanwhile, healthy products might become less healthy without crossing into unhealthy territory (the purple arrow), and unhealthy products might become even less healthy (the red arrow). This creates a situation where the binary measure shows improvement, but the continuous health score tells a different story.
It is more cost-effective for a company to achieve the same level of health impact using a continuous measure rather than a binary one. Because continuous measures encourage reformulation across the entire product range, companies can concentrate on improving the products that are least expensive to change.
The potential costs of improving the healthiness of sales will differ between products and categories. Some are cheaper to reformulate, some have more readily available healthy alternatives, and new product development is more economical in certain categories. By incentivizing change across the whole portfolio, continuous measures enable businesses to pursue the most financially efficient strategies for enhancing the healthiness of their sales; by contrast, binary measures again incentivise companies to reformulate a narrow range of products which may not be cost-effective.
Multinational companies face various health related regulations across different countries. Companies will often respond to these regulations and improve the health of products in order to, for example, pay less sugar tax or avoid black warning labels on their products.
However, when a company uses a binary measure to report on health at a global level, many of these product improvements go unrecognized. With a binary measure, changes only count if a product’s health crosses a defined threshold. Because many regulatory driven health improvements will not push a product across that line, they have no impact on the company’s global health measure.
In contrast, a continuous measure recognises every improvement in healthiness – no matter the product’s initial healthiness or whether it crosses a specific threshold – ensuring any regulation driven health improvements are captured in a company’s continuous global health measure.
Whether using a binary or a continuous measure, companies must choose whether to weight this measure by sales, and if so, how. Three options include:
We don’t recommend an unweighted approach as this does not capture companies sales and therefore will not give an accurate view of health impacts.
We don’t recommend weighting by sales value (eg, pounds) because it creates a perverse incentive to lower the price of less healthy products and increase the price of healthier ones. For example, if a company halves the price of its least healthy product and sales of that product consequently double, it would still count the same towards the company’s health score. This is because the overall sales value remains unchanged, despite the company selling twice as much of this unhealthy product.
Under a kilogram-weighted scenario, however, the product would now count twice as much in the company’s score, reflecting the increased sales volume of less healthy products. Therefore, our recommendation is to weight by sales in kilograms in order to maximise health impacts. This method of weighting captures the sales of a company without creating perverse incentives on price.
The case for the use of continuous measures is clear. However, one advantage of binary measures is their simple, intuitive communication value. Using our example above, saying that "33% of our sales come from healthy products" is more immediately understandable than saying "our sales-weighted NPM score is 60." This is why we recommend that when it comes to reporting, companies could report a binary measure of health in addition to a continuous measure of health.
Continuous health measures give a clearer picture of companies' progress on health, incentivise them in the right direction and help investors understand the full picture.
We recommend that companies:
To maximise health impacts all these measures should be weighted by kilograms.
Nesta is able to support companies which operate in the UK with measuring and improving the healthiness of their sales using continuous health measures. We are currently working with Asda on doing just this. If you have any questions or are interested in working with us please reach out at [email protected].
In short, yes. The longer answer is that the arguments set out here apply to any nutrient profile model that first calculates a continuous score and is then used to split products into “healthy” and “less healthy” categories. So the benefits of continuous measures could include commonly used NPMs such as the UK NPM, Healthy Star Rating, and Nutri-Score. [1][2]
Yes. The principals set-out here apply to single-nutrient measures. For example, it would be better for organisations to report the sales-weighted average salt per 100g they sell than reporting the percentage of their sales that are “low” or “high” in salt.
No. If a business is already reporting a binary health measure, it will only take a couple of lines of data code to calculate a continuous health measure. The most resource intensive part of the reporting process is calculating NPM scores for each product which is required both for binary and continuous health measures.
Yes. Whilst it is early days for continuous health measures, Asda have committed to using continuous health targets for the same reasons set out in this blog (see page 15 in their ESG report).
At Nesta we are supporting Asda in improving the health of their sales as measured by a continuous sales-weighted average NPM score. We are happy to discuss how to do this with manufacturers, retailers, the out-of-home sector, investors, and policy makers. Please get in touch at [email protected].
All contents and statements within this publication are Nesta’s own findings based on their independent analysis.
[1] We note there are nuances relating to how binary measures apply for Nutri-Score and Health Star Rating nutrient profiling models eg there is no consensus on where a binary cutoff of ‘healthier’ vs ‘less healthy’ is drawn for Nutri-Score and Health Star Rating (whereas this is set in legislation for the UK NPM as the ‘high in fat, salt and sugar’ threshold).
[2] Technically speaking, the UK NPM is considered a discrete or semi-continuous measure (as it only takes whole-number values within a large range), whereas the Health Star Rating and Nutri-Score are discrete ordinal measures with fixed increments (e.g., 0.5 steps for HSR). However, when averaged across multiple products, these measures become continuous, as they can take any decimal value within a range. Therefore the principles outlined in this blog regarding the benefits of continuous measures apply to all such averaged scores, so we do not draw a distinction.
[3] Nesta’s analysis of Kantar’s Worldpanel 2021 Take Home data.
