🎬 Every week I share a short (fun!) product video teardown and some hot takes on behavior change. Pro tip: The video only takes 〜5 minutes at 2x speed. And it will be more memorable. ;) ⏩
While in France recently, I kept getting distracted by the color-coded letter grades plastered across food packaging. This is the Nutri-Score, a voluntary A-to-E rating system for nutritional quality introduced in 2017.
At a glance, you know where a product ranks. When researchers first proposed the idea in 2014, not a single food company agreed to use it. Only six brands signed on by launch in 2017. But momentum shifted, and today over 1,500 brands have adopted the system.
Participation isn’t universal, though: in a single aisle, one brand of gnocchi wears its C proudly, while the competitor right next to it opts out entirely, declining to advertise its own C rating.
What it measures
Nutri-Score assigns negative points for nutrients you want less of (calories, saturated fat, sugar, and sodium) and positive points for the ones you want more of (fiber, protein, fruits, vegetables, and legumes). However, it ignores food processing entirely.
If you survey French consumers, most assume an A label means “unprocessed” and an E means “highly processed.” In reality, the opposite often happens. Because the system focuses purely on nutrient counts, manufacturers could improve a product’s grade simply by cutting calories or sugar and replacing them with artificial additives and ultra-processed ingredients.
In a classic case of “you get what you measure,” a label designed to encourage healthier eating ended up rewarding higher levels of food processing. (France has since updated the system to address some of these loopholes.)
Does the label even work?
Credit to France: they ran a randomized controlled trial before rolling out their A-to-E nutrition labels, and that format proved most effective at shifting consumer behavior. However, cigarette packaging already established a clear benchmark for warning labels. Text-only warnings, like the U.S. Surgeon General’s, have little to no effect on smoking rates. In response, several countries escalated their approach by introducing graphic images, such as severe tooth decay, which proved far more impactful. Applying that same visual approach to soda, using images of rotting teeth, similarly outperformed plain text warnings. Even so, one graphic-label study found that the strongest immediate reaction was people hiding their cigarette packs in public, a sign of shifting social norms, even if overall consumption takes longer to drop.
Some research suggests that what actually determines whether a label, especially a calorie count, changes behavior is surprise.
If you didn’t realize a Starbucks muffin was loaded with calories, seeing the exact number can prompt you to choose something else. But if you already know a food is healthy or unhealthy, a label simply confirms your baseline assumptions. (And no one is rigging the system to favor junk food: under France’s Nutri-Score, French fries earn an A based on their raw nutritional profile before frying.)
It moves you toward salad more than away from cake
Early on, a controlled online trial found the Nutri-Score label worked: shoppers who saw it chose lower-calorie, lower-fat foods. Then the bigger studies came in. The most effective Nutri-Score label lifted purchases of food in the top third of its category by 14%, but did nothing for the medium, low, or unlabeled products. Net, it improved the nutrition quality of a basket by 2.5%.
So the answer isn’t “it works” or “it doesn’t.” It works in one direction by moving people toward the healthier option, but it doesn’t stop them from buying the cake. Same as walking into a McDonald’s: a label won’t change the very unhealthy purchase, but it can move you at the margin. The trouble is, that margin isn’t big enough to cover the calorie change you’d actually need.
There’s a broader lesson here: when you’re trying to change behavior, you can either make a good behavior better or make a bad behavior less bad. You can get someone to eat a salad every day, or get them to drink less soda. Both matter. But often, there’s more room to move at the bottom of the distribution: making an unhappy user less unhappy can have a bigger effect than making an already happy user slightly happier.
So when you evaluate an experiment, don’t just ask, “Did it work?” Ask: “Who did it work for, and where was there actually room to improve?”
Reorder the options, change the choice
Now for the interesting part. Because labels alone barely shifted purchasing habits, researchers tested a different lever: default sorting order.
In an experiment using an online grocery interface (similar to Instacart), researchers changed the default product display from alphabetical order to healthiness rank. Average diet scores rose significantly, and purchases of A- and B-rated items surged. The team didn’t add new features. They simply grouped healthier items at the top where they were easiest to reach.
This is the ordering effect, which proved far more effective than search filters. Fewer than 10% of shoppers even bothered to apply a filter. You don’t need to force behavior changes; you just have to put the right choices first.
The ordering effect works on small decisions and big ones. In a recent study, researchers reordered a list of teaching jobs, placing hard-to-staff schools at the top instead of sorting alphabetically. More teachers applied to those high-need schools.
Because human attention is scarce, top-ranked items do double duty: they are easier to process, and they carry an implicit endorsement. The first option becomes the baseline against which every other choice is measured.
We see the exact same principle drive product growth. On a past podcast episode, Kyle (who led growth at Credit Karma Money) described a signup form requiring users to select their industry for SEC compliance. Once they moved the most common industries to the top of the list rather than keeping them alphabetical, users no longer had to scroll to find their match and conversion rates climbed instantly.
Reordering defaults is one of the simplest, highest-leverage UX changes you can test today.
Reworked something with the order rather than the content lately? I’d love to hear what moved. Email me at kristen@irrationallabs.com.
Have a friend who would enjoy these teardowns? Click the button below to refer them.👇
