Ten Sri Lankan offices got a diet programme matched to how ready each worker was to change. After 12 weeks, 39% of their staff met the healthy-diet mark, against 3% in ten offices left alone

Most diet advice is handed out the same way to everyone: eat more fruit, cut the sugar, here is a leaflet. Behaviour researchers have long argued that this misses half the audience. Someone who has never thought about changing needs a reason to start. Someone who is already trying needs help keeping it up. A trial published on 9 October 2026 in BMC Public Health tested that idea on desk workers in the south of Sri Lanka.

The team, from the University of Ruhuna's medical faculty and Sri Lanka's Health Promotion Bureau, picked 20 government offices in the Galle district, each with at least 30 staff and more than 5 km apart so news of the programme wouldn't spread from one to the next. Of 610 workers invited, 518 agreed to take part: development officers, management assistants and administrators whose jobs keep them in a chair. Ten offices were drawn by lot to get the programme (266 people) and ten to carry on as usual (252 people). Most participants were women, and their median age was 39.5.

Each worker was first sorted by the classic 'stages of change': not thinking about changing their diet, thinking about it, getting ready, already doing it, or keeping it up. A 24-hour food recall checked the answer against what they actually reported eating. People in the first two stages were given three 30-minute talks by a doctor, one a month, on chronic disease, what an unhealthy diet costs, and how to switch, with a recorded version sent to anyone who missed one. People in the later stages instead got a daily serving counter, a sheet or phone app on which they ticked off servings of each food group against Sri Lanka's national targets, with the too-little and too-much zones marked in red. An investigator came by at the end of each month to check the sheets.

After three months, 62.0% of those in the programme offices had moved up at least one stage, against 11.2% in the comparison offices. The diet results tell more. In the programme offices, average daily fruit rose from 0.7 to 1.1 servings and vegetables from 3.0 to 3.3. Rice, bread and other starchy foods fell from 6.4 to 5.9 servings, closer to the 6 to 7 the authors set for desk workers. Servings of foods the paper counts as unhealthy, things that are mostly sugar and fat or highly processed, fell from 2.2 to 1.5 a day. In the comparison offices every one of those numbers stayed where it was. Fish, meat, pulses and dairy didn't change in either group.

The researchers' main test of a 'healthy diet' was meeting the national serving targets for starchy foods, fruit and vegetables and a protein group, with no more than one unhealthy food a day. By the end, 38.7% of programme participants qualified, against 2.7% of the comparison group. After adjusting for other factors, the odds were about 25 times higher. People who were already in the doing or keeping-it-up stages at the start were the most likely to get there.

Those are big gaps for a few talks and a tally sheet, and that is reason to read the method closely. Nobody was blinded: workers knew whether their office had the programme. Diet was measured by asking each person, once before and once after, what they had eaten the previous day. People who have spent three months ticking off servings, and who know a researcher is interested in their fruit, may well report differently from people who haven't, even if their plates changed less than the recall suggests. The study measured no weight, blood pressure or blood tests, so there is nothing yet to show a change in health.

The authors list other weaknesses. The two groups weren't evenly matched at the start: the programme offices had twice the share of men (24% against 12%) and fewer people with a family history of chronic disease. The final analysis didn't account for the fact that workers in the same office tend to behave alike, which the authors justify by a low measured office effect but concede can't be ruled out entirely. Twelve weeks is short for a diet change to settle, and government offices staffed mostly by women may not tell us much about private firms or male-dominated workplaces. The study was paid for by the lead author, and the authors report no competing interests.

Even so, fruit at 1.1 servings a day is still about half the national target of two to three, and the authors are careful to call for a larger, more real-world test. What the trial does suggest is that the same message doesn't need to go to everyone in an office. A reason to change for the people who haven't started, and a simple way to keep score for the people who have, moved the reported numbers further than doing nothing did.