Abstract Randomized controlled trials (RCTs) are increasingly popular for assessing development programmes. This study investigates the validity of extrapolating RCT results to large‐scale programmes, using the example of the national bed net subsidy programme in Tanzania that later added a free distribution campaign. Although the RCTs suggest steady sales of nets throughout, an interrupted time series model reveals a 34 per cent monthly decline immediately after the campaign compared to ‘normal’ sales. After 6 months, the total unsold nets reached 45 per cent of normal sales, or 347 000 nets nationwide. We hypothesize that these differences in results arise from various scale and spillover effects. Copyright © 2013 John Wiley & Sons, Ltd.