Despite the complexities of GLP-1 treatment, a simple recommendation question may be surprisingly revealing about whether patients will switch or stop.

‍Net Promoter Score, or NPS, has traveled widely since Fred Reichheld introduced it in 2003, drawing inspiration from Enterprise Rent-A-Car’s approach to customer loyalty. Its appeal is simplicity: ask people how likely they are to recommend a product or company. Those answering nine or ten are “promoters”; those giving six or below are “detractors”. Subtract the share of detractors from the share of promoters and management has a number to watch.

A metric inspired by car rental might seem an unlikely guide to the fortunes of prescription medicines. GLP-1 treatment decisions are complex, involving a prescriber’s judgment, an insurer’s rules and a patient’s willingness to live long-term with the trade-offs between effectiveness, side effects and cost. Can a single score really anticipate behavior amid all this complexity?

Using longitudinal data from more than two years of dQ&A US Patient Voice research, we tested whether recommendation ratings anticipated subsequent treatment changes. We followed people with type 2 diabetes taking Mounjaro or Ozempic, the two most popular current GLP-1 treatments for T2D, linking their ratings with later reports of switching brands or quitting the drug class altogether.

One survey can establish that patients are unhappy today. Following the same people reveals whether that unhappiness precedes a change in treatment. Our analysis tracked 776 participants from early 2024 through mid-2026. We measured treatment status every three months and collected recommendation scores every six months, and weighted the sample using age, sex and regional benchmarks for GLP-1 users from the 2024 National Health Interview Survey.

The results were striking. An average of one in six patients were detractors. They were roughly twice as likely as promoters to switch brands, and twice as likely to quit the GLP-1 class entirely. Specifically, 43% of detractors reported switching brands compared with 21% of promoters, and 35% of detractors reported quitting the class compared with 18% of promoters.

The steep early rise in departures makes a detractor score a flashing red alarm. By three months, 16% of detractors had left their original drug, compared with 7.8% of promoters. By six months, 30% of detractors had left, versus 12% of promoters. The relationship held across four successive measurement waves.

Our AI analytics platform, now available to customers, makes this deep analysis of a research dataset containing 48,000 completed surveys and over 25 million data points possible in hours rather than days.

What lies behind the recommendation score? We checked whether the severity of six potential problems (e.g. side effects, injection difficulties, cost) and satisfaction with six treatment attributes (e.g. weight management, glycemic control, device experience) predicted switching or discontinuation.

A patient’s overall verdict can contain useful information even when its component parts are difficult to disentangle. Only gastrointestinal side effects—nausea and vomiting—independently predicted class quitting, but not brand switching. None of the individual attribute satisfaction scores showed a strong relationship with future switching or quitting. This suggests that NPS delivers a broader assessment of treatment than the prevalence of any one problem or attribute.

This is a reason to pay attention to NPS in the diabetes drug space. For commercial teams and insights managers, a low score signals an elevated risk of near-term departure. For brand teams, it’s a cue for investigating where patient support or communication may be falling short. For leadership, it adds a forward-looking dimension to the interpretation of prescribing and sales trends.

The true value of NPS stems from using a research method that tracks the same respondents over time. A succession of random-sample surveys can chart the market’s mood, as long as the sample frame is consistent every wave and all the respondents are legitimate: increasingly challenging in 2026. In our US Patient Voice data, repeated observations of the same people connect sentiment to later decisions, which is consistently more useful to commercial planning because an opinion is linked to subsequent behavior change.

NPS cannot capture all the complexity of the decision to start or stop GLP-1 treatment, but that’s not the point. A simple recommendation question turns out to have useful predictive power for attrition and should be the springboard for qualitative and quantitative efforts to learn more.

The rental-car question has traveled a long way. In the GLP-1 space, its value becomes clearer when we ride along for more of the patient’s journey.