Stated switching intent consistently overstates observed switching. The direction is usually right; the magnitude is not — and the gap narrows sharply when the question names a specific patient type rather than the therapy in general.
What is stated switching intent worth?
Directionally a lot and quantitatively very little. Prescribers reliably identify which therapies they find interesting and unreliably predict how often they will actually prescribe them.
That is a general property of stated-intent research rather than anything specific to clinicians — B2B software buyers overstate it the same way — and it is well documented — which does not stop forecasts using the raw figure.
The asymmetry between direction and magnitude is what makes the data usable at all. A study reporting which segments are most open to switching is reporting something reliable; the same study reporting a percentage who will switch is reporting an artefact of the question. Both come out of the same fieldwork, and only one of them should reach a forecast.
How big is the intent gap?
Large enough to change a launch case, and consistent enough to be worth measuring rather than guessing. The consistency is the useful part: a systematic bias can be corrected for, a random one cannot.
It also varies by how the question was asked, which means the same study run two ways produces two different overstatements.
Consistency is also what makes a correction factor possible in principle. If the same instrument overstates by a similar margin across studies in a therapy area, that margin can be estimated and applied. What defeats it in practice is that most organizations do not keep the observed outcome alongside the original stated figure, so the comparison that would calibrate the next study never gets made.
How should the question be asked?
Anchored on a specific patient type. "For a patient presenting like this, what would you do" produces answers materially closer to behavior than "would you consider switching", because it engages clinical reasoning rather than general disposition — which is how a prescriber survey should be written.
It also surfaces the segmentation. Prescribers who would switch for one patient type and not another are describing the real shape of adoption.
Vignettes have a second advantage: they make answers comparable across respondents. Two prescribers asked in general terms are answering about different patients without either of them saying so, while two answering about the same described patient are answering the same question. That is what allows the segmentation to be read as a segmentation rather than as noise.
What actually stops a switch?
Habit, formulary friction and the effort of monitoring a change — rarely a clinical objection. Prescribers who say they would switch and do not usually cite none of the reasons a clinical study would have anticipated.
That makes barrier research a different exercise from efficacy research, and one a handful of prescriber calls can settle.
Monitoring effort is the barrier most often left out of a launch plan. A therapy requiring an additional test, a dose adjustment or a follow-up appointment imposes work on a practice that is already full, and that work sits outside anything the clinical case addresses. It is also the barrier most amenable to being designed around once it has been named.
How should a forecast use this?
As a ranking rather than a rate. Intent data reliably orders segments by likelihood of adoption; it does not reliably estimate how many, and models that treat it as a rate build the overstatement into the forecast — before the coverage decision caps it again.
Where a rate is needed, the correction factor has to come from somewhere — either an observed analogue or a measured gap, not an assumption.
The ranking is also more durable than a rate. Segment ordering tends to hold as a market moves, while any absolute figure ages with the competitive set, which makes the ranking the part of the finding worth carrying forward into the next planning cycle.