People chose a hypothetical quit app: did they actually use one?
Separate a discrete-choice experiment’s stated tradeoffs from downloads, continued use and demonstrated quitting outcomes.
No Vape · Updated

Find what was on the choice card
A report might vary developer description, presentation, cost or another attribute across imaginary services. List only the attributes and levels the study actually offered. Unlisted features were not silently tested. Also check whether people could choose neither option, since a forced comparison and a choice including an opt-out ask different questions.
Szinay and colleagues’ original smoking-cessation-app experiment explicitly describes hypothetical alternatives and distinguishes stated preferences from real-life uptake in its limitations. Its UK smoking population and numerical findings are not transferred to US vaping or No Vape. We use the task distinction to examine what a public claim means.
Use an invented tradeoff without creating a sales forecast
Imagine a fictional study offering app A with a shorter description and app B with a longer description, alongside other specified attributes. Someone selecting A has answered within that particular set. They have not necessarily chosen A over every existing app, committed to a purchase or said that short descriptions cause successful quitting.
Write three lines: what varied, what respondents selected, and what an advertisement claims. If the last line says most adults will download the product, identify the additional real-world evidence needed. The exercise contains no actual percentage, willingness-to-pay estimate or appraisal of a specific brand.
Keep preferences and model predictions on their own scales
A report can estimate attribute preferences or predict choices for a hypothetical combination. Read the model’s stated assumptions and scenario. A predicted choice share is not an observed installation count or a guarantee about the market. Do not turn a preference parameter into the probability that a feature helps someone quit.
The presented attribute range also frames the comparison. A result identifying the strongest influence among selected attributes cannot automatically rank all possible product qualities. Keep recruitment, context and stated limitations visible. That does not make the experiment useless: it can inform a focused design question while leaving behavior outside the task unmeasured.
Ask which later behavior was actually observed
Look separately for downloads, purchases, continued use and defined quitting outcomes if a claim mentions them. These are different stages, each requiring its own observation and method. Satisfaction among existing users is another task, and a voluntary poll should not be relabeled as a discrete-choice experiment just because respondents selected an answer.
Finish with a bounded description of the choices studied and the inference you can verify. For your own app decision, evaluate the actual task you need, current access terms and trustworthy evidence for any health claim. A hypothetical preference study neither chooses a treatment for you nor establishes the effectiveness of the app whose attributes resemble a favored profile.
Sources and scope
Sources inform the guidance. The worksheets and examples are No Vape editorial suggestions, not tested treatments.
Written with AI assistance and checked against the cited sources. No clinical review is claimed. This guide covers practical planning, not individual medical treatment. How we research and write