“Definitely would buy” is not a purchase. It's a sentence. Sometimes a human says it. Sometimes a model generates it. Either way, the credit card hasn't moved.
Purchase intent can help compare reactions, expose objections, and decide what to test next. It becomes dangerous when a team quietly relabels it as conversion, revenue, ROAS, or product-market fit.
The short answer
Key takeaways
- Keep human stated intent, synthetic intent, and observed behavior in separate columns.
- Read distributions and conditions, not just an average score.
- Move from concept clarity to human response to real commitment as the decision gets more expensive.
Why intention and action split up
Real purchases include price, shipping, timing, trust, alternatives, cash constraints, partner approval, stock, checkout friction, and the sudden realization that the old product is probably fine. A survey stimulus strips away much of that mess.
People also misremember, please interviewers, or answer an idealized version of themselves. Models add their own problems: prompt sensitivity, training-data bias, stereotype, and no real budget.
Sources: Morwitz, Steckel, and Gupta: When do purchase intentions predict sales?, Sun and Morwitz: Stated intentions and purchase behavior
Read an intent distribution without fortune-telling
Look for shape. Is the result polarized? Mostly neutral? Strong but fragile around price? Do the written reasons repeat one condition? Those patterns suggest hypotheses.
Don't convert top-box share into an expected conversion rate. Don't multiply mean intent by traffic and call the result revenue. Different instruments, different samples, different worlds.
The evidence ladder
- Comprehension: can people explain the offer correctly?
- Synthetic screen: what objections and audience assumptions appear?
- Human stated response: what do relevant people say about the same stimulus?
- Low-friction behavior: will they click, save, or join a transparent waitlist?
- Costly behavior: will they deposit, preorder, or purchase?
- Business quality: do margins, returns, repeat use, and retention hold up?
Sources: Shopify: How to test a business idea, Shopify: How to run an A/B test
How to read PaperPMF
PaperPMF returns synthetic intent distributions. The free 10-response preview and separate 100-response report are generated independently. The full report may help you inspect reactions and subgroup filters, but it still doesn't measure customers or conversion.
Use a strong result to fund the next question. Use a weak result to revise or stop. Either way, the diagnostic has done its job when it changes what you test—not when it gives the launch deck a heroic-looking number.
Sources: PaperPMF methodology
Sources and verification
Product details are based on official documentation reviewed on August 4, 2026 unless noted. Features and pricing can change; verify them with the provider before making a purchase.
- Morwitz, Steckel, and Gupta: When do purchase intentions predict sales? — Peer-reviewed evidence on moderators of the intention–purchase relationship; reviewed 2026-08-03.
- Sun and Morwitz: Stated intentions and purchase behavior — Peer-reviewed modeling of systematic gaps between intention and behavior; reviewed 2026-08-03.
- Shopify: How to test a business idea — Official staged validation guidance; reviewed 2026-08-03.
- Shopify: How to run an A/B test — Official behavioral experiment guidance; reviewed 2026-08-03.
- PaperPMF methodology — First-party report documentation and interpretation limits; reviewed 2026-08-03.
