How do you prioritize features with customer input?
TL;DR
- Don't ask "how important is this feature?" alone. Everything scores high.
- Best-worst ranking (MaxDiff) gives a forced order; the importance-satisfaction gap shows where unmet need sits.
- In What Customers Want (2005), Anthony Ulwick set out an opportunity score that adds importance to the gap between importance and satisfaction.
- Neon Blue runs both methods on a simulated audience of your users and hands back one ranked list per segment. Run your first study free
Neon Blue builds simulations of human behavior. You prioritize features with customer input by asking buyers to make trade-offs, not to rate each feature on its own. Rate-everything surveys come back with every feature marked "important". Two methods fix that: a best-worst ranking that forces choices, and an importance-versus-satisfaction gap that shows where current options let people down. Run both on the same list, by segment, and the roadmap argument moves from opinions to a ranked table.
Why do feature surveys usually fail?
Most feature surveys ask people to rate each idea on a 1 to 5 importance scale. People are generous, nothing costs them anything, and the result is a wall of 4s. Requests from the loudest customers fill the gap, and the roadmap tilts toward whoever emails most. The same generosity inflates purchase intent scores when nothing is at stake. Two fixes help: make people choose, and ask how well today's options already do the job.
How does the opportunity score work?
In What Customers Want (McGraw-Hill, 2005), Anthony Ulwick described outcome-driven innovation. Customers rate each desired outcome twice on a 1 to 10 scale: how important it is, and how satisfied they are with how they get it today. The opportunity score is:
Opportunity = importance + max(importance minus satisfaction, 0)
An outcome that matters a lot and is badly served scores high. One that matters a lot but is already well served scores lower, because a new feature there has less room to win. One that doesn't matter scores low no matter how badly it is served. The method works on outcomes ("know what I'll pay before checkout") rather than feature names ("price preview"), which keeps people from voting on jargon. Writing outcomes in buyers' words is the same skill used in value proposition testing.
How do best-worst ranking and the opportunity score fit together?
| Method | What people do | What you get | Best for |
|---|---|---|---|
| Best-worst ranking (MaxDiff) | Pick the most and least useful from sets of 4 or 5 features | A ratio-scaled order of every feature | Cutting a long list to a shortlist |
| Importance and satisfaction | Rate each outcome twice | An opportunity score per outcome | Finding underserved needs |
| Both on the same list | Both tasks | Features that rank high and solve an underserved need | Roadmap calls with a budget limit |
A feature that ranks near the top in best-worst and maps to a high-opportunity outcome is a strong candidate. One that ranks high but maps to a well-served outcome often wins because it is familiar, not because it is needed. Choosing between MaxDiff and a full trade-off design is covered in MaxDiff or conjoint. Earlier-stage teams sorting must-haves from delighters can use the Kano questions in testing a product idea.
What does a feature prioritization study look like on a simulated audience?
Hypothetical: a national grocery chain's app team has 12 candidate features and engineering capacity for three next half.
- Audience: US adults who order groceries online at least twice a month, split by app users versus web-only shoppers, households with kids versus without, and weekly versus occasional orderers. Finding your target audience covers drawing those cuts.
- Questions to field:
- Of these five features, which would be most useful to you, and which least? (repeated across sets until each feature appears several times)
- How important is it to you to [outcome, for example "know the final total before checkout"], from 1 to 10?
- How satisfied are you with how you do that today, from 1 to 10?
- If the app added only one of your top three, which would make you order more often?
- Which of these would you never use? [Multi-select]
- Design: the 12 features mapped to 9 outcomes, fielded on a simulated audience of shoppers, both tasks for each respondent.
Example output layout (illustrative, not a real result)
| Feature | Best-worst rank (app users) | Mapped outcome | Opportunity score | Would order more often | Shortlist? |
|---|---|---|---|---|---|
| Price preview at checkout | __ | Know the final total | __ | __% | Yes / No |
| Substitution approvals | __ | Get what I ordered | __ | __% | Yes / No |
| Reorder last basket | __ | Save time on repeat orders | __ | __% | Yes / No |
| Recipe-to-cart | __ | Plan meals faster | __ | __% | Yes / No |
The three that pass both screens go to design. The "would order more often" column ties the ranking to a business result the growth team can test after launch.
How do you turn the ranking into a roadmap?
- Read by segment before totals. A feature in the middle overall can top the list for weekly orderers, who may be most of revenue.
- Ask about the bundle. Three features that each rank well may overlap. A short follow-up on the shortlist shows whether people want all three or just one.
- Check the price effect. If a feature is meant to support a paid tier, run a short pricing research read on the tier with and without it.
- Pair with live data. Usage logs and support tickets show what people do today; the simulated read shows what they would choose among things that don't exist yet. For the shortlisted features, a few live interviews with real users add the words the design team needs.
Product marketers who also own the launch message can pick up the thread in the product marketing research workflow.
Questions about feature prioritization
What is the best feature prioritization technique?
There isn't one best. Use a forced-choice ranking to cut a long list, and an importance-satisfaction gap to find unmet needs. Using both on the same list catches the errors each makes alone.
How many features can you test in one survey?
Best-worst ranking handles 15 to 30 features comfortably because each person sees small sets. Importance and satisfaction ratings get tiring past about 15 outcomes, so group or split them.
Should you ask customers which features they want?
Ask about outcomes and trade-offs, not wish lists. "Which would you give up to get this?" says more than "would you like this?"
How does Neon Blue run a feature prioritization study?
Neon Blue fields your feature list on a simulated audience of your users, runs best-worst and importance-satisfaction tasks together, and returns a ranked shortlist by segment, all marked as simulated. Start a free first study.