Rflct

Product development starts with the decision context

A new product does not succeed because it is technically superior. It succeeds when it solves a decision problem better than existing alternatives, in the context where the purchase actually happens.

Most product development processes start with the idea. We believe they should start with the decision context: when does the customer buy? Which alternatives do they consider? What decides the choice?

If you do not understand the decision context, you risk building products optimised in a vacuum — strong in the test, mediocre on the shelf.

Reflect maps the decision context first, tests concepts second, and calibrates against observed purchase patterns where data is available.

Upcoming visual
Decision context mapping — category behavior × purchase situation

Key takeaways

  • Start from the decision context, not the idea
  • Understand which alternatives the customer actually considers
  • Test in realistic context, not in isolation
  • Calibration against observed data strengthens predictive power
  • The product that wins the test does not always win in reality

Example

A beverage company developed a new flavour variant that topped blind tests. But in-store — surrounded by established brands and price anchors — consumers chose the safe option. The decision context (low involvement, fast choice, strong recognition need) did not favour the newcomer.

Related articles

How to evaluate concepts

Concept evaluation is not about asking consumers whether they like a concept. It is about understanding which attributes drive preference, what separates the concept from alternatives, and whether enough people accept the price point.

From insight to concept prioritization

Insights without a prioritization framework rarely lead to action. Reflect connects consumer insights to a quantified ranking of concept alternatives, based on expected market potential, not just appeal.

See related service

Discuss product development with us

Contact us
Back to Product development with data, not guesswork