Skip to content
PansophIQ

Capabilities · Decision modeling

Choice models that predict the market, not just describe it.

Your client is deciding how to launch, price, or position. The research has to say what will happen to share — and hold up when their forecasting team checks the work. We design the experiment, estimate the model, and hand over a simulator they run themselves.

Intelligence from the market’s own choices.

Discuss a project Sign in

Designs that look like the market

  • Alternative-specific. Each alternative carries the attributes it actually has. A feature that exists for one product is asked about for that product only.
  • Real choice sets. The alternatives that compete, the availability that constrains them, and a none option where the market has one.
  • No trade-offs among things that never co-occur. A design that asks people to trade off what the market never offers produces weights that predict nothing.
  • Full-range extraction. The design is extracted across its entire used range; truncation fails loudly, never silently.

Uncertainty that survives to the answer

  • Behavior-predicting weights. Estimation yields weights that predict choice, validated on held-out tasks — not descriptions of attitudes.
  • Bayesian, and simulated from the posterior. Every scenario is simulated from the full set of posterior draws, so the simulator shows a range, not a false point.
  • Reported as shares, in percentages. Every effect is a share change — the language of a commercial decision.

The simulator is the deliverable

An interactive market simulator your client opens and runs: change price, availability, or profile and read the share change with its interval. Scenarios saved and compared. Client-ready, in your branding, with the base case and the assumptions on the face.

Market simulator — scenario vs. base
Alternative A38.2%
Alternative B27.5%
Alternative C21.0%
None of these13.3%
Shares shown with 90% intervals from posterior drawsScenario differs from base on availability only

Market simulator — scenario vs. base. Representative Illustrative values.

The decisions it answers

Launch sequencing · pricing and contracting · positioning against named competitors · line extension and portfolio · forecast inputs (share by scenario, with ranges).

Methods that survive the client’s stats team

A methods appendix with the design, the estimation, the validation on held-out tasks, and the diagnostics — written for their document. Naive-reference comparisons and sensitivity bands travel with every deliverable.

Common questions

What sample do we need?

It depends on the design’s size and the segments your client needs to read; we size it in the proposal and say what each option buys.

How long?

Design and programming in days once the brief is settled; the field runs on schedule; estimation and simulator typically within a week of field close.

What do you need from us?

The decision, the alternatives and attributes that matter, the market’s availability constraints, and your client’s template.

Can our client’s statisticians see the model?

Yes — the appendix is written for them, and we join the call if you want us to.

Do you do rating-scale “importance” instead?

Not for a decision this size; a choice experiment is what predicts.

Bring us the decision. We design the experiment that answers it and deliver the simulator that shows it.

Discuss a project Sign in