All posts

Quantitative

Conjoint vs. asking what people want

July 28, 2026 · 9 min read

Stated importance lies. Trade-offs reveal preference — when the instrument is designed correctly.

Stated vs. revealed

3.2×

Average spread between top-ranked stated-importance features and actual choice drivers in conjoint.

Sample quality

#1 factor

Instrument failure almost always traces to wrong sample or unrealistic attribute levels — not model choice.

Triangulation win

91%

Pricing recommendations that held post-launch when conjoint was paired with qual follow-ups.

Why stated importance fails

Asking respondents to rate feature importance produces socially desirable, undifferentiated scores. Everything is "important." The method measures salience, not preference.

Bundle A · $49
Bundle B · $79

Forced trade-offs between realistic bundles reveal preference — rating scales do not.

Insight

Table-stakes trap

Teams that rely on stated importance over-invest in features everyone rates highly and under-invest in differentiators that actually drive choice at realistic price points.

Product teams that rely on stated importance alone often over-invest in table-stakes features and under-invest in differentiators that actually drive choice.

Trade-off design fundamentals

Conjoint and discrete choice methods force trade-offs: bundles of attributes at realistic price points, with experimental design that separates main effects.

3–5 each

Attribute levels

8–12

Tasks per respondent

2+

Holdout concepts

Fantasy bundles respondents would never see in market. Too many attributes causing cognitive overload. Sample that includes non-buyers for a product-only decision. Missing holdout validation concepts. Any one of these produces precise nonsense.

Instrument quality depends on attribute selection, level realism, and sample definition. A conjoint on the wrong sample or with fantasy bundles produces precise nonsense.

When to use each method

Use Van Westendorp and Gabor-Granger when the decision is primarily price-led. Use choice-based conjoint when feature bundles and willingness-to-pay for combinations matter.

Price sensitivity curves from Van Westendorp anchor conjoint price levels.

Insight

Conjoint tells you what; qual tells you why

Adaptive interviewing after conjoint explains the narrative behind trade-offs — especially when a segment chooses unexpectedly or when price sensitivity breaks the model.

Use adaptive interviewing when you need the "why" behind the trade-off — conjoint tells you what they choose; qual tells you what the choice means.

Integrating quant and qual

Our strongest pricing engagements triangulate: conjoint for structure, adaptive interviews for narrative, and behavioral data where available for validation.

Qual
Quant
Market

Quant structure + qual narrative + behavioral validation = decision-grade pricing.

A simulator without limits is a liability. A recommendation with limits is an asset.

Insight

Ship limits with the model

Every deliverable includes segment-level holdout accuracy, confidence bands, and explicit notes on where the model breaks — so leadership knows when to trust the numbers and when to run a follow-up wave.

The deliverable is not a simulator alone — it is a recommendation with explicit limits, confidence grades, and the segments where the model holds versus breaks.

Want the full research brief or a similar engagement for your team?

Contact us →