Conjoint or Tradeoff Survey¶
Assessment instrument — instantiates Acceptable Substitution Mapping
Estimates how stakeholders value different attribute combinations by asking them to choose between bundles, then infers the exchange rates hidden in their picks.
A Conjoint or Tradeoff Survey is a structured choice experiment that never asks people what they value directly. Instead it shows them competing bundles — each a different mix of attributes at different levels — makes them pick, repeats this across many carefully designed comparisons, and then infers from the pattern of choices how much of one attribute a respondent will give up to gain another. Its defining move is that it converts a fuzzy, socially awkward question ("how much is speed worth to you versus price?") into an estimated exchange rate derived from behavior a respondent finds easy: choosing between two concrete options. Because the rates are estimated statistically over a sample, it also tells you whose preferences produced them and how they vary across segments — which is exactly what a substitution map needs when it must decide whether a swap acceptable to one group is acceptable to another.
Example¶
A regional transit authority is redesigning its bus network and needs to know what riders will actually tolerate. It cannot afford to be more of everything, so it must trade — a lower fare but longer waits, or shorter waits but more crowding. It fields a conjoint survey. Each respondent sees a series of paired trips: Trip A — $2.50 fare, 12-minute wait, standing room versus Trip B — $3.25 fare, 6-minute wait, a seat. They just pick the one they'd take, twenty times, over varied combinations. From the choices, the analysts fit a model whose coefficients reveal the trades: on average, riders will accept about seven extra minutes of waiting to save a dollar, but crowding is weighted far more heavily than the planners assumed — a seat is worth roughly what a five-minute-shorter wait is worth. That single finding reshapes the plan: the network protects seating capacity on long routes even at some cost to frequency, because the survey showed frequency is the more substitutable attribute and comfort the less. Crucially, the model also splits by rider type, revealing that commuters and occasional riders have almost opposite exchange rates — so the map can't assume one acceptable trade for everyone.
How it works¶
The instrument's distinguishing machinery is that it measures trades indirectly and statistically:
- Define attributes and levels. Choose the dimensions that vary (fare, wait, crowding) and a small set of realistic levels for each. This is the frame; everything downstream is expressed in these terms.
- Design the choice sets. Combine levels into bundles using an experimental design so the attributes vary independently, which is what lets the analysis separate the effect of each one.
- Force repeated choices. Respondents pick among bundles many times; no rating scales, no "how important is X" — just picks that mimic real selection.
- Fit a preference model. A discrete-choice model turns the choices into part-worth utilities; ratios of those utilities are the estimated exchange rates — how much of one attribute compensates for another.
Tuning parameters¶
- Attribute and level count — how many dimensions and steps enter the design. More captures a richer trade space but overloads respondents and inflates the sample needed for stable estimates.
- Number of choice tasks — how many comparisons each respondent faces. More data sharpens the estimates but fatigues people, and tired respondents start using shortcuts that bias the result.
- Sample composition — who is surveyed and how they're segmented. This decides whose exchange rates you learn; a convenient but skewed sample yields precise answers to the wrong population's preferences.
- Willingness-to-pay anchoring — whether price is an attribute so trades can be expressed in currency. Money anchors make rates legible but drag in income effects and can crowd out non-monetary values.
- Full-profile vs. partial — whether bundles show all attributes or a subset per task. Fuller profiles are realistic but harder; partials are easier but assume the omitted attributes don't sway the choice.
When it helps, and when it misleads¶
Its strength is putting a defensible number on trades people struggle to state, and exposing compensations that intuition gets backwards — like the transit finding that comfort dominated frequency. Grounded in random utility theory,[n1] it treats each choice as evidence of an underlying preference and recovers the exchange structure from many such observations, segment by segment.
Its central weakness is hypothetical bias: a survey choice costs nothing, and stated trades routinely diverge from what the same people do when real money and consequences are on the line. Pile on too many attributes and respondents cope by ignoring some, so the model fits a simplified decision that isn't theirs. The classic misuse is reading an estimated rate as a hard equivalence — treating "seven minutes per dollar on average" as license to swap freely — when the average hides wide variation and says nothing about the attributes that must never be traded at all. The discipline is to validate estimated rates against real behavior where possible, keep non-negotiables out of the traded set entirely, and carry the segment spread rather than collapsing to one deceptive average.
How it implements the components¶
The survey realizes the measurement slice of the archetype — it quantifies preferences and trades, and stops there:
substitution_rate_or_exchange_ratio— its primary output: estimated rates at which one attribute compensates for another, read from the fitted utilities.stakeholder_preference_source— it grounds the map in a real, sampled population's choices and reports how trades differ across segments, so acceptability is tied to whose values were measured.attribute_or_resource_dimensions— the attribute-and-level design defines the dimensions along which bundles are compared and traded.
It does not name the outcome_requirement that every acceptable bundle must preserve, nor convene stakeholders in deliberation — that is the Preference Elicitation Workshop, its nearest twin: the workshop deliberates trades qualitatively in a room, while this survey estimates them statistically from many independent forced choices. Nor does it render the results as a preference_surface_or_equivalence_region — that is the Indifference Map.
Related¶
- Instantiates: Acceptable Substitution Mapping — it supplies the measured exchange rates a map turns into substitution rules.
- Sibling mechanisms: Preference Elicitation Workshop · Indifference Map · Design Tradeoff Curve · Equivalency Table · Approved Alternative Catalog · Procurement Equivalence Review · Service Tier Mapping · Substitution Matrix · Exception Review Workflow
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: Estimates how stakeholders value different attribute combinations by asking them to choose between bundles, then infers the exchange rates hidden in their picks, making its operative form a bounded trial, probe, simulation, or adversarial exercise that generates evidence from performance.
Independent corroboration: The frozen evidence defines Conjoint or Tradeoff Survey as 'Estimates how stakeholders value different attribute combinations by asking them to choose between bundles, then infers the exchange rates hidden in their picks', so its operative form is Experiment, Test & Rehearsal.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Psychology
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Mathematical psychology first cohered conjoint measurement as recovery of latent attribute tradeoffs from judgments over bundled alternatives.
Related originating lineages:
- Economics & Finance — Random-utility and demand analysis recast bundle choices as marginal rates of substitution and willingness to pay.
- Statistics & Experimental Design — Factorial choice designs and estimation supply the survey instrument and segment-level inference.
Review resolution: The reviewers agree on psychology with economics_finance and statistics_experimental_design as the relevant lineages. The survey form converges conjoint measurement, experimental tradeoff elicitation, and random-utility choice analysis, and it is used beyond a single application domain.
Attribution caveat: The survey form was shaped jointly by conjoint measurement, marketing choice research, and random-utility economics.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] Random utility theory, the foundation of discrete-choice / conjoint analysis, models each option as carrying a utility that is partly systematic (a weighted sum of attribute levels) and partly random; a respondent picks the highest-utility option, so observed choices identify the attribute weights — and their ratios, the marginal rates of substitution. ↩