Jason Duke, Founder, Kronaxis
Tag: Research
Conjoint analysis has been the gold standard for pricing research since the 1970s. The method is elegant: present participants with combinations of product features and prices, ask them to choose, then statistically decompose the relative value each attribute contributes to the decision. If you want to know whether your customers value collaboration more than analytics, or whether a £20 price increase is worth adding priority support, conjoint gives you a defensible answer.
The method is sound. The bottleneck is everything that happens before and after the statistics.
The Traditional Process
A well executed conjoint study requires 200 to 500 participants who match your target market. Recruiting them takes two to four weeks through a panel provider, longer if you need a niche segment. Each participant completes 20 to 30 choice tasks, which takes 15 to 25 minutes per session. Data cleaning removes speeders, straightliners, and inconsistent responders, typically 10 to 20 percent of the sample.
Then comes the analysis. Hierarchical Bayes estimation produces individual level utility estimates, which are aggregated into market simulations. The statistician builds a model, runs sensitivity tests, and writes the report. End to end: four to eight weeks and £20,000 to £50,000 depending on sample requirements and analytical complexity.
For many organisations, this means conjoint happens once or twice a year, on major pricing decisions. The cost and timeline preclude using it for iterative product development, rapid feature prioritisation, or testing multiple pricing architectures in parallel. The method that could answer the question most precisely is too slow and expensive to use at the pace the business actually moves.
The Synthetic Approach
Build a 500 persona panel with census weighted demographics and DYNAMICS-8 personality profiles. Present the same choice tasks you would present to human participants. Each persona evaluates the options through its personality model, economic context, and life experience.
A high Discipline persona evaluates feature bundles systematically. It calculates value per pound, compares features against stated needs, and selects the option with the highest perceived utility relative to cost. A high Impulsivity persona gravitates toward the option that feels most appealing in the moment, weighting emotional resonance over analytical comparison. A high Yielding persona looks for signals of popularity: which tier is marked "most popular," which has the highest implied adoption.
These are not random variations bolted onto a demographic shell. They are coherent personality driven evaluation strategies that produce internally consistent choice patterns across dozens of choice tasks. When you run hierarchical Bayes estimation on synthetic conjoint data, you get individual level utility estimates that segment cleanly by personality type.
What You Gain
Speed is the most obvious advantage. A synthetic conjoint study that would take six weeks with human participants runs in under an hour. Panel generation takes minutes. Stimulus presentation and response collection are parallel across all personas. Data cleaning is unnecessary because synthetic personas do not speed through tasks or straightline responses.
Cost follows from speed. A 500 persona synthetic conjoint study costs under £200 in compute. Not £50,000 in panel fees, incentives, platform costs, and analytical time.
But the genuine advantage is not speed or cost. It is the personality segmentation.
Traditional conjoint produces utility estimates segmented by demographics: willingness to pay by age group, by income bracket, by region. Useful, but limited. Synthetic conjoint produces utility estimates segmented by personality dimension: willingness to pay by Discipline level, by Impulsivity, by Acuity.
This distinction matters because personality segments cut across demographics. A 28 year old startup founder and a 52 year old NHS administrator might both be high Acuity, high Discipline. They evaluate your feature set using the same cognitive approach despite having nothing in common demographically. Your pricing page needs to serve both of them, and personality segmentation tells you how.
Reproducibility is the fourth advantage. Run the same conjoint with the same panel twice and get the same results. Change one variable (swap a feature, adjust a price point, modify the framing) and the difference in results is attributable to that variable alone. Traditional conjoint has sampling noise, order effects, and day of week variation. Synthetic conjoint has none.
What You Lose
Synthetic willingness to pay is modelled, not observed. No money changes hands. No persona reaches for its wallet, hesitates, and puts it back. The psychological simulation produces realistic choice patterns, but those patterns are derived from personality models and population distributions, not from real purchase events.
Validation against real purchase data is the critical gap. Synthetic conjoint results align with known patterns from published pricing research and with traditional conjoint studies run in parallel. But "aligns with known patterns" is not the same as "predicts actual revenue within 5%." The field is building that evidence base. It is not complete yet.
Extreme edge cases may be underrepresented. A persona combination that exists in 0.01% of the real population will not reliably appear in a 500 persona panel. If your pricing question depends on the behaviour of a rare segment, you need either a much larger panel or a targeted generation run.
Worked Example
A SaaS company offers a project management tool and needs to price its new enterprise tier. Four features are under consideration: realtime collaboration, advanced analytics, API access, and priority support. Three price points: £29, £49, £99 per seat per month.
They build a 300 persona panel representing UK based technology professionals. Full factorial design: every combination of features and prices presented as pairwise choices.
The aggregate results show a clear preference hierarchy: analytics is the most valued feature overall, followed by API access, collaboration, and priority support. Willingness to pay peaks at £49 with all four features and drops sharply at £99 with fewer than three features.
That much, a traditional conjoint would also tell you. The synthetic conjoint adds the personality layer.
High Acuity personas value API access 3.1 times more than collaboration. Their reasoning traces reference integration with existing tools, automation potential, and workflow efficiency. They tolerate the £99 price point when API access is included because they evaluate the feature against the cost of building the integration themselves.
High Sociability personas value collaboration 2.2 times more than API access. Their traces reference team coordination, shared visibility, and reduced email. They are price sensitive at £99 because collaboration feels like a baseline expectation, not a premium feature.
High Discipline personas are the most price sensitive group overall, but only when features are bundled opaquely. When each feature has a clear, separate value proposition, high Discipline personas are willing to pay more than average because they can quantify what they are getting.
High Impulsivity personas respond disproportionately to the framing of the £49 tier as the "recommended" option. Their choice pattern shows a strong anchoring effect that barely registers in low Impulsivity personas.
The Practical Recommendation
These personality segments shape specific commercial decisions. The API access tier should lead with technical specifications and integration documentation: that is what high Acuity buyers evaluate. The collaboration tier should lead with team workflow stories: that is what high Sociability buyers respond to. The pricing page should present clear per feature value breakdowns rather than opaque bundles: that converts high Discipline evaluators.
None of these insights come from demographic segmentation. Age and income do not predict whether a buyer evaluates your product through a technical lens or a social one. Personality dimensions do.
The Hybrid Path
The smartest approach is not synthetic or traditional. It is both, in sequence.
Run synthetic conjoint first. It costs under £200 and takes an hour. Iterate rapidly: test five pricing architectures instead of one. Identify the two or three candidates that perform best across personality segments.
Then validate the winner with a traditional conjoint study at a fraction of the usual sample size. You are no longer exploring the full design space. You are confirming a single hypothesis. A validation study with 100 to 150 participants costs a quarter of a full conjoint and runs in half the time.
The synthetic phase handles exploration. The traditional phase handles confirmation. Together, they deliver more insight at lower cost than either method alone.