Share of preference is the percentage of respondents in a conjoint market simulator predicted to choose a specific product configuration given a defined set of competing products and prices. It is not a direct forecast of real-world unit or revenue market share. A simulator measures what demand would look like if every buyer had 100% awareness of every product on the screen, every product had 100% retail distribution and stock availability, and zero switching friction existed.

Ordeen's market simulator computes posterior-expectation shares of preference across every respondent and posterior draw — complete with composite 95% credible intervals and paired-draw scenario deltas — so your team can measure the causal impact of price and feature changes while keeping the boundary between preference and physical distribution clear. See the math in our methodology.

Why simulator shares differ from real-world sales data

When a team builds their current market baseline in a conjoint simulator, they often notice that a newly launched challenger brand scores 22% share of preference in the simulator while holding only 6% market share in retail scanner data, whereas an incumbent sits at 30% in the simulator vs. 55% in the real world. Four structural factors explain the gap:

1. Equal shelf presence vs. asymmetric distribution. In a survey choice task, a challenger product sits right next to the market leader in equal font size. In the real world, the market leader may be stocked in 90% of stores or rank #1 on search results while the challenger is in 15% of stores.

2. Instant awareness vs. habit and inertia. Survey respondents actively read the feature specs and prices on every card. Real shoppers frequently grab their habitual default without checking whether a competitor quietly dropped its price by £2.

3. Unmodelled brand equity vs. surveyed attributes. Even when Brand is included as an attribute in the conjoint design, real-world sales are shaped by salesforce relationships, enterprise procurement lock-in, and promotional end-caps that don't appear on a spec card.

4. The Outside Good (None option) calibration. How often respondents click “None of these” in a survey depends on how they were screened and framed, which rarely matches the exact rate at which shoppers leave a store empty-handed.

Avoiding the “Representative Consumer Fallacy” inside the simulator

While no conjoint simulator can observe your retail distribution footprint, many simulators introduce an avoidable mathematical distortion of their own: plugging average part-worth utilities into the logit formula.

Because the multinomial logit (softmax) function is non-linear, the share of the average utility vector does not equal the average of individual shares (softmax(E[β]) ≠ E[softmax(β)]). If half your market strongly prefers a feature and half strongly dislikes it, the average utility is zero — predicting mediocre share everywhere instead of strong capture of a dedicated niche.

Ordeen avoids this at two levels:

• Across respondents: Share is calculated individually for each respondent n (including their personal walk-away threshold u_none,n) and averaged across the sample.

• Across posterior draws: Rather than plugging in each respondent's posterior-mean utility, Ordeen evaluates the softmax inside every saved posterior draw d and takes the posterior expectation across draws, combining parameter uncertainty with finite-sample survey uncertainty to form the 95% credible interval.

How to use Share of Preference for real pricing decisions

If share of preference is not raw market share, how should a pricing or product team use it?

• Read scenario deltas and relative ratios, not raw baseline levels. The power of a conjoint simulator is isolating a controlled counterfactual: holding awareness and distribution constant, what happens to demand when we move our Pro plan from £49 to £59, or when a competitor adds a feature?

• Rely on Paired-Draw Credible Intervals for scenario comparisons. When you compare a proposed scenario against your Market Baseline in Ordeen, both scenarios are evaluated on the exact same paired posterior and bootstrap draws (Common Random Numbers). Shared baseline noise cancels out, giving you a tight 95% credible interval on the change in share (Δshare). Ordeen's Movers view filters out products whose 95% delta interval straddles zero, so you only act on statistically verified shifts.

• Check where lost buyers go (Substitution vs. Walk-Away). When a price increase drops your share of preference by 4.0 percentage points, check how much of that 4.0pp flowed into the None column (category exit) versus specific rival SKUs.

Related reading: Using a market simulator · What is Nested Logit and IIA? · What is willingness to pay? · Full methodology

Want to test your product line-up in an interactive market simulator? Book a demo or email im@ordeen.com.