Suppose you are running a choice-based conjoint (CBC) study for software sold in three plans (Starter, Pro and Enterprise), or for a drink sold in 250 ml, 1 litre and 2 litre packs.
In a standard conjoint design, each attribute varies independently of the others. If you set one price ladder that covers the whole range (say £19, £49, £199 and £499), respondents will see a Starter plan at £499 and an Enterprise plan at £19.
Cards like that cause two problems. Respondents stop taking the survey seriously, and every £19 Enterprise card becomes an obvious pick, which distorts the estimates.
Why prohibitions are the wrong fix
The tempting shortcut is to keep one ladder and add prohibitions: rules that stop Starter appearing at £199 or £499, and Enterprise at £19 or £49. Avoid this for price tiers:
Conjoint relies on attributes varying independently, so the model can separate the value of the tier from the effect of the price.
If Starter only ever appears at £19–£49 and Enterprise only at £199–£499, tier and price move together.
The model can no longer tell whether people turned down Enterprise because they didn't want it or because it cost £499. The estimates become unstable.
Keep prohibitions for combinations that genuinely can't exist, such as a glass bottle in a letterbox-sized flat pack.
How conditional pricing works
Conditional pricing (also called alternative-specific pricing) gives each tier or pack size its own realistic price ladder, while keeping the design balanced.
Worked example: plan tiers
Instead of one price list, you set a price table that depends on the plan attribute, with four price steps in each row:
Starter
Step 1 (low): £19
Step 2 (base): £29
Step 3 (high): £39
Step 4 (stretch): £49
Pro
Step 1 (low): £59
Step 2 (base): £79
Step 3 (high): £99
Step 4 (stretch): £129
Enterprise
Step 1 (low): £249
Step 2 (base): £349
Step 3 (high): £449
Step 4 (stretch): £599
Behind the scenes:
The design works on steps, not pounds. The design balances the plan against the price step (1, 2, 3, 4), so every plan appears equally often at its low, base, high and stretch price.
Every card shows a realistic price. A Starter card at step 2 shows £29; an Enterprise card at step 2 shows £349.
The model compares like with like. The prices still differ a lot between plans, so Ordeen estimates price on the logarithm of the price actually shown, centred within each plan:
$$\tilde{p} = \ln(\text{price shown}) - \text{average } \ln(\text{price}) \text{ for that plan}$$
Centring within the plan removes the difference in price level between plans, so the plan's own value and the effect of price stay separate. Using the logarithm means the model measures sensitivity to percentage changes: a £10 rise means more on a £29 plan than on a £349 plan.
What happens in the results
In Ordeen, conditional pricing works in both CBC and ACBC. In the results, share of preference, willingness to pay, attribute importance, brand-portfolio reach, the elasticity sweep and the price-response curve all use each product's real tier price.
Willingness to pay is converted into pounds at one reference price: the median of all the tier prices.
Related reading: Methodology · CBC vs ACBC · Using a market simulator
Sources
Sawtooth Software, Lighthouse Studio help: conditional pricing in CBC (checked October 2026).
