Each of 600 simulated shoppers is a random-utility agent with its own price weight wp and quality weight wq, sampled around the population's price-sensitivity setting. For every option k on offer, the agent computes a utility and picks the option with the highest value:
U_k = w_q · quality_k − w_p · (price_k / 60) + ε
ε ~ small Gaussian noise (idiosyncratic taste)
choice = argmax_k U_k
This is the standard random-utility / discrete-choice model behind conjoint analysis and multinomial-logit demand estimation. The Target product is deliberately more expensive but higher quality than the Competitor; on its own, shoppers split roughly along their personal price-quality tradeoff.
The Decoy is built to be asymmetrically dominated: the Target beats it on both price and quality, but the Competitor does not (Competitor is cheaper yet lower quality than the Decoy). A dominated-but-comparable option is hard to evaluate on its own, so shoppers anchor on the easy Target-vs-Decoy comparison — and that comparison makes the Target look like the obvious win, pulling choice share away from the Competitor even though the Decoy itself is barely ever chosen. This is the asymmetric dominance (decoy) effect, documented by Huber, Payne & Puto (1982) and popularized by Dan Ariely's Economist-subscription experiment.
- Decoy price / quality sliders — move the decoy toward or away from dominating the Target; drag it far enough and the effect collapses or reverses.
- Decoy: ON/OFF — toggle the decoy out of the choice set to see the two-product baseline instantly.
- Price-sensitivity slider — shifts the whole population's average price/quality weighting.
- Δ Target — percentage-point lift in Target's share caused purely by adding an option almost nobody picks.