What the decoy-effect research actually says about "good-better-best" pricing pages

Almost every SaaS pricing page uses some version of a three-tier "good-better-best" structure, and the usual justification is "the decoy effect" — the idea that adding a deliberately unattractive middle option makes a more expensive option look like the obvious choice. The research behind that idea is real, but it's older, narrower, and more fragile than the pricing-page folklore built on top of it.

The foundational work is Huber, Payne and Puto's 1982 paper on "asymmetric dominance," which showed that adding a third option that is clearly inferior to one existing option (but not comparable to the other) increases the choice share of the option it's dominated by. The classic demonstrations used simple, easily comparable attributes — things like beer priced on quality and cost, or cars on mileage and price — not the multi-dimensional, feature-heavy comparisons typical of real SaaS pricing tiers.

Two things matter for anyone trying to apply this to a real pricing page. First, the effect depends on the decoy being easily, directly comparable on the same attributes as the target option — it works because the comparison is effortless, not because the option merely exists. A middle SaaS tier with a confusing mix of different features than the tiers on either side of it may not create a clean dominance relationship at all, and can just as easily confuse buyers as nudge them. Second, later replications have found the effect is considerably weaker, and sometimes absent, when real money and real purchase stakes are involved rather than hypothetical choice tasks in a lab.

The practical takeaway isn't to abandon three-tier pricing — it remains a reasonable default because it typically reduces the number of comparisons required per decision — but not to assume the decoy tier is doing more psychological work than the evidence supports. If you want a middle tier that nudges people upward, keep the comparison to the top tier simple and direct on the attributes buyers actually care about, and test the effect on real purchases rather than assuming a 45-year-old lab finding transfers automatically to your product.