Decoy Effect

011 min

The decoy effect at a glance

• What it is: A choice between two options changes when a third option that loses to only one of them is added.
• Other names: Asymmetric dominance effect, or attraction effect.
• Who defined it: Joel Huber, John Payne and Christopher Puto, in the Journal of Consumer Research, 1982.
• When it works: Buyers are unsure between two options, and the comparison is easy to see.
• Decision it changes: Which tiers a pricing page shows, and whether an unsold tier should stay.

022 min

How the decoy effect works

Every decoy setup has three roles. The target is the option the seller wants people to choose. The competitor is the other real option. The decoy is a third option placed close to the target but worse than it on every attribute buyers compare.

The precise term is asymmetric dominance. One option dominates another when it is at least as good on every attribute and better on at least one. The target dominates the decoy. The competitor does not, because the competitor is better on one attribute and worse on another. That difference between the two relationships is the asymmetry.

An illustrative case makes this concrete. A subscription app sells two plans, and then adds a third:

Without Pro Lite, a buyer must decide whether 15 extra projects are worth $10 more. That is a real trade-off with no obvious answer. With Pro Lite on the page, the buyer also sees a comparison that needs no thought. Pro costs $1 more than Pro Lite and includes twice as many projects. Pro now has a clear reason behind it. Basic has no such reason, because nothing on the page is clearly worse than Basic.

The decoy does not change what Pro or Basic offer. It changes which comparison is easy to make. Buyers who were close to indifferent between the two real plans use that easy comparison to settle the choice. This is why the effect depends on buyers being unsure. A buyer who already knows they need only five projects will ignore Pro Lite.

The mechanism also works in the other direction. A $12 plan with 4 projects would be dominated by Basic but not by Pro, so it would push buyers toward Basic instead.

032 min

Where the decoy effect comes from

Joel Huber, John W. Payne and Christopher Puto described the effect in 1982, in a Journal of Consumer Research paper titled "Adding Asymmetrically Dominated Alternatives: Violations of Regularity and the Similarity Hypothesis". They were not trying to invent a pricing tactic. They were testing a rule of choice theory called regularity. Regularity says that adding a new option to a set can never raise the share of an option already in it. The rule is built into formal choice models such as those of the mathematical psychologist R. Duncan Luce. Luce had written in 1977 that regularity was the only axiom of rational choice theory that had not been violated.

University students chose between pairs of products in categories such as beer, cars, restaurants, films and television sets. They then chose again with a decoy added next to one option. In every category except lottery tickets, adding the decoy raised the chance that the target was chosen.

In the original studies, the decoy itself was chosen only about 2% of the time. That small number matters. It shows that people saw the decoy was worse, so the target's gain did not come from confused buyers.

The paper broke a second assumption too. The similarity hypothesis says a new option takes most of its share from the option most like it. The decoy was most like the target, yet the target gained share. The effect is also a violation of a decision-theory rule called independence of irrelevant alternatives: an option nobody chooses should not change the choice between the others.

042 min

A decoy effect example: The Economist

The best-known example comes from the behavioral economist Dan Ariely. In a 2008 TED talk and in his book Predictably Irrational, he described a subscription offer on The Economist's website. It listed three options:

Print-only is the decoy. Print and web costs the same and includes everything print-only has, plus web access. Web-only is not dominated by print-only, because it costs less than half as much.

Ariely gave the offer to 100 MIT students. With all three options on the page, 84% of Ariely's students chose the $125 print-and-web subscription. Nobody chose print-only. He then removed print-only and gave the two-option version to a different group of 100 students. With the print-only option removed, 68% of students picked web-only and 32% picked print and web.

Removing an option that nobody chose reversed the result. For a seller, that is the difference between most buyers paying $125 and most buyers paying $59.

Two limits apply to this case. It was a classroom demonstration with hypothetical choices, reported in a talk and a popular book, not a peer-reviewed study with real payments. And Ariely says the offer disappeared from the website after he phoned The Economist. So the case shows the shape of the effect, not a measured change in The Economist's revenue.

051 min

A decoy effect case on a real shelf

The Economist case used hypothetical choices on paper. A field study tested the effect with shoppers spending their own money. Huber, Payne and Puto describe it in their 2014 review. In a 1999 field study by Doyle and colleagues, Heinz baked beans priced at £0.31 served as a decoy for the same Heinz beans priced at £0.29.

The setup was the simplest decoy possible. The decoy was the same can of beans at a higher price, so the target was better on price and equal on everything else. The competitor was a lower-price, lower-quality option. With the higher-priced cans on display, sales of the £0.29 target rose relative to that competitor.

The variable that differs between the two cases is the stakes. Ariely's students made a hypothetical choice. The shoppers in the Doyle study made real purchases. Together the two cases show the effect in a classroom and in a shop.

The shelf case also shows why retail decoys are often the same item on worse terms. That form makes the dominance impossible to miss, which is one of the conditions the effect needs. Huber, Payne and Puto give a second retail case of this kind. In a display of oranges sold at one price each, the noticeably smaller oranges would work as a decoy for the larger ones.

062 min

How the decoy effect shows up in tech

Most software products sell through a pricing page with two to four tiers. That page is a choice set, so the decoy effect is part of its choice architecture, the way options are arranged and presented. Several roles make decisions that it changes.

• Product managers decide how many tiers to show and what each includes. A tier that almost nobody buys is not automatically waste. Before removing it, check whether it makes a neighbouring tier look better, as print-only did for The Economist.
• Designers decide the layout. Huber, Payne and Puto (2014) note that small formatting choices matter. One is whether the decoy and the target sit next to each other. Another is whether their values are easy to compare. A comparison table with aligned rows makes dominance visible. Limits hidden behind tooltips make it invisible.
• Founders set the prices. A tier priced just above a clearly weaker tier sends buyers to the stronger one. A tier priced far above everything else works through a different mechanism.
• Engineers and growth teams build the experiment. A decoy is a testable prediction: the target's share should rise when the decoy is present. That calls for a proper A/B test, not a comparison of this month against last month.

Huber, Payne and Puto suggest that online shopping may make dominated options less rare. They point to Amazon, where almost every product shows a price and a 1–5 star rating. When every option is described by the same two numbers, one option can clearly dominate another. Comparison grids on software review sites have the same structure.

072 min

How to use the decoy effect on a pricing page

The steps below follow the conditions Huber, Payne and Puto describe in their 2014 review of the evidence. Each step depends on the one before it.

1. 1

   Pick the target first.

   A decoy helps only one option. Decide which tier you want chosen more often, usually the one with the best margin or retention.

2. 2

   Check that buyers are unsure.

   The effect works when buyers are close to indifferent between the target and the competitor. If one tier already takes most sales, a decoy has little room to change anything. Look at current tier shares before you design.

3. 3

   Build the decoy to lose to the target on every attribute buyers compare.

   Usually that is price plus one headline limit. The decoy should cost slightly less than the target at most, and be clearly worse. If it beats the target on anything some buyers value, those buyers will choose it, and it will take share from the target.

4. 4

   Make the comparison easy to see.

   Place the decoy next to the target, with the same attributes in the same rows. Pettibone (2012), as summarised in the 2014 review, found that the effect disappears when people do not have time to assess the information.

5. 5

   Place the decoy where your buyers are looking.

   Heath and Chatterjee's 1995 meta-analysis, cited in the same review, looked at where decoys sit. High-price, high-quality decoys work better with buyers who want and can afford high quality. Low-price, low-quality decoys work better with buyers on limited budgets.

6. 6

   Test it, then keep watching the decoy's own sales.

   If the decoy starts to sell, it has become a real option, and the design needs to change.

Step 3 has no meaning until step 1 is done, because a decoy is only worse relative to a chosen target. Step 4 is wasted if step 3 produced a decoy that is not clearly worse.

081 min

How to measure the decoy effect

The standard measure is the target's choice share with and without the decoy. Show one group of buyers the two-option set and another group the three-option set. The decoy effect is the difference in the target's share between the groups. In Ariely's demonstration, print and web went from 32% of choices to 84%, a difference of 52 percentage points.

Two checks keep that number honest:

• The decoy's own share should be close to zero. Huber, Payne and Puto (2014) argue that a clean test needs a decoy almost nobody picks. If many buyers choose the decoy, it is competing with the target, and the measure mixes two effects.
• Track revenue per visitor as well as share. A decoy can raise the target's share and still lower total sales. This happens if the extra tier makes the page harder to read. Measure paid conversions and revenue per visitor for both versions.

A second method measures money instead of shares. Paolo Crosetto and Alexia Gaudeul (2016) measured how much more a person will pay for the target over the competitor. They compared that amount with and without the decoy present.

091 min

Common mistakes with the decoy effect

Trap 1: teams assume any expensive third tier is a decoy. A decoy must be clearly worse than the target. A top tier that costs more and offers more is not dominated by anything. It may still change choices, but through a different mechanism with different rules.

Trap 2: teams assume the effect moves every buyer. Huber, Payne and Puto (2014) list five conditions that weaken the effect: strong prior trade-offs, a dominance relation that is hard to see quickly, buyers whose values differ widely, a decoy that people strongly dislike, and a decoy that people actually like.

Trap 3: teams assume a real product can be a pure decoy. Real plans differ on many attributes. A cheaper plan with fewer projects may include one feature some buyers want, so for them it is not dominated at all.

Trap 4: teams add tiers without limit. Every extra tier is one more option to read. Past a point, more options make a decision harder, as described under the paradox of choice. One decoy next to the target is the usual design.

Trap 5: teams remove the tier nobody buys. A tier with almost no sales may be the reason the target sells. Test the removal before making it.

102 min

Where the decoy effect evidence is contested

In 2014 the Journal of Marketing Research published two papers that questioned how useful the effect is outside the lab. Frederick, Lee and Baskin (2014) reported that the effect appeared when the options were described with numbers, but not otherwise. When the same choices were shown as pictures or described in words, the effect weakened and sometimes reversed. Yang and Lynn reported 91 attempts to produce the effect, with only 11 reliable effects. The marketing researcher John Dawes summarised both papers in 2016 under the heading "The Attraction Effect debunked".

The original authors replied in the same journal issue. Huber, Payne and Puto accepted that the effect has boundary conditions. They argued that it still appears whenever the conditions of the original studies are repeated. They pointed out that Frederick, Lee and Baskin's own studies with numeric gambles largely reproduced it.

The reply also made two concessions. In a commercial conjoint data set, Huber could not detect any consistent increase in the target's share when it dominated a decoy. That data set covered 586 respondents who each made 20 choices, and the authors think practice gave them fixed decision rules that the decoy could not move. The authors also wrote that they suspect the effect "occurs rarely in the marketplace today, at least in its strict regularity-violating form." Their reasons are that real products have many attributes, and that sellers stop producing options nobody buys.

This leaves a practical position. The effect is most likely where buyers compare options on a price and one or two numbers, and have weak preferences. A pricing table often fits that description. It is weak or absent when products are rich, visual and personal. Treat any decoy as a claim to test on your own buyers, not a guaranteed result.

111 min

Decoy effect vs. compromise effect and anchoring

The nearest neighbour is the compromise effect. The deciding question is whether the added option is dominated. In the decoy effect, the added option is clearly worse than the target. In the compromise effect, the added option is not worse. It is an extreme, such as a much more expensive, higher-quality plan, which turns the target into the middle choice. Buyers who avoid extremes then pick the middle. Huber, Payne and Puto (2014) note that the decoy effect is stronger in lab studies, but sellers use the compromise effect more often in real shops.

Price anchoring is a third, separate mechanism. An anchor is a number seen first, which later prices are judged against. It does not need a third option at all.

?8 questions

Questions people ask

§9 sources

Sources