Prospect Theory in Action
How to use Prospect Theory to improve product positioning
Intro
This post summarizes some of the research that was led by Amos Tversky and Daniel Kahneman around decision making. These two psychologists from Israel have transformed the field of economics. Their outsider perspective was required to up-end an industry steeped in traditional thought, similar to many disruptions. They challenged the thinking that people make decisions that align with economic models. Through the process, they also made it popular to run experiments to analyze behavior — which is now commonplace in User Research and Product Development organizations around the world.
I’ll start with an overview of a few of the key research outcomes they discovered through their careers. I will share examples from their work and then turn the attention towards how to apply their work towards developing products.
The Frameworks
The below summarizes a few of the key theories/frameworks that stood out to me from the work that Tversky/Kahneman led in their careers.
System 1 vs. System 2 Thinking
A fundamental part of their theories is around the idea of System 1 vs. System 2 thinking. System 1 refers to automatic decision making and 2 refers to reflective thinking. System 1 is the fast reaction that you have to a situation and System 2 is when you are required to think. Driving is a System 1 for most people being able to quickly respond to situations. System 2 is used for solving harder problems, which could be driving in snowy conditions or doing math. Below is a question I like to ask during interviews (that comes from Tversky/Kahneman’s research):
A bat and ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?
- If you are like many people your system 1 response is $0.10
- If you have a strong system 2 checks in place you will do some quick math and realize that isn’t possible. You will eventually get the right answer. Though very few people have a system 1 that would spit out the right answer.
This insight from Tversky/Kahneman is critical to understanding the results of their work. Here is a great quote on what to do with this work:
The best we can do is compromise: learn to recongise situations in which mistakes are likely and try harder to avoid significant mistakes when the stake are high. — Daniel Kahneman, Thinking Fast and Slow (page 28)
This is great advice for product and business leaders. We often have an initial reaction (system 1) to a situation. Many times these are small decisions, but for larger decisions we should be aware of how the mind works.
Heuristics
While not technically part of Prospect Theory, this research led Tversky and Kahneman to go deeper into the psychological reasoning of unreasonable decisions.
- Representativeness: Captures the probability that an individual belongs to a specific group.
Which of these sequences is more likely to occur if you were to flip a coin, H-T-H-T-H-T or H-H-H-T-T-T? The answer is both are likely given Heads has a 50% chance.
The other day I was at the gym and an athletic black guy that was probably 6' 8" was on the basketball court. Is he more likely to be a (current or former) pro basketball player or a school teacher? There are 3.6M teachers in the US and each year there are ~3000 current and former pro basketball players. Thus there is a 1200x likelihood to be a teacher. In the research from Tversky/Kahneman they also added a factor like a school teacher that was a former college basketball player. People would often say it was more likely that the individual was a school teacher and former college player vs. just a teacher. This can’t be true as the basketball player component is a subset of the overall teacher population.
- Availability: This is the bias that human tendency thinks of examples that come readily to mind are more representative than is actually the case.
How likely is your house to be broken into? If you are set up with Nextdoor alerts you would think there is a high likelihood. This may then inspire you to buy a security system.
If you live in an affluent area you are likely to assume that everyone is affluent. This also plays out in the “keeping up with the Joneses” idea.
- Anchoring: is a cognitive bias where an individual relies too heavily on an initial piece of information offered
In their research they shared an example of asking people to estimate the product of a set of numbers:
- 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8
- 8 x 7 x 6 x 5 x 4 x 3 x 2 x 1
In this comparison, people who were shown the first set of numbers had a median estimate of 512 vs. 2,250 for the second set. The actual result is 40,320. This type of anchoring applies to many other scenarios. If I told you that the average cost of a concert ticket was $100 and then told you that our favorite band was playing next week and tickets were only $50 you would be more willing to go than if I said the average price was $10. The price of the show didn’t change, but you think it is a $50 savings in the first situation vs. a $40 premium in the second.
Value / Loss Function
Here is a quick excerpt on the summary of the value (also known as the loss) function that Tversky and Kahneman discovered through their research and published in their paper on Prospect Theory.
In summary, we have proposed that the value function is (i) defined on deviations from the reference point; (ii) generally concave for gains and commonly convex for losses; (iii) steeper for losses than for gains. A value function which satisfies these properties is displayed in [the figure below]. Note that the proposed S-shaped value function is steepest at the reference point, in marked contrast to the utility function postulated by Markowitz which is relatively shallow in that region.
There are a number of valuable insights that stem from this model that can be applied to many scenarios where decisions are not straight forward. A great example from their research was around how companies fund growth initiatives, like the options below.
- Project A: 50% chance of success to deliver $10M in value or 50% chance to lose $5M in values
- Project B: 90% chance of success to deliver $1M in value or 10% chance to lose $500k
The expected value of A = $5M whereas B = $650k. Clearly, a company would want to fund lots of A investments.
The research aligns with this theory when discussing the decision with a CEO who has multiple initiatives to pursue. Though many managers would choose B as there is a higher likelihood of success (or said differently a lower likelihood of failure). Most companies have more extreme punishment for failure than success. Companies can change their performance plans to address these issues.
Weighting Function
The last major aspect I’ll touch on from the research of Tversky and Kahneman is their Weighting Function to assist in decision making. This research is really profound. The table below shows the probability of an event occurring and the decision weight that someone places against it. People put a lot of value on certainty, as can be seen in the table. This insight can be a key driver for creating value in products. In a previous post, I shared the OpenDoor model and how they can leverage certainty to maintain high margins.
Security system example. There is a less than 2% chance that your home could face a loss due to burglary and if so would cost you ~$2000. The economic argument is that you’d pay up to $40 per year to see the same expected value (2% x $2000). Though using decision weights the value is 4x higher, meaning you’d be willing to pay $160 per year to have a 0% probability. This value is aligned with the $10/month plans that Amazon recently announced via their Ring Security system, though is much lower than the $30-$50 per month that ADT and others charge. It must be that homeowners either are weighing the economic impact higher or they view the likelihood higher. Assuming it was likelihood you would need to think you had a 20% likelihood of an event occurring or that damage was as high as $8000.
Framing. I won’t cover it here but will discuss the ideas in a future post. It is important but adds too much to this single post.
Examples in the Real World
The home sales process is a great example of many of these theories. Most sellers will not have a pre-inspection done to know what may be found. This is due to an obligation to fix something they know will happen.
This is not too different than some of the scenarios Tversky and Kahneman ran. Here is the scenario for a homeowner looking to sell:
- Option A: Pay $500 for a home inspection, then a 10% chance you have to pay $10,000 to fix the issues, and finally a 90% chance to close on time
- Option B: Don’t have a home inspection, then a 10% chance the buyer asks you to pay $10,000 to fix the issues, and finally a 50% chance to close on time
In Option B, the seller has a level of certainty around the pay off to make the fixes and avoids the certainty of losing $500 on a home inspection, though this comes at the cost that the buyer may walk away from the deal.
The rational argument would be to do the pre-inspection, but due to the loss potential of the scenario, the seller is likely to be risk seeking (as seen in the table of scenarios above.
Applying Prospect Theory
There are many ways to learn from Prospect Theory. I have started to use these frameworks to guide User Research and to stretch the thinking of the team.
Loss framing. If you know the decision for the individual and their alternatives you can frame the decision to align with your goals. This understanding also helps you to understand the opposing forces that you may face to launch a new product.
Add certainty to ambiguous situations. People are willing to pay a premium for certainty. The benefits to ensure certainty may outweigh the costs to deliver certainty, as was the case for OpenDoor and a host of clones. I would argue that Uber/Lyft have benefited from this effect as well. The whole on-demand industry is essentially looking to create certainty out of ambiguous situations.
Make Products Free. There is a massive difference between a minor cost and free. The decision weights and loss function show that customers will place more than a 1x1 rate on these products.
Conclusion
Tversky and Kahneman were pioneers in ushering in behavioral economics. There are still so many areas where their learnings can be applied to user research to develop killer products.
I am much more aware of these principles now and it has reshaped the research agenda I have around our product strategy to both prove and disprove various product hypotheses we have.
This post is part of a series of posts that I will do in 2019 to summarize various articles/books from leading individuals in the field of economics.
