“A little learning is a dangerous thing.” - Alexander Pope
In March 2018, United Airlines made headlines.
Company managers decided to introduce an alternative incentive scheme for employees. They decided to change from the fixed bonus to a lottery system:
Instead of all employees receiving a moderate but not insignificant amount, the lucky six (decided by the lottery) would get prizes like $100,000 cash, a Mercedes-Benz C-Class Sedan, or a tropical vacation. And the rest would get nothing.
Most employees were not happy.
Backed up by their union they revolted against it.
Management was forced to go back to the original scheme within days.
The basis for management's decision
There is a good reason why a chance of winning something of a bigger value might be a stronger incentive than the certainty of something of lesser value: we tend to overweigh small probabilities. A 1/100 chance of winning $100 appears more attractive than a certain $1.
For example, traffic at Dishoom café in London was low before 6pm on weekdays. To solve for that, behavioral scientists suggested they introduce a lottery:
If you pay your bill by 6.00pm Monday to Thursday you get to roll a die and if you throw a six you get your whole meal for free.
To an economist, that's identical to a 16.6% discount to all customers.
But to the human brain it is not identical.
A 16.6% discount is boring and doesn't encourage you to do anything.
Whereas a 1/6 chance of getting your meal for free is really, really exciting. Adrenaline kicks in. You want to try your luck.
Plus, the lottery encourages you to spend more because you don’t want to miss out on getting a bigger treat if you skimped on everything and then ended up getting it for free. What a disappointment that would be.
This intervention was very successful for Dishoom...but not so much for UA.
What UA management missed
So yes, lotteries have been very useful in many contexts. The keyword here is: context.
If management (or their advisors) had dug a bit more they would have found cases where lotteries backfired compared to traditional (fixed) incentives.
Moreover, specific aspects of the scheme were red flags to begin with. For example, perfect attendance was required to be eligible for the new bonus system, which was unfair to single parents who may have to take care of a sick child at a moment's notice.
What could management have done?
Human nature is messy.
We can’t change that.
What we can do is change our approach to it:
Ask, don’t assume. Use mixed methods of inquiry to explore the context: narratives, self-concepts, and cognitive processes related to the behavior(s) we’re designing for.
Don’t assume tendencies (like the overweighing of small probabilities) apply unconditionally, and don’t overgeneralize findings from one context to another one, and blindly throw the most common or exciting ones into our product or program.
Develop intervention concepts that fit your context based on your findings, and test on a small scale before launching full scale.
This shouldn’t sound simple because it’s not. But if you get yourself an excellent behavioral scientist, it's very possible—and can revolutionize your outcomes.
The number 1 rule in behavioral science, as UA found: Context matters.
Question for you
What is one change you can test first before applying it more broadly, so that you make sure it helps solve your bigger puzzle?