The Best Way to Predict Consumer Behavior: Add "Nothing" to Your Model


Presidential general elections in the last century were often focused on making a case to swing voters. After primaries ended, both Democrats and Republicans ran to the political middle in order to convince people to choose their side. But that all changed in the new millennium. Candidates don’t care about the ‘winning’ the middle nearly as much as they care about turnout. They want to do anything they can to excite their base of likely voters and get them to the polls. It’s clear that both Obama and Trump are very good at exciting their respective core voters.

Why the shift? Some of it is the “us-or-them” rhetoric that is compelling on cable news or social media. But it’s also the realization that voters are really not deciding between the Democrat or the Republican on Election Day.  They are deciding if they can leave work early to vote, if they need to cook or pick up dinner that night, if the election is close enough to matter, if someone is watching the kids, if any candidate actually excites them, and 100 other factors before actually showing up to the polls and voting.

Why is that? Choices and action take work, and we only have so much willpower to make the choices and actions we need to just get through the day. Anything beyond our day-to-day tasks is difficult to fit into our lives. So when it comes to doing more things, we often would rather do nothing. In the words of comedian John Mulaney, “It’s so much easier not to do things than to do them, that you would do anything is totally remarkable.”

Let’s get academic a bit. From Prospect Theory (that’s right, from Kahneman/Tversky, all you fans out there), we know that there is a bias in favor of the status quo, or the solution that requires less energy or risk. Without capturing that bias, we are likely to overestimate changes in future behavior even when the models themselves are sound.

Prospect Theory sounds a bit much, so we call it “Comfort Bias.” Basically, you have to know the comfortable choice of the action you are trying to predict and bake it into any prediction you make if you want it to be accurate.

When making a consumer choice model like a conjoint, too many researchers are willing to take the winning product and pretend like it’s going to “win” in market without adjusting for this bias.

A better way is to embrace Comfort Bias and work it into your model:

  • Make sure “doing nothing” is always an option in every choice task

  • Define “doing nothing” – it can literally be nothing, but can also be their current solution

  • Pair each choice question with a question measuring the likelihood of each selection made

  • Understand what the default choice will be in-market (e.g. if there are multiple tiers in an online service) and adjust your model to promote that choice.

  • Use past research and in-market results to better calibrate the impact of all of the above factors.

It can be complex stuff, but we have an in-house analytics team (The BrainTrust) that can help you navigate this.  We hope the “comfortable” choice for you is to just drop us a note at and We can help you with the hard choices from there.