The one time it’s morally acceptable to shove a fat guy

This week’s newsletter is longer than normal. If you just want the practical “steal this” survey question then feel free to skip to the bottom.

You know the “Trolley Problem”?

The thought experiment where you can pull a lever to divert a runaway trolley and save 5 at the cost of 1, or do nothing and 5 people die?

Debating the lever is fun.

But I like the scenario itself.

Especially the variants, like “The Fat Man,” where pushing a large man onto the tracks saves the five others, and “The Self-Sacrifice," where you can divert the trolley towards yourself.

Good survey questions function similarly. 

They reveal our motives and values, simply by being read.

Although shopper insights are typically extracted from spreadsheets, a survey question can expose truths before it collects a single response.

I hadn’t thought about the “Trolley Problem” since college, where I encountered it first. 

It came to mind last week while I answered a poll about a proposed amendment to the New York City Charter for primary elections.

The change would let any registered voter participate in any party's primary—a shift from the current system where only party members vote in their primaries.

What I found interesting was not how the survey measured my opinion but how it used various scenarios to scrutinize it.

One suggested the amendment could make New York elections fairer, which increased my support. But another warned that it might harm working families, prompting me to withdraw my support.

My view shifted yet again after learning it could benefit the opposing political party, leading me to select, “I strongly oppose the amendment.”

Do I care about fairer elections and working families?

Sure.

But mostly, I just want my side to win.

If the survey had simply measured my opinion, it wouldn't have uncovered this blunt truth.

However, by crafting scenarios that critically examined my opinion, it revealed not just what I thought, but why and what could change my mind.

You see this approach more in political polls and focus groups, where skilled moderators engage participants without pressuring them — but it’s rarer in customer surveys, which emerged from a tradition that prioritizes neutrality and precision.

The idea isn’t to outsmart people.

It’s to find inconsistent beliefs so you can ask questions like:

In the previous section you said so-and-so but just now you said such-and-such. Can you explain why?”

Valuable insights emerge when conflicting beliefs are pitted against each other -- the friction reveals the mental fault lines that guide our decisions.

I could go on, but let’s get more practical and get to an example.

 

Steal This
Suppose you’re researching online streaming platforms like Netflix and HBO Max. Here’s a sequence of questions you could use to understand “subscription fatigue,” with the key insights coming from Q3. 

 

Q1: How many streaming platform subscriptions (e.g., Netflix, Hulu, Disney+) do you currently have?"

Response: _________


Q2: Do you ever feel like you have too many streaming platform subscriptions?

• Yes
• No
 

Q3: If you feel like you have too many subscriptions, what's stopping you from canceling one or more of them?
Response: ____________________

 

For the record
I’d pull the lever but hesitate to push the fat man, unless that fat man was me, in which case I’d take the exact opposite position.

If it were my two children tied to the tracks, I’d sacrifice myself.

That’s an easy decision.

Next steps
Speaking of easy decisions: if you're considering creating a survey, but you're unsure about your approach, then consider getting a Survey Roast.

Send me your survey draft, and for $145, I’ll make a 15 minute Loom video with copy-and-paste edits and suggestions.

I’d love to help.

https://www.sammcnerney.com/survey-roast

Cheers,
Sam

Interested in more weekly research insights and survey tips like this one?
Click here to subscribe.

Previous
Previous

How to ask shoppers why they choose one brand over another

Next
Next

Want to add some quant rigor to a qual-heavy segmentation?