Deep Dive: User Types
The user you pick shapes every answer that follows. A generic choice produces generic problems and generic solutions. A too-specific one limits your creativity for the rest of the problem.
This skill comes in very handy when the question prompt is less defined.
At OpenAI, product sense prompts are frequently intentionally under-specified. One candidate who went through the loop described "a brutal product-sense interview with an under-specified memory-machine prompt and almost no feedback."
The lack of structure is part of the test. When you commit to a specific user, the rest of the answer follows: the pain point is theirs, the feature is built for them, and the success metric reflects whether it actually worked for them. When you stay generic, everything downstream is generic too.
What interviewers are scoring
- Are the segments genuinely distinct, or do they overlap?
- Are they specific to what this technology enables, or could they appear on a list for a completely different product?
- Does the prioritization show judgment, or did you just pick the biggest group?
- At Senior+: Does the choice reflect company strategy, not just user need?
Generate your segments
Step 1: Start with people, not categories
Ask: who would actually benefit from this product? Not in the abstract. Picture real people whose lives this changes. For the animal mind-reading technology, that might be a dog owner anxiously checking their phone at work, a vet trying to assess a patient who cannot describe pain, or a conservation researcher watching an elephant herd from a distance. Let those people surface naturally before you start organizing them into buckets.
Step 2: Give them names
Once you have a sense of who they are, name them. Not labels , names. The Worried Dog Owner. The Field Researcher. The Conservation Scientist. This is not a stylistic flourish. Naming your user types forces you to actually picture a person, which makes the pain points you surface in Step 4 feel specific and earned. Interviewers notice the difference between a candidate who says "pet owners" and one who says "the Worried Dog Owner who has no idea whether their dog is calm or distressed while they are at work."
Step 3: If you're stuck, use a lens
If real people are not coming to mind, four lenses can help generate candidates:
By use case: what is the user trying to accomplish? Maps to intent rather than surface attributes.
By product usage stage: new vs. casual vs. power users. Useful for growth-stage products; less useful when the product is novel and no one has usage history yet.
By demographics: age, life stage, professional role. The most familiar lens and the most overused. "Kids, young adults, elderly couples" could apply to any product. If your segments could belong to anything, go deeper.
By consumption behavior: what kind of value they want to extract. For media: movie watchers vs. episodic TV viewers vs. old-catalog browsers.
The sharpening test
After generating your segments, ask: would any of these groups appear on a segmentation list for a completely different product? If yes, they are still generic.
There is also a ceiling on the other side. A segment that is too narrow effectively pre-solves the pain point. When you get to Step 4, there is only one problem left to name. A useful rule of thumb: your segment should be specific enough that you can picture their daily rhythm, but not so specific that it describes just one part of their day. "Dog owners" works: you can imagine their mornings, their commutes, their walks, their evenings. "Dog owners dealing with separation anxiety" has already named the problem, and Step 4 collapses.
Prioritize one
Score each segment against three criteria: breadth (how many people have this problem), depth (how acute the pain is), and mission (how well it fits what this company is trying to do). You do not need to work through every cell. Glance at the scores, say which segment is standing out and why, and commit. The goal is a clear, reasoned choice. Not a full analysis.
A competent answer produces segments that are distinct and sensible, with a prioritization choice backed by clear rationale (typically the largest or most accessible group). The logic holds up.
A Senior+ answer goes further. Segments are specific to what this technology enables. Prioritization names a criterion tied to the company's mission or the technology's actual constraints, and states what the chosen segment gives you that others do not.
Running example: animal mind-reading at OpenAI
The prompt: "Our researchers have developed new technology that can read the minds of animals. How would you think about what to build?"
After clarification: vague emotional states only: calm, distressed, engaged. Requires a wearable sensor.
Three candidate segments:
- Pet owners trying to understand a companion animal
- Researchers and veterinarians needing ground-truth cognitive data
- Conservation organizations tracking animal wellbeing
Apply the sharpening test. All three could appear on lists for different products. "Pet owners" is closer: it is tied to companion animals. But still too broad. "Dog owners" is right. Specific enough that it reflects what the technology actually enables (a wearable emotional signal for a daily companion), broad enough to surface multiple distinct pain points in Step 4: separation anxiety, uncertainty about whether the dog is genuinely happy, confusion about whether training is working or not.
"I am going with dog owners. They have a daily, intimate relationship with an animal whose emotional life is largely invisible to them. Vague emotional signal (calm versus distressed) is sufficient to be genuinely useful here, without requiring precision the technology does not yet have. It maps to OpenAI's mission of extending understanding to sentient beings who cannot speak for themselves, and it gives us fast, observable feedback on whether the core signal is trustworthy."
Common pitfalls
Generic cuts. "Casual users, regular users, power users" describes engagement levels, not people. It tells the interviewer nothing about this specific product.
Picking the largest group without a criterion. The biggest group is often the hardest to design for. Their needs are the most diffuse. Name why large and right are the same thing here.
Leaving the segments open. Pick one. If you keep all three alive, your pain points will be as generic as your segmentation.
Weak prioritization. "This group seems like the right fit" is a missed opportunity. Name the criterion, name what this segment gives you that others do not, and tie it to something specific about the company or the technology.
Doing some research on the company you are interviewing at goes a long way.
A Shopify PM candidate noted the product sense interview "was specific to Shopify's business line, so I had to understand their user segments."
Knowing the company's actual users before you walk in is not optional at this level.