there are a lot of parallels between what researchers and product managers do.
i noticed this because my roommate colin, who comes from an ai research background, seems to be really good at product. i believe in him and think he will be very successful, which requires belief in his ability to build a really big consumer product, which requires product skills of course.
on the surface level, research may be thought of as slow and theoretical, whereas product is often thought of as the opposite (fast and pragmatic). but on a more fundamental level:
- they both focus on truth-seeing
- they both rely on a similar framework: hypothesis → experiment → result → interpretation
you can approach product-building like a scientist - every launch is an experiment, and getting to product-market fit involves running controlled trials to figure out what people want
breaking it down a bit more:
hypothesis selection involves taste
while great product managers / researchers have strong analytical thinking skills, there is a certain amount of intuition involved at the very beginning of the process (hypothesis → experiment → result → interpretation)
intuition helps you decide where to look and rationality helps you decide what to keep
choosing the right question to ask in the first place is not a trivial task, that involves some taste, rather than pure logic. you’re trying to make a bet on what the truth is, or where signal may exist:
- can we make this model better with less compute?
- is this pain point strong enough that people will care / switch their behavior?
deciding what not to build is important. you can be a great engineer, but judgement is a very helpful skill to have as you advance your career in research / product. this judgement is hard to describe and doesn’t sound very scientific, but it can determine the quality of your experiments (consistently starting with wrong / uninformed hypotheses can waste a lot of time)
the experiment loop
on to the quantitative / scientific part: after the question is selected, you need to run experiments, measure the results, and repeat until you find the truth (about the universe, or about human behavior). these experiments (or in product, a/b tests) have to be repeatable (or in research, the term would be reproducible)
in the experimentation phase, speed matters a lot. unfortunately, experimentation speed can be harder to achieve on the research side as opposed to the product side. in the earliest stages of a product or company, speed matters a lot (the more assumption you can validate / invalidate in a short amount of time, the higher the chances of survival are)
the review process
once you ship something (a paper, or a product), you can outsource the review to other people. in research, you publish something and get peers to review. in product, you ship something and get your users to give you feedback. in both cases, you may have to repeat this a lot if you’re not where you want to be.
in both research and product, if you’re really seeking the truth, you have to be honest in evaluating results. it’s easy to cherry-pick the results you want to see, but that won’t get you closer to the truth. when the experiment fails, you can’t be emotional about it
researchers can make for great product people
researchers who enjoy real-world applications of their theories (and enjoy talking to people) already have a lot of good product dna. you have to structure ambiguity, run a bunch of experiments, and interpret results that might look messy
again, you have to enjoy talking to people and have some level of empathy (important for understanding human behavior). this empathy can be tough to learn if it doesn’t come naturally
if you combine researchers’ truth-seeking nature with product-builders’ empathetic tendencies, and apply these skills towards founding a startup, the persona you end up with is a builder who ideally has the capacity to learn faster than anyone else, and iterate into product-market fit
you could think of “truth” or “information” as the “product” that researchers are trying to reach
or you could think of great product people as “researchers in human behavior”