Do you know of any frameworks, standards, statistical models or rules of thumb that help prescribe the number of people that should be interviewed for customer research?
It depends on what kind of research you’re doing. If you’re doing usability tests/interviews, usually 5 or so will suffice. Here’s a resource from Nielsen Norman Group that I’ve found helpful:
We always do five, and if there’s a common pattern, we’re good. If not, then we will interview a few more. I believe the standard is five because, after that, you start to hear similar info.
Any framework standard or statistical model that you would apply for customer research depends entirely on your research question and how you hope to use any insight garnered. The way in which you’re asking this question implies that you have never actually worked with customer researchers or understand how to participate with them from the seat of product management.
My advice to you is get on LinkedIn look through your network and one degree of separation and see if you can get 30 minutes with somebody who actually does customer researcher or leads a customer research group and learn about how to participate with them and what’s important from the seat of product management.
The details you are asking for here are usually structured by the researcher your partnered with. If you’re company doesn’t value customer research enough to provide access to customer research in-house or through contractors and you have to do this yourself best of luck to you.
There are some tricks for qualitative research that’s non-generalizable if you want those reach out with a direct message.
@MichaelYoffe, I’m doing this as part of a three person startup designing a physical product with simple user interface. So looking for a starting point as to key words to google and beginner resources and books to read.
It reminds me of looking for a function in Excel you know it can do, but having no clue what it’s called so I can’t google it!
For physical products, it’s more important to get the right set of participants for your intended purposes.
5/95 percentile for ergonomics: can it be used by the majority of the population? For instance, if it is a wearable, are you testing with the 5-95% of the wrist circumstances of the target population? Look up human factors/scale design references. Some of the data is government issued and available in public.
Literacy: can it be used by users with different technical literacy? What about different languages?
Accessibility/inclusion: can it be used by low vision users? Can it be used by different body types? (ex. many car seats were only designed for men in the past).
Privacy: how does it inform when a camera is on and how does the user control it? (ex. a hidden status LED won’t help)
(Not exactly human subject test but including) drop testing: trying on different heights/surfaces to ensure the robustness.
For quality, where the aim is directional insight and not blind and misplaced confidence in a ‘statistical model’, a good rule of thumb is sets of 5-10 people with your target need/cohort. If you see a relatively consistent signal from your marginal 5-10 people, then you’re probably in a good spot for the particular thing you were testing for.
Of course you should have a basic understanding of methods to know the right sort of research approach for the given question/hypothesis to test.
Know when to actually work with a good quality or quantity researcher. And thread lightly with surveys as we tend to write garbage ones and cherry pick results.
I’m curious too as far as qualitative interviews go, it’s easier to get a sample size with quantitative…
And your specific case what is the research question or questions that you’re planning on tackling?
If you’re a part of a three person startup working on the type of product that you described. Your common use is for customer research methods would either be ux research to make sure your interfaces are understandable and usable, or trying to understand who your customers might be within a market space.
If it’s the latter check out competing against luck and then search Bob Modesta jobs to be done. That’s customer development research.
In that case you’re relying on qualitative methods that are interview based. You can usually get themes within about 5 to 10 interviews and you shouldn’t expect your results to be broadly generalizable but indicative.
Applying simple logic, the bigger the sample, the more meaningful your results.
Sample size is a thing. Isn’t it?
@Karan, surely the size doesn’t automatically make it superior?
What if you had sample of 5 that you got via voluntary survey vs sample of 20 that came to do the survey after an significant incentive?
@Jesus, There are tons of nuances and getting people to participate is hard but yes it’s important.
Especially the lower you go on the lean development pyramid. For example it’s WAY more important to get as big a sample as you can when you’re validating a product idea be validating a UX functionally on a product you already know people want.
Some good example are the many many 10,000 people polls that get the presidential candidacy wrong.
There are more things that matter but sample size is definitely a thing.
Making sure you have the right audience. Qualitative data matters too.
It’s better to do continuous discovery so you can keep making sure your assumptions are valid.
There isn’t a model/ answer for everything. Your goal, thesis, industry will dictate how you go about it. I usually go by 10% as a sample size and extrapolate up and down from there. About 5-10 people will give you about 80 percent of the needed qualitative info.
Also it should be noted that “sample size” is a technical term