Don’t Fear the Small Numbers: Why a Few Design Research Participants Can Provide Big Insights
1. #HxR2016
Don’t Fear the Small Numbers
Why a Few Design Research Participants
Can Provide Big Insights
Dan Berlin
VP, Experience Research, Mad*Pow
dberlin@madpow.com
@banderlin
Susan Mercer
Experience Research Director, Mad*Pow
smercer@madpow.com
@susanamercer
2. #HxR2016
Topics
Prevalent requests for large numbers of participants
How small numbers provide big insights
Determining the “right” number of participants
21. #HxR2016
Usability studies are qualitative
Positive: Users were able to find the
“Buy Now” button.
3 out of 8 participants mentioned
that the picture had no relevance to
them.
Consider: replacing this with an
infographic or another picture that
may draw in users.
6 out of 8 participants mentioned
that the text in the buttons was hard
to read.
Consider: increasing the contrast of
the text to ease readability.
22. #HxR2016
Priority Finding Recommendation
Medium Participants said that the text was hard to read Increase the contrast of the text
High Participants were unable to locate the product; they said
they expected it under Products & Services
Move XYZ product to the Product &
Services area
High Participants were unable to discern the top from the bottom
of the vial
Include a visual cue as to which is
the top and bottom of the vial
Medium Participants were able to figure out how to use the insulin
pump, but said that the instructions were very hard to
follow
Add detail throughout the
instruction manual
Typical Usability Study Results
29. #HxR2016
• 12-16 participants per project segment & user group
Discovery Interviews
• 8-10 participants per project segment
Usability Studies
Very General Rules of Thumb
We’ll start with two things that Design Research is not.
Marketing is great for it’s traditional uses: determining who to sell the product to and at what price point
When determining what people will use and how much they will pay for something, statistically significant data makes sense – you want those large numbers as “proof” for very important company decisions
This is still very important, so still use your 4 Ps or 4 As, the ARF, and Ogilvy’s book (DON”T ASSUME PEOPLE KNOW THESE)
4 P’s: Product, price, promotion, and place
4 A’s: Acceptability, affordability, accessibility, & awareness
ARF: Advertising Research Foundation
We’re not designing print advertisements or TV ads that are a one-shot deal
We’re (typically) designing digital interfaces that can and should easily change
This has facilitated iterative design: where we design something, learn from users, update the design, and repeat as necessary
Even if it isn’t a digital interface – rapid physical prototyping will allow similar design processes soon enough (if not already)
Design research is a very new thing and people often confuse it with market research
Hence the desire for large numbers of participants when familiar with more traditional marketing research, not design research
Iterative processes allow designers to check-in with users to ensure that the design meets their expectations
We can course correct as we learn from/about users
Also allows us to see if user wants and needs align with business goals
Companies can easily release an early beta to get real-world feedback
Can also continue to iterate after release – making incremental improvements
To make a vast understatement: just as with market research, clinical trials are extremely important
Yes, you definitely should have very statistically significant data when determining the efficacy and safety of a treatment
But again, this is not typical for design research – we don’t need to “prove” that an interface is intuitive
Naturally, the exception is when you’re conducting usability research for FDA approval of a device
But that’s an final summative study; there should be iterative research along the way with smaller numbers of participants.
We are talking about design research here today – research that informs the design of interfaces, workflows, devices, and other experiences
We’re not determining who will buy the product and at what price – that’s for marketing to determine
We’re also not determining the efficacy of a medical device or treatment – that’s for the clinicians to determine
Design Research focuses on:
How people complete specific tasks today
What interfaces, devices and services they use today
What challenges they have and why
How these interfaces, devices, and services can be improved
Design Research activities are mostly Qualitative
Problem:
The design of this injection pen required patients to push the two halves together and then press the button to inject.
However, patients with small hands couldn’t do both things at once with one hand
Patient Solution: Use a rubber band to push the two halves together and use their hand to press the button
We would never have seen the use of the rubber band unless we went to the person’s home
Things we have learned from studies in the OR:
Nurses having to change a battery that’s at the top of a stack
Medical device reps often stand in the OR to troubleshoot and advise live
In the 90’s, Bob Virzi found that 5-8 participants will generally find 85% of your usability problems.
More complex now, but it’s still a good rule of thumb.
Minimum of 5 participants per demographic, if they truly represent different interaction/information needs.
Typically, 12-18 participants per usability study, depending on the number of demographics.
There are typically diminishing returns above 10 participants, which can waste precious project time and budget
There are myriad UX design research methods, some of which may be familiar, some may be new
Part of our jobs, as UX researchers, is to determine the right tool to answer a particular question during a certain project phase at the right price
More studies are needed as to how many participants are needed for the different methods
But, for 95% of these methods, under 100 or even under 50 participants will yield plenty of useful data
The nice thing about having all of these methods is that we can be nimble
Remember the iterative design process? Each iteration is different and we seek to learn different things:
Usefulness
Usability
Does the navigation make sense?
Which design is more appealing? Or engenders a particular emotional response?
Again, for each of these questions, the trends typically emerge after just a few participants
UX research is not definitive – we’re not “proving” anything.
We’re giving the design direction based on user input.
When deciding what product to sell or if a treatment is effective, yes, you need research that proves it
But not with design research; we take direction from users, design, check-in with users, and so on
It’s a collaborative, participatory process
At Mad*Pow, we conduct hundreds of interviews each year across a wide array of domain
These are typically conducted in sets of 8-24 participants
Smaller numbers when we just want to “check-in” with users; larger numbers when looking at a broader scope and set of user needs
There are also the studies where we DO have 30-70 participants:
Wide user-base with different needs – Example: Working with a University about a research grant tracking database used by 100s of different users in different departments for different types of gratns
Deep product functionality
FDA approval
In conducting all of these interviews, we typically find that high-level trends emerge after just a few participants
Whether it’s something tactical, for example: most participants clicking the wrong path during a usability study
Or strategic: most participants expressing the same needs for an injection pen diary
So in just a few days of interviews, we can give a design project actionable insights that allow designs to create something that aligns with user expectations and needs
As we conduct additional interviews during a study, there are tidbits of good information that we gather that aren’t part of the overall trends
There’s always that one participant who finds a different way to complete a task, or someone that goes off on a tangent the proves to be insightful
While this can be useful, the ROI on adding participants to get incremental data is quite low
This is an example of what a usability report looks like.
Researchers will point out prioritized opportunities for improvement.
Reporting on the positive findings is important so we don’t break what is already working.
The overall goal is to report on the study findings and the associated recommendations.
Always report a recommendation for each finding.
Ensure the findings are prioritized, a H/M/L scale is typically sufficient.
Focus on your goals and the decisions that you need to make based on the findings
Let the professionals (researchers) recommend research methods and participant numbers
Provide budget ranges and target users – this helps us scope projects
Every project is different and conversations make for a successful project – not RFPs with subsequent Q&A documents
We’re not aiming for statistically significant data – it’s not worth the cost.
We’re not proving anything, we’re getting design direction.
Personas are documents that capture a “typical” user’s wants and needs for an interface, design, product, etc.
If you feel that a demographic needs to be part of the user audience specification, ask yourself “Why?”
What BEHAVIOR or ATTITUDE is this demographic suggesting to you?
Use that behavior/attitude in your user description, instead of the demographic.
Similar to Maslow’s hierarchy of needs, this is the UX hierarchy of needs, first posited by Stephen Anderson, a UX thought-leader
At the base is functionality/usefulness – this is where you need to segment your users for research recruiting
What are the different transactions and information needs that your users have?
Confirm these assumptions with users early in the project and adjust from there
If these needs are very similar, you don’t need that many participants; If these are very different – you will need more participants
This is why the scope of a project is also important when determining the number of participants to interview
Are you looking to update a microsite with limited functionality? You probably don’t need that many participants.
Are you designing an EMR? You definitely need a lot of participants – and a variety of research methods
Chunk larger projects into bite-sized pieces and conduct multiple studies
There may be overlap between projects, but that’s an opportunity to confirm assumptions
The project as a whole may have a lot of participants, but each individual study can have just a few to move the project along
As a vast generalization, these are numbers that tend to work and have proven to be efficient
Discovery interviews tend to be more open-ended and we want to cast a wider net at the beginning of a project
Usability studies and the like tend to be more tactical: Can the user accomplish the primary task on the website?
Tactical findings take fewer participants to uncover
When you keep in mind that design research is neither market research nor a clinical trial and that small numbers of participants can advance the project you:
Get direction from users throughout the project
Save project time and money – participant recruitment can be quite expensive, depending on the demographic
Produce a product that better aligns with target users’ expectations