Derisking Ventures through UX Research: A chat with GV’s Michael Margolis.

How can you find your TAM and validate your product-market fit if you don’t know who your bulls eye customers are?  UX research is a way to accelerate the learning curve of startups and derisk the investment.

This week my student founders and I had the pleasure of welcoming Michael Margolis, UX Research Partner at GV (Google Ventures), to our H4H impact venture incubator course at Columbia Business School.

Michael has been collaborating with GV portfolio companies on UX research for 11 years.  He was the first person doing this at a venture capital firm.  It is still pretty unusual for anyone to be doing this at a VC.  The experiences and effectiveness of his group have been written about in the highly useful bestseller, Sprint:  How to Solve Big Problems and Test New Ideas in Just Five Days by his GV colleagues Jake Knapp, John Zeratsky, and Braden Kowitz.

Michael shared with us highlights of his approach developed over his vast experience working with Uber, Slack, Foundation Medicine, One Medical, Flatiron Health, GitLab, Lime, Kindbody, Gusto, Blue Bottle Coffee, StockX, Cockroach Labs, and Nest.  These are all companies known for effectively scaling based on maximizing the value of their customer/user experience.

I asked Michael to explain the value GV places on UX (User Experience) research for its portfolio companies.  Michael explained to my Columbia student founders that UX research is a way to accelerate the learning curve of startups.  

“It is a way to derisk the decisions you have to make, to validate concepts, to help figure out product-market fit, and to help clarify who exactly is your bulls eye customer.

If we are investing in a company, we want it to be successful.  They have a certain amount of runway, which you can think of as the amount of time to get it to work and make progress.  Research is a way to accelerate that.”

Michael and his GV colleagues assist hundreds of companies by teaching, advising, doing skills training, holding office hours, and helping them hire designers and researchers.  It is like a funnel.  At first, they help a lot of people, then they focus on a smaller number in collaborative research projects.

I asked Michael what were the triggers to embark on such collaboration.  The triggers are those critical inflection points:  expanding, entry into new markets, launching a new product which requires figuring out the value proposition and messaging.  

Research Sprints Defined.  Research sprints are cross-functional team efforts designed to answer a fundamental product question and increase the team’s exposure to their customers.  

“I want them to see, hear, and understand stories, and build some empathy about who really are these people.  A lot of the output is building team momentum and alignment.”   

Michael explained to the class the elements of GV’s research sprints and how they are designed to be conducted in a matter of days.  

Step one:  What do we want to learn?  It is important to define key goals and research questions.  What are the big open questions?  What is the thing everyone is worrying about?  This is typically whether the product is going to be appealing to people and who it is going to be appealing to.  It is important to be very specific.

Step two:  Who are our bull eyes customers?  

“I push companies to best define who is included as well as the exclusion criteria.”

Step three:  Create a simple prototype.  We call this an MVP in my class which is defined as the least effort to create a representation of the offering to get actionable feedback from a potential customer.  The prototype/MVP is designed to answer specific questions.

“I encourage people to use free prototypes.”

Free prototypes?  I was intrigued.  

“A free prototype to me is a competitor’s product, a competitor’s homepage.  Because it is out there.  It exists.  You can learn an enormous amount about other things that are in your space.  And it is usually important to use more than one [prototype/MVP].” 

Multiple prototypes/MVPs give you more actionable information.

“It is not as helpful to provide one MVP.  You need to present two or three.  This is a very powerful thing that you can do in an interview, people comparing and contrasting the goods about this one (this is what I like) and that one (I don’t like that). 

It is not about which is the winner but about comparing and contrasting, and teasing out the best points that I can take into my own next version of things.”

Founders, it is all about iterative learning cycles.  It is also about letting go of stuff that doesn’t work as well.

“When you compare two to three [prototypes/MVPs] you are a little more willing to throw things out.  If you are focused on only one, you get pretty wed to it and it gets a little dangerous early on.”

Step four:  Go recruit people to talk to based on defined criteria.  You recruit them in different ways depending on who they are (ophthalmologists versus busy dads).

Step five:  Hold a party.  Did someone say party?  Step five is to hold all the interviews in one day and hold a UX watch party.  These means that the whole team is in attendance, observing and taking notes.  

“At the end of the day, there are key takeaways and the team has all of this alignment and momentum to move forward.”

This is how you speed up traction.  It is valuable since the clock is ticking as your startup uses up its cash runway.

So why wouldn’t everybody want to do this?  What holds founding teams back from doing such critical UX research?

Founders are very confident.  Those who know they are right may not see a need to do this.  There are certain teams that may not want to risk learning that they might not be right.

Too busy, too time consuming.  The number one reason is that founding teams may worry that they might not have time to do this, that it would slow them down.  

The emphasis of a research sprint is on speed.  A sprint focuses on the one thing keeping the CEO up at night.  It is worth it for a CEO to spend a day to validate this and go talk to customers.  Yet talking to people that use their products is not easy for some people.

“People get nervous.  We get companies with a lot of engineers and they are more comfortable building stuff than stopping and talking to people, their customers, in this way.  It can feel uncomfortable.  It just takes practice.” 

So, what are the dos and don’ts of a research sprint? 

·      Be specific.  Know what you are trying to learn.

·      Know clearly who your bulls eye users are or at least have a hypothesis.  

“Each time you do another round of this, you learn who (and who is not) your customer which helps you figure out your TAM [Total Addressable Market].  How many people are there who are affected by this?  How big of a problem is it?”

·      Be specific about recruiting.  It is how to do this quickly.  If it is too amorphous, it will take a lot of work to dig through it all and discern the key things you want to learn such as who is the customer.

·      Do those interviews in a clump.  Doing all the interviews in a day, versus spreading them out over a week or more, makes it easier to observe patterns.

“At the end of the day, the lessons are pretty clear.”

·      Don’t get too fancy with your prototype/MVP.  Keep it simple.  

“Build just enough to answer your key questions.  You want to simulate an experience.  Make it look real, but just enough.  You should be able to build it in a day or two.”

·      Schedule the interviews.  

“This is the secret power behind research because it acts as a deadline.  It is a way to drive a huge amount of work.”

TAM and Customer Segmentation.  I asked Michael what customer segmentation characteristics he finds most useful to evaluate the market opportunity:  static demographics or behavioral qualities?

“I think it is a mix.  I tend to lean on the behavioral- the experiences of the need, the goals- much more than demographics.  Part of this is that teams are usually more focused on demographics.” 

Michael gave us the example of a grocery delivery company whose team rallied around its perceived customer persona of a busy young mom with two kids who didn’t have time to go grocery shopping.  The team even had a name for the mom.

“We dug into their data to see who actually were their best customers, who were spending the most.  We found out that these customers were not who the team thought they were.”

A key learning was that the target customer was not best defined by age or other static demographic but by behavior.  It was about finding out how much time a customer has available.

“The thing that you are solving is saving somebody time.  You want to find people who are really busy.  It is not really a demographic thing.  You want to find people with jobs and responsibilities so they are time starved a bit.  This is a behavioral thing that we have to look for and tease out when we are recruiting.”

It had to be more even focused than that.

“We wanted to make sure that it was somebody who was reacting to the value proposition of this particular offering.” 

Michael and his team were not looking for people to talk to that were debating whether or not to have food delivered, but how food delivery services were chosen.  This was a different value proposition and it was not based on demographics.  Michael brought up a demographic typically cited that he doesn’t find very useful, household income.  

“The more important detail is how they are spending their money.  Maybe it’s a values thing.”

Michael looks for a proxy.  It might be what kind of car a person drives or their last vacation.

The Power of Data Science & Decision Intelligence.  My last question for Michael was about how GV incorporates data science and decision intelligence in UX research. 

“Great question.  I think these play very well together.  Thinking about qualitative and quantitative is very, very powerful.”

Michael referred back to his previous grocery example where the team had a very specific idea of their primary persona.

“We pushed them to look into their own data.  I had a bunch of specific questions I wanted them to go answer.  Go look at your usage data.  Go look in your sales data.  Who are the people who are your most successful, your bright spots, among your customers?”

This portfolio company was able to dig into their own data and discover that their target customer was not busy moms but this other group of people that they needed to go talk to for actionable feedback on features of their offering.

“Then, the qualitative gives us these very interesting hypotheses about the patterns of behavior.  These are questions that we can pose to the data scientists to see if this is pattern.”

This is a way to quantitatively validate a qualitative observation.  The data scientists can go validate if there is a common pattern across a large number of people.  It leads to aha moments the team has not noticed before.  This leads to insights and further questions not initially asked of the data.

“The quantitative and qualitative go back and forth.  We can use that to determine what is true across a larger number and see what is going on.”

The heart of understanding what is true, who the customers are to validate a value proposition and how representative these customers are to a wider market opportunity, is a core imperative of derisking a venture and enabling startup success.

Thank you for sharing your incredible expertise and vast experience with our students at Columbia, Michael!

Thank you all for helping us build community while addressing problems that affect people and the planet!