As a founder, you're constantly doing discovery. Of new markets to conquer. Of worthy problems to solve. Of insanely great people who you want to hire.
When people ask me how I spike as a founder, this is it. But specifically, it's the 30 min 1-on-1 session.
I realized recently though that I've never tried to put into words what it is that I'm doing. This was written as a way to reflect on some things I've intuited over the years:
- What makes a question good?
- How to prepare for a discovery session?
- What are some techniques for bringing a conversation back on track?
- How do you come up with good questions on the fly?
Good questions
Questions are the basic unit of discovery.
A good question moves you closer towards the larger learning objective. A bad question is obvious because it will feel like standing still or circling.
2 big unlocks for me in thinking about questions to ask:
- Setting a clear intention and learning objective
- Thinking about how the other person might respond
Setting a learning objective grounds every question in a "why". If your question doesn't advance the "why", it's a bad question.
Putting yourself in the other person's shoes forces you to anticipate how they might respond. I like to think about this as a matching problem. Is this the best question to be asking for my goal? And if so, is this the best person to ask it to?
Here's a very common scenario in discovery:
- You're talking to someone about their work, and they raise frustrations with another department (ex: purchasing department)
- You realize you have questions about the origins of the PO process, but the person you're talking to won't have the answers
- Instead of diving deeper into this topic, just politely ask for the best person to follow-up on this with and steer the conversation back
The goal is to make sure every question is maximally applicable to the person in front of me. I'm obviously being extreme of course. In many cases, I will still ask a question to probe because it's cheap to do so (exception: ramblers, see more below).
But the point is this: if I sense that the question was not a good one, I switch tracks immediately. That way I'm not discouraged from asking questions, but I can still control the general flow of the discovery session.
The last thing that changed how I ask questions is making them personal. Instead of asking about what happened, ask about how it affected them. In most cases, by asking a more personal question you'll (A) get a much more interesting response, (B) open up the emotional aperture of the conversation, and (C) get a much more specific answer.
And the reason why I've broke this out is that you can apply this in any setting. Q&A after a fireside chat. Questions during your company all-hands for leadership. Every question can be 10% more personal and yield more interesting responses.
Discovery flow
If discovery were a game, each good question is a successful move towards winning. But what does winning mean? And what is the overall strategy?
To me winning in this context means:
- Walking away with a shared mental model of how things work
- Fully internalizing the problems this person is sharing
- Clarity on next steps in a solution
Having a shared mental model means that you can describe how this just work. You understand the lingo, can form useful analogies, and have concrete examples. It's impossible to understate how important this is. If you don't have this foundation, you won't fully understand anything else.
If you find yourself running out of time in a single session, abandon goals 2 & 3. Just wholly focus on 1, and book more time.
The other 2 goals are fairly self-explanatory. But the goal is to bake this as a layer cake. Understand how things actually work, understand when they don't work (problems), and understand what the delta is between reality and ideal (solution).
The strategy is the agenda and how you navigate it. I like to prep a little going into things, but it's also super important to let the natural flow of the conversation determine where things go.
Phase 1. Getting the download
The goal here is to develop that shared mental model I described above. At this stage, it's very important that you are honest about what you don't understand. It's ok. You won't look dumb. No one expects you to download their entire job in a sitting. I constantly say "Can I read it back to?" to establish that I'm following. And in return, they are excited because they see you are clearly engaged.
Some questions I typically start with:
- What is your role?
- What are you responsible for?
- How does this fit into the overall business?
The failure mode at this stage is not developing a rigorous enough understanding. Maybe it's because they glazed over some details you failed to double click on. Or maybe you didn't stop them enough with dumb questions because you're embarassed.
Other tips:
- Share what your overall learning objectives are before starting. This way they can orient what they show you towards that.
- Share what you already know and who you've already talked to. This will save them time from stating what you know already. Also do this before you start.
- Keep asking them to not just tell you, but show you. Offer to meet them at their workstation so everything is easy to access for them whether it's on computer or on their desk (nice side effect: also very trust building).
Don't move on until you've gotten the full download. In fact, I will often ask before moving on "Did I get the full download?"
2. Diving into problems
The goal here is to fully understand the pain. Who, what, when, where, what.
Some questions I ask:
- What's painful about the process we just went over?
- Why is it painful? How painful is it (in hours, dollars, etc.)?
- What is the consequence of not solving this problem? Who else cares about solving this?
- When was the last time this happened? Can you walk me through an example?
The most common failure mode I've experienced at this stage is letting them pull you too deep into their solutions. If they have tried to solve the problem before, that's a great sign. It means that (A) the pain is real and (B) they are the one that's motivated to solve it. But letting someone give you a full product walkthrough of what they've vibe coded with AI is a waste of time.
If you've read the Mom Test, the idea of "they own the problems, you own the solution" applies here. To the extent that it's useful to see what they've built to better understand the problem, do it. But everything else is eating into valuable discovery time.
3. Exploring solutions
Last but not least, the solutions! Don't get me wrong – I love this stage. But the discipline of discovery is not to jump to this too early without establishing the right foundation.
Some questions:
- What have you tried before to solve this? Why didn't that stick?
- What would a successful solution need to have?
- How will we measure its success?
If you've already started working on a prototype, pulling it up and having them react to something concrete is very helpful. And all feedback is good feedback, but don't get caught up in how the buttons look. Walk through features and map them back to the problems you heard earlier by asking "Would this solve for X?"
If you haven't started yet, you should have clarity on an MVP by the end of this conversation. AI notetaking is great for making sure nothing slips through the cracks. But if you're deferring all of your product judgement to an AI that has access to the transcript, you probably did something wrong.
The failure mode here is spending too much time co-designing a solution. Yes you want them to feel bought in, but that should happen as a result of you fully internalizing the problem. Not from them nitpicking the UI.
Common archetypes
Discovery sessions are best done 1 on 1. Because they are generally open ended, you need to make sure you can control the flow of the conversation while also leaving space to interject and ask the dumb questions. In a big group, you will lose control very quickly.
I also typically don't schedule anything more than 30 mins. If you need more time, just book more time. But in most cases forcing you and the other person to get it done in 30 mins is enough.
This will set you up for success, but some personas will still make it tricky. Here are some common archetypes and how I navigate them:
The Rambler
For every question you ask, they give you a very long and unfocused answer. The risk here is that you really only get to ask a handful of questions, and before you know it time is up and you've learned nothing.
The best way to deal with this is to always share your learning objectives upfront and use that to reign them back in. I'll often say "I'm sorry I asked a question that put us on this tangent, but coming back to X" as a way to not make them feel at fault while gently putting us back on track.
Sometimes people get that as a cue to ramble less for the rest of the session. But if it keeps happening, I'll start to interject more often with things like "Sorry but before we move on can we X?"
It may feel rude but don't worry. People who ramble are also wondering "When are they going to stop me?"
The Explainer
Similar to the rambler, some people like to really get into low-level details that aren't relevant at all.
If they are doing this, it's most likely coming from a good place. They want to share, but you haven't told them at what level of detail it stops being helpful. In such cases, I will often ask to "Zoom back out" to make sure we are focused on the right things.
The Hammer
Everything is a nail to a hammer. This person is the builder, and has already built many solutions to solve these problems. They'll often have multiple projects they'll walk you through proudly. All sorts of dashboards and widgets.
While it's definitely helpful to explore what worked and what didn't, you need to make sure you don't get stuck in full product walkthroughs. Ask "why did it not get adopted by the team?" and keep things at a high level.
Closing thoughts
None of this is rocket science or anything new. The same people that have excelled at this for traditional product discovery will continue to it very well in delivery-oriented roles like FDE.
What is different though is that now we have AI. For taking notes and making sure nothing slips, and for doing an exorbitant amount of background research for you in between conversations. Here's how I like to use it:
-
Taking notes and synthesizing all the action items at the end of the day. And because I know AI is taking notes, I'll use that by just always verbally confirming next steps as a part of the meeting so the summaries capture everything.
-
Background research. I maxmize sample efficiency by getting a few good and bad examples from the client, and then I'll have agents pattern match and find more. That creates valuable follow-up questions because some things true at n=5 do not hold at n=100.
Having intelligence too cheap to meter doesn't mean the problems will go away. In fact, there will probably be even more problems to discover and work on. And hopefully better ones.
Have fun out there.