I’ve learned a lot about building startups over the past few years, and have become much more opinionated on how to do it well. I spent the last 2 years at Clubhouse, where my learning was supercharged by the caliber of our team, the rate of growth, and the very nature of building in a new product category.
When I started, the app had ~1000 users. Rooms didn’t have titles and “clubs” didn’t exist yet. The team was just the founders. Clubhouse has been built in public since, and as an early employee and Head of Community, I became a quite public person at the company and got to contribute to nearly every part of it.
In recent months I craved more time to explore new ideas again as a founder, and decided to leave Clubhouse while staying on as an advisor. I’ve taken away many personal lessons and refined perspectives on how to build and scale startups — especially products, communities, and teams — and how to do it in public.
Lesson #1 — Earn the right to build the next thing
Commit to following through on what you’ve already planned or launched before you chase the next creative, experimental, or shiny idea. “Earn the right to build” and “have respect for sequencing” are principles I’ve internalized.
At the earliest stages, it can mean doing diligence on user needs and testing with no-code MVPs before coding anything. While scoping, it can mean being disciplined about what needs to be in a version 1 of a feature. Post-launch it can mean rigorously ensuring feature quality and evaluating feature success before thinking about what’s next. And so on.
At a meta level, it’s valuable to set up a consistent product development process that considers strategic fit, prioritization, scope, launch, evaluation, and iteration.
Lesson #2 — Strive for a great UX and a good enough UI
Simply put, how things work matters more than how they look. This is especially true in the early stages of a product when you have supportive early-adopters.
A great UX facilitates your product’s core interactions — clear, easy, and simple. A bad UX makes it hard for users to do the thing you want them to do (or even know you want them to do it). Bad UX can turn users off a good product idea.
On the other hand, a “good enough” UI is … good enough (especially since it can be more subjective). For example, color and font schema can influence perception of a product’s purpose and audience (e.g. target age and personality). But, a differentiated product with strong utility and great UX should overcome this.
Some evidence that this holds true: bug reports and product feedback are almost always about UX (or new feature requests), and the ones that seem like they’re about UI usually overlap with usability issues.
Lesson #3 — Product judgment is more skill than talent
Good product judgment is hard to define because it’s empirical; you’re only proven “right” after the fact and over time. It’s also a skill more than it is pure talent; you have to practice to get good and to stay good at it. Arguably the most important way to invest your time — to deeply understand how users:
experience your product (so much so that you can even anticipate how various target user segments would react to changes in your product)
experience other products (know the evolving socio-cultural trends and understand the products that may compete to serve similar needs)
Product thinkers say that some combination of expertise, empathy, or creativity underlie good product judgment; deeply understanding the product and the market directly through users will help with all three. This doesn’t have to mean building a UX research function early. The best results can come from doing it yourself (see lessons #6 and #22) and drawing on the collective wisdom of a team.
Lesson #4 — Beware of complexity creep
The first version of your product is usually simple. As you keep building, the product becomes more robust in quality, utility, and experience. But, there’s a tipping point after which “robust” can sneakily switch to “complicated.”
Some signs your product might be too complicated: You’ve launched a bunch of features and haven’t killed any of them. You’re running out of literal surface area to put new features. Users aren’t clear what to do with the product and how.
Complexity creep is a human problem. When we make things, we get attached. We don’t want to seemingly discard people’s work. We may not rely enough on data to make decisions. Or, we just aren’t clear on the one or two core interactions that users need and value, so it’s hard to choose what stays and what goes.
Anticipate this, then build in early checkpoints. Define success and failure before a launch. Build a solid process and track the right metrics to help make objective, decisive decisions — and don’t forget to create a framework to sunset things!
Lesson #5 — A ‘community’ is really a collection of many communities
If you have 2 users, you probably have 2 user personas. It’s illustrative but even a small group of users can have a vastly differing set of experiences, perspectives, goals and needs for your team’s attention and your product. Hence, groups of users form and represent many distinct communities as a network grows.
User segmentation isn’t an academic exercise; done right, it’s a strategic and operational one. Use qualitative insights, data, and reference users to help define personas. At a strategic level, match value propositions to user segments. Choose to prioritize one user segment and hence not another. At an operational level, classify users into a specific segment and support them accordingly. Then build systems and allocate resources to do it well at scale.
Lesson #6 — Talk to users (even if you don’t think you need to)
This shouldn’t be a one-off thing, and it should be something every team does. You should talk to users when you’re trying to figure out what product to build, when you’re ready to launch, or even when users reject it. Talk to them not just about product but also about their motivations and needs from a community. Bonus points if you can engage users in a way that’s native to your product!
There are many traps (logical and emotional ones) that can get in the way. The Jobsian philosophy might be cited – “some people say give the customers what they want, but that's not my approach. Our job is to figure out what they're going to want before they do.” Even if this is true, talking to users is still valuable. Just use the insights to understand problems and refine solutions, not to generate them.
Another potential trap — “the most vocal users aren’t the users we want.” If you’re distinguishing between users you want and don’t, make sure you really know the difference (see lesson #5). Then do the work to reach the users you do want. I’m confident you can find both thoughtful advocates and critics of every feature, program, or strategy you might be considering.
Lesson #7 — Create value through user-to-user interactions
A true community is defined by the interactions, relationships, and value exchanged among its members. Hence, building a community alongside your product shouldn’t focus just on team-to-user interactions, but also on user-to-user interactions. Figure out where team needs (e.g. product insight, user education, user growth) align or don’t align with the needs and interests of engaged users. Use this to help identify and test new engagement initiatives.
Importantly, don’t just tell, rather show users how you want them to engage. Collaboration between community and product can be very valuable here. For example, to tackle user education, demonstrate the feature in use via the product (e.g. onboarding flow) and in the community (user-led onboarding sessions). If you want users to create, engage existing creators to show what a good process and output looks like. Your users and supporters can augment your team’s impact.
Lesson #8 — Trust & safety is a continual effort, not an end point
Managing user trust, safety, and privacy has to be a conscious and consistent effort; it’s never too early to start and it’s never fully ‘done.’ The first thing to do is anticipate what challenges your product uniquely presents in this domain. Then, figure out how to tackle challenges with product, process, and team.
Create an initial set of rules (read: Community Guidelines!) for how you want users to engage with your product, with each other, and with your team. Get input from your existing users. Avoid the legalese up front so it’s a quick exercise. This also helps give clarity to your team and guidance for decisions that arise.
This is especially important if you want to build an open network. Figure out which rules you want to centralize versus leave up to users. Anything centralized sets a network-wide standard where your team may be the arbiter; anything decentralized can be taken as permissible and may feel out of your control.
Go-To-Market & Growth
Lesson #9 — Your friends are false positives until proven otherwise
Often the earliest users of your product are your friends, family, colleagues — people you personally onboarded and who have a vested interest in your success.
If you can’t get your day 0 supporters to use and like the product (assuming they’re in the target market), that’s probably an accurate, and negative, signal. But, if they do like it and even act like an ‘ideal’ user would (e.g. sharing the product with their friends), you should still consider it a false positive! It’s valuable feedback but isn’t an objective indicator of quality or growth potential.
Extend out one more degree of separation — if you start to see friends of friends invite their friends repeatedly, now that’s a stronger type of signal. You’re finally seeing intent that doesn’t rely on your direct relationships and influence.
Lesson #10 — Product-GTM fit comes before product-market fit
Product-market fit is an end state; a go-to-market (GTM) strategy however is an input you control! And without product-GTM fit, you can’t create product-market fit; you can only stumble upon it (and who wants to depend on luck?).
One example of product-GTM fit is in choosing the right user to serve. If you’re building a group-based product, figure out if there’s a group “leader” archetype and if that’s the user you should explicitly target and build for. Tl;dr — make sure the user you’re courting is right for the product you’re building, and vice versa.
Another example is in picking an effective seeding strategy. If you’re building a closed or default private network, you can pick many, distinct cohorts to test with since they’re effectively walled off from each other. If you’re building an open or default public network, the first cohort matters more — seed it deliberately in character, size, and pace. This group will set early user behavior and culture, influence direction of growth, and give feedback that shapes the product.
Lesson #11 — Start building a growth engine before you need it
You don’t get to choose your growth rate. You can influence it or slow-drip it (e.g. gating early access), but the input-output relationship isn’t linear, and it takes time to figure out the formula too. Whether you onboard early users one by one or are fortunate enough to have viral growth, you will have to build a growth engine.
Start understanding how growth works for your product early so that when, not if, growth rate shifts (up or down), you know what your levers are and can quickly move into execution mode. Treat the growth engine as a set of “loops” (circuits that compound) rather than “funnels” (one-way channels that don’t compound).
Lesson #12 — Invest in your community-led growth flywheel
Traditional product-led growth (PLG) means building a great self-serve product that can basically acquire and retain users itself. This is a direct product-to-user sell and a “best product wins” approach. Community-led growth (CLG) engages the community in the process, enabling a “best recommendation wins” approach.
PLG may work best for products that are utility-based, can provide value on day 1 (e.g. without a network of users), or once a product is socially or culturally validated. CLG stands out when the product’s primary purpose isn’t utility (e.g. instead it’s entertainment, status, new social connections) and people may have lots of options — here a vibrant, vocal community can make a difference.
For instance, even a small group of early users that are community organizers or ‘network nodes’ can champion a product and informally acquire, onboard, and help retain users. The tradeoff with CLG of course is that it’s operational, hence it’s slower and likely harder to find great economies of scale. But, validated community-led approaches can also be built into product in creative ways!
Strategy & Operations
Lesson #13 — Strategy is not doing everything
Strategy is equally about defining what you want to do and what you don’t want to do. A good strategy can be written down in a one-pager. It’s actionable and measurable. It forces you to tackle distinct (or diverging) priorities in series instead of attempting them in parallel. It drives team alignment, hiring, and resource prioritization. A good strategy is communicated transparently and reiterated constantly inside the company and the essense of it shared to users too.
Most teams do start with a strategy. But, once a product launches, there’s no guarantee it’ll be used how you expect or want it to be. When you start to see multiple use cases and user segments form, it’s exciting, but it risks muddying your strategy (especially if one demands a change in your product roadmap). In times like this, you are forced to decide which directions fit your strategy or are a distraction — because doing everything isn’t a strategy.
Lesson #14 — Expand scope in good times, narrow it in the bad
Know when to push and when to pull back, but things should never be stagnant. When things are going well (e.g. the product is strong, growth is good, team is operating well, morale is high), expand the scope of work and experiment more. If things aren’t great, in one or more ways, revisit priorities and narrow scope. This is helpful at both a macro and micro level — for a company, team, or individual.
Lesson #15 — Every team needs a strong operator
Operations is not just a function, it’s also a skill. In the early days of a startup, you’re basically just one cross-functional team. One strong operator is enough. But as startups grow, you might set up independent teams for each function, and “operations” can get siloed. Or, you might set up several, distinct cross-functional teams that are each responsible for a business goal.
Regardless, make sure you still have a strong operator in every team — whether someone native to that team or by embedding a member of the operations team. Part of operating well is breaking down big challenges into actionable workplans, getting team alignment, then relentlessly executing on them. Make sure someone in every team has the skills, agency, and support to play this valuable role.
Lesson #16 — Know when and when not to measure ROI
You can think about ROI on money or time (either your own, your team’s, or even the company’s). Consider this upfront before committing resources, but also when things have been done a certain way for a long time; there’s usually room to prioritize and focus further. Sometimes this can mean building operating models, and other times it’s a simple 80/20 exercise based on gut and team experience.
But sometimes ROI isn’t the best measure — often for sub-scale, experimental, or other creative initiatives. Resist the urge to believe that things without a clear cost-benefit equation today aren’t valuable. Instead, outline the objectives clearly and create a process to measure inputs vs. outputs, impact expected vs. delivered.
Building & Managing Teams
Lesson #17 — Focus on skills, not functions
For early-stage and / or high-growth startups, traditional organizational functions are too rigidly defined. I’ve started to think in terms of skills instead.
There are only 4 hard skills you might need to do any job — data, design, engineering, and operations. Every “job” requires a unique weighted average of these 4 skills. Being a product manager can be a heavily operational role. Being a marketer can be a data-oriented role. “Functions” tend to focus on the application of these skills with a defined objective and in a defined context. Functions don’t always translate across orgs; skills usually do — consider both when hiring.
With this framework as a starting point, you can better break down any role, understand skills required, and nature of day to day work. It can help hire more effectively, better structure and integrate teams, and help employees develop.
Lesson #18 — Teams shouldn’t be “upstream” or “downstream”
Simplistically, there are 3 top-level domains in most organizations — make, sell, and support. In a consumer tech startup for example, the product, design, and engineering teams might constitute the “make” domain. The “front of the house” teams may “sell” (e.g. marketing), and the “back of the house” teams may “support” (e.g. customer success). Importantly — don’t think of these domains as upstream or downstream of each other; think of them as collaborative!
It’s easy to forget to build strong yet flexible touch points between domains, especially as you scale. This is also true within sub-teams and sub-sub-teams (you get the idea). Make sure to build in timely cross-team communication, collaboration, and shared decision-making.
Lesson #19 — In hypergrowth, hire startup-stage chameleons
Hypergrowth feels like you’re running a huge company with a tiny team, and like it happened overnight. Hiring yourself out of the imbalance is challenging. First, you should be reasonably confident that you have product-market fit so new hires will match the strategic direction. This is of course more critical for specialized and senior roles. Timing is also important because growing the team will paradoxically slow you down as you onboard, ramp up, and add layers.
New leadership hires should have 3 types of expertise: building processes and systems to scale, hiring and managing teams, and applying industry experience (ideally multiple reps at tackling similar challenges end-to-end). Bring on people who have this expertise and are extremely curious, use the product, want to talk to users, can run fast and lean, execute as an IC or manage people, and work cross-functionally. Yes, it’s a lot to ask. You want people who can build 0 to 1 and have built the 1 to n — i.e. they can play early or growth stage roles.
Lesson #20 — Communicate with optimism about the future but realism about the past
Optimism is very valuable but it’s only forward looking. When looking at the past, it’s important to reflect with a heavy dose of realism. Team retros are a great tool. Which hypotheses, decisions, outcomes were good? Which weren’t? Why?
It’s also critical to communicate what’s going well and what’s not. Teams are smart and they’ll know regardless, whether because they’re close to the data, talk to their teammates, spend time with users, or just use the product a lot themselves. Trust and morale aren’t based just on optimism or success; they’re built up by a consistent sense of transparency, alignment, and having a unified plan of action.
Lesson #21 — Company culture is built in DMs and 1:1s
It matters far less what you say in all-hands meetings or post in a #general slack channel than what happens the rest of the time! Behaviors demonstrate culture, and most behaviors — good and bad — happen behind the scenes, in the slack DMs and the 1:1 meetings that are far more frequent than any public gestures.
Keep this in mind when hiring too! The first 10-20 employees have an outsized impact on the culture and direction of a company, directly and indirectly. The people they hire, the team culture they set, and how they engage at all levels and in all channels will be hugely influential, so evaluate cultural fit holistically.
Working at a Startup
Lesson #22 — Be a power user of your own product
If talking to your users is important, being a user of your own product is just as critical. (Luckily in consumer tech you’re almost always a potential user.)
There are lots of reasons to do this: 1) it builds empathy for your users 2) it builds empathy for your team 3) it helps clarify user feedback and context-poor data 4) it can reveal new product ideas and, hence, 5) it improves your product judgment!
Pro tip: know your own usage patterns and those of your team. Find the power users across your team and draw on them for insights. Keep track of how usage changes, with or without product changes, and understand why. Product and user-facing teams in particular need to be immersed to keep learning.
Lesson #23 — Unblock yourself and everyone else
Whenever possible, don’t rely on or wait for other people to remove an obstacle. Tangible obstacles are things like skills, information, access, or time. Intangible obstacles are often related to communication or a psychological barrier like lack of alignment or fear. Be proactive about recognizing and removing your own obstacles, communicating them transparently (see lesson #24), and making it very easy for others to help you when necessary. Take it one step further and try to help unblock others proactively.
Lesson #24 — Overcommunicate, don’t overpromise
If I had to give a single piece of advice to anyone working in a fast-paced, high ambiguity, cross-functional team environment, it’s this — overcommunicate. This applies to the macro and micro, to every level, and on every day. Examples:
Be responsive to requests and set context even if you won’t be able to fulfill them quickly or possibly at all (read: don’t overpromise).
Keep people updated on the status of your work. E.g. share weekly progress updates in an easily digestible form. Link to a central place where people can dive deeper into relevant work or catch up on progress any time.
Don’t ask for permission,
ask for forgivenessbut give a heads up. Share your thought process and plan of action and leave room for debate or disagreement. You’ll usually just get confirmation and avoid bad surprises.
Give constructive feedback to everyone (not just people who report to you) and ask for it in return. Do this for the good and bad, at regular intervals.
Set boundaries to respect your personal health, relationships, and goals. It goes without saying these should be reasonable; then communicate them!
Lesson #25 — Know which lane you want to be in
As I pondered leaving Clubhouse in recent months, I talked with mentors and friends in the company and externally. They all echoed a version of the same questions. “What do you want to spend your days doing?” “Which direction do you want to take your career?” “What’s important short-term vs. long-term?”
A common framework for startup career paths emerged – founder, executive, or investor. There’s no hierarchy because they are fundamentally different lanes, but it’s important to know which lane you’re in right now and which you want to be in. Some people stay in one lane; others zig and zag between them.
I’ve also found it clarifying to think about energy. People often ask how energized you are about your role; I like to ask how energized you are about the idea of doing anything else. Relative energy is just as impactful as absolute.
Once you know what you want, make sure you’re 1) building the relevant skills 2) being direct about your goals 3) tracking your own progress 4) regularly zooming out to think big picture and 5) finding a community of like-minded people. My two cents – no matter what you want to do, do it with intention and energy!
A huge thank you to everyone that I had the pleasure of working alongside and learning so much from. 💛 And, for those reading — if any of these lessons resonate strongly with you, let me know!
Thanks for reading Working Theories! Subscribe for free to receive new posts.
Great nuggets of knowledge and experience! I loved how you broke down functions into skills. Was wondering if sales would make as the 5th?
"There are only 4 hard skills you might need to do any job — data, design, engineering, and operations."
Thanks for sharing!