AMA: Atlassian Head of Product, Jira Product Discovery, Tanguy Crusson on Building 0-1 Products

There are multiple things you need to get right before you start building a product, because the most likely outcome of creating one is that it will fail. To see an example of this you can watch this talk for how we did problem and solution discovery when creating Jira Product Discovery * Validating the problem space is important. For that I think the recipe is pretty straightforward: talk to customers and prospects and let them guide you. It's quite important to go in there open minded about what you're going to learn - because there's a high chance that all the assumptions you made when going into these conversations are wrong, that people don't face the problems you thought they faced, or not with the intensity that you imagined. I'd recommend getting coaching from a user researcher on interview techniques - otherwise you can easily meet with 50 customers and learn nothing. It's one of the most important skills for PMs (and humans are naturally pretty shit at this so it REALLY takes coaching). * You want to land on problems that are felt very strongly and urgently by the people you talk to, and are pervasive. * By doing all this you'll form a view on who your target customer is - you want to keep refining this. And you'll identify potential customers you'll want to work with to shape the solution. We call them "lighthouse users" and it's who you would want to build the solution for: they feel the problem strongly, are crafty and have tried many different solutions to solve them to no avail, they're great communicators, easy to collaborate with and have urgency in solving the problem. Find them, and the rest of the journey will be much easier! Read this post to learn more about lighthouse users. * But on its own it's not enough, there are other aspects to tease out. * You want to be clear on if there is a market - check out which companies play there, try to get a feel for their revenue and growth (there's a lot you can glean about that online); if you can, try to chat with investors who gave them money; understand the maturity of this market (early adopters, early/late majority, etc.). * You want to have a thesis for how you can win - e.g. do you have a distribution advantage, access to tech that's hard to reproduce, other? Note that the best product doesn't necessarily win distribution is going to be key to answering this question. * Distribution is super important and what will make or break your product. It's a good idea to validate demand and your ability to reach potential customers. In the past I've done this by creating and advertising a landing page on a website stating the problems and asking people to join a waitlist. That's how we knew we were onto something in the early days of Jira Product Discovery - within 3 weeks we had 3000 people on the waitlist (which contained zero information about the solution). * You need to have a clear answer on "why now?" - the problem needs to be felt really strongly and urgently by customers, maybe there is new tech opening new possibilities, maybe it's the right moment for your company to invest because of strategic reasons, etc. But you need to be clear on why this needs to happen now vs next year. * If you're in a bigger company with cash at hand, assess build/buy/partner opportunities - before jumping straight to creating something a new product: is there another company you could buy instead of building, and fast track everything by a couple of years? * Even after answering all this I wouldn't jump straight into asking for engineers. I'd only do that if we're not clear the solution is feasible, e.g. if it requires new tech that we don't master. After problem discovery, don't skip solution discovery! There's a lot you can/should validate without writing code: * In the past I've shown people slides with value propositions and asked them to explain back to me how the solution would help them. There's a lot I learnt by doing that! * Create prototypes, in Figma, or live prototypes than people can play with. You can gauge the reaction of the people you give them to: if they "just" look enthusiastic about the solution you can throw it away. If they're straightaway asking you if they can please please please start using it tomorrow then you know you're onto something * Even if you ask for an engineering team: start with technical spikes and prototypes, the goal is to validate the solution solves the problem and is feasible. All of this will also help you refine your understanding of the problem space (problem <> solution work hand in hand) You can also read more about it in the Atlassian Product Discovery handbook, that we wrote to help with things like this. Check out the "Ideas" section.

I don't really have a "scientific" answer to this - I've always done this in 2 ways: creating a bottom-up model, and trying to find data from other companies/competitors to benchmark against. Bottom-up model I've usually done this in a giant spreadsheet, trying to build a model with a lot of documented assumptions. Here's how we did it for Jira Product Discovery: * First, decide on how we would package and price the product, which we did via conversations with our early lighthouse customers, a survey, etc. We landed on: per user pricing, some users pay, some users don't, at a price of $10/user to promote collaboration use cases (we didn't want the price per user to be a blocking point to invite more people to collaborate). I'm not an expert at this but I highly recommend this conversation between Lenny and Madhavan Ramanujam on "the art and science of pricing" * Then, estimate how much revenue, based on that, we could get from the current customer base in the beta. We ran a beta for about 18 months and had > 1000 companies using the product for free. We made assumptions about how many of them would buy the product based on their usage of the product and what we knew about them from our other products (e.g. if they're already paying for Jira, if they're paying for more than one product, etc.). We already knew how many users they had on Jira Product Discovery, so with a price of $10/user we were able to estimate initial revenue in the months following general availability * After that we estimated sign-ups and conversions. * For that we used the data we had from sign-ups and conversions from the beta, we adjusted it (because beta was free), and we compared with sign-up and conversion rates from other products at Atlassian - some were a good benchmark (early products), some were not (more mature products). * We also estimated the impact of experiments we would run to try and recommend the product to existing users of Jira. We knew this was going to be our main distribution strategy, which we had validated via a few experiments already, so we had some data to work with. * Based on this we came up with a guesstimation of signups / month, conversion rates from trial to free/paid (which would go down in time as we reach more customers who are not early adopters), we took into account an element of churn, we estimated the number of users per customer (based on beta users, and also assuming we'd progressively reach bigger companies), etc. * We added to the model future initiatives that could impact revenue - for example the launch of a Premium plan at a higher price point. * Etc. - basically we built a bottom-up estimation based on everything we knew (data) but also bets into the future (new features that make the product appealing to more customers, new plans) The model is usually wrong at the beginning, but it does get better over time as real data comes in. Benchmark with competitors At the very beginning of the product, before we even wrote a single line of code, we tried to assess the TAM/SAM/SOM. There's a lot of literature online about how you do that. What I found works best for me is to assess this by proxy, by looking at other players in the market/competitors. You can find a great deal about a company by crawling its blog, finding out who is in charge, look for talks they gave, opinions they give online, etc. There is so much information online. For example: a competitor might claim they have X users, and you can match that to their pricing structure (public information) and estimate their ARR (Annual Recurring Revenue). They might boast on a podcast that they have 80% growth in the past year - that gives you an idea of their momentum. Do that for a few companies and it gives you an idea of the current size of the pie and how big the pie is growing. And you can decide how big of a slice of that pie you want/need in 1-3-5 years for this to be worth it. Careful though: another company making $100M ARR is not a guarantee it will be the same for you: it also depends on your business model, pricing and packaging. -------------------------------------------------------------------------------- By using a combination of these 2 approaches (bottom-up, top-down) you can usually get to a decent answer. But remember all this is bullshit up until the moment where customers take out their credit card! So make sure to find ways to keep testing your prospects' willingness to pay from very early days.

The following stand out to me for what I've witnessed: * The team is acting like the product they're creating is going to be successful. I'm amazed at how many successful companies try to build new products like they do for their existing successful ones: same processes, same success metrics, etc. They don't have the urgency, move too slowly, learn too little, are complacent. The market is a graveyard of failed products and businesses: the most likely outcome, when creating a product, is that it will fail. As a result we shouldn't assume success but instead focus on doing everything we can to de-risk, test and validate - and make sure we don't over-invest too early. I recently was interviewed in Lenny's podcast to talk about that: if you're working on 0-to-1 in a bigger company I think it's worth your time. * The team built a solution to their own problem and they assumed that many more people would face it in the same way. We hear a lot of success stories that started like that, and while it's true that some successes began like this, it's also where a lot of failures came from. You're not the customer of your product: work with real customers! Or it's only going to work for a small population, and that may not make it a viable business. Mehdi, the founder of Cycle, shared openly a great example of this for how they changed their approach to user onboarding. * The product manager tries to get a big team from day 1. Honestly a small team can move 10x faster than a big one. You need a team of people you can trust who can move mountains on their own and continents together. For the first year of creating Jira Product Discovery we only had 5 engineers, and we moved really fast. It forced us to focus on the most important things, to be nimble and creative when testing solutions, and we could do this with very few meetings (it's very easy to get everyone on the same page at that scale). Once you've proven you have a solution that solves real problems for a small number of users - you can then go back and ask for a (slightly) bigger team. Resist that urge to grow too fast! * The team is fighting too many battles at once. This happened to us at Atlassian when we tried to rebuild Hipchat into another product called Stride. We were trying to launch a new product built on top of a new platform that could be used for all other Atlassian products. We faced this tension product<>platform from day 1, and it was much slower to move on the product front. Like with any failure there's of course more than one reason this product didn't pan out, but this is definitely one of them. * The PM fails to get the right level of buy-in internally - if it's about creating a product inside a bigger company. It's very important to "build chips" in the company (gain trust from leadership) so you can spend them, gain more chips, etc. And it's key to communicate your learnings and progress in a way thatmake your team look like a high speed train - no one wants to get in front of a high speed train. This is one of the topic I cover in an interview on Lenny's podcast.

You should think of it as: it should be ready to be shipped when it's the first shittiest version, and when it's the best version of itself. The question should not be not so much about readiness of the solution, but readiness for whom? I've witnessed teams who got too excited when creating a new product and opening the floodgates to let anyone try it too early. That usually doesn't end well... You don't want many people jumping on the solution when it's not ready for them: first impressions won't be good, they won’t stick around and you'll need to work hard to get them to try again. Here's the approach we take in my teams: basically, for anything we ship, whether that's big new features to Jira Product Discovery or new products (including Jira Product Discovery itself), we work with progressively more customers (10-100-1000) before making it generally available. It helps us test the solution very early with a handful of customers to get feedback and make sure the solution delivers on its promise for them, and ensures that everyone only gets the solution when it's ready for them - thereby increasing retention & minimizing churn. If you want to learn more you can watch this talk I gave about the process we went through when creating Jira Product Discovery. You can also read more about this topic in the Atlassian Product Discovery handbook, that we wrote to help with things like this. 0->10 customers (preview) In the first phase of early stage bets we only work with max 10 customers. We unpack the problem and test solutions and iterate fast, which is a process that’s best done with a small number of users/customers that feel the problem the most. And we get the solution right for them. It’s easier and much more focused to do it this way than “throw it out there and see what sticks”. Users who feel the pain the most will be happy to work with us, we can chat with them on Slack/Zoom/Email easily. They're happy to work with incomplete solutions (we can take a LOT of shortcuts) and we can remove the need to pile on untested assumptions. If the solution doesn't work we can throw it away, it's cheap, no harm done. To demonstrate progress to leadership we use metrics that best represent what we're trying to prove at each phase. We started with the following for the private preview of Jira Product Discovery: 10 active teams have been using the product for more than 3 months and plan to continue using it when we enter beta 10->100 customers (alpha) Then we progressively invite more customers to use the solution. At this stage we usually need to polish rough edges and address more scenarios based on what we learn from working with customers of varying needs, and maybe less willing to work with a rough prototype. It's also harder to collaborate with every customer 1-on-1 so we need to create better onboarding material, demo videos, etc. We move support to a community forum. Our success metric at that stage when creating Jira Product Discovery was still focused on problem-solution fit, but for more customers: Product market fit score of 40% or greater with 100 active teams (I highly recommend the Product market fit score survey, and you can read all about it here) 100->1000 customers (beta) At some point we become ready to share with more users, in our case for Jira Product Discovery it was when we reached 50% PMF score for more than 100 active teams. At that stage we needed to focus on making the solution fully self-service - we couldn't afford to have to talk to every customer once in their journey. So we focused on in-app onboarding, we improved usability, we polished the design, we scaled the technical implementation, we trained support and sales teams, etc. It's really about making it ready for prime time. It was also time to change how we measure success by introducing more metrics that represent the health of the funnel as more customers discover, try and adopt the solution on their own. That’s when we started adopting pirate metrics (AARRR), and here's a great read about them. In the early days of that stage we also focused on validating a pricing model, but that would be a topic for another post. Once we were done with all that, had high PMF score for 1000 customers, healthy conversion rates and retention, low churn, validated pricing - that's when we decided to make Jira Product Discovery generally available. These phases work for us - it doesn't mean they will work for you, but I believe that this general approach should apply to a lot of contexts. Basically: don't focus on "when the solution is ready to be shipped?" but "who should we be working with right now based on the state of readiness and validation of the solution?"

Great question. It's really hard to prioritize small, iterative product improvements against large new features/bets. In my experience you need both, as well as a few other aspects. The way we do it in my teams is to think of it as balancing investment levels between different buckets, and to dedicate capacity to each of these buckets. Otherwise it's a constant struggle. We've tried to describe it in this section of the Atlassian product discovery handbook talking about ideas, and that one about prioritization. A couple of different types of buckets: * Boulders, rocks and pebbles * Boulders: large investments with potentially big payoff but high uncertainty, too. E.g. one or multiple teams over one or multiple quarters. * Rocks: medium sized investments with fewer risks, but potential for delighting users. E.g. one team for a month. * Pebbles: Small, typically straightforward change. E.g. one person for a week. * Don't underestimate the impact of rocks and pebbles! In my experience users LOVE to see the app they use get better every time, that's a great way to create fans. * RUF: Reliability + Usability improvements + new Features. Think of the RUF framework as a pyramid: * At the base of the pyramid there's Reliability. Reliability is about building trust. Trust takes a long time to build, but can be destroyed very quickly — a single event of data loss or security breach can be a serious source of churn, let alone repeat incidents. So you need to invest in your product's reliability first and foremost. * Usability Improvements comes second: a feature is rarely “done” — it’s part of a system and that system needs constant tuning. In your roadmap, it is important to allocate budget and resources to keep investing in improving your current feature set. * At the top of the pyramid is new features, both large and small. Then you decide how you want to invest in each: E.g. for a super early stage app you might be spending all your time on boulders and rocks, and little in reliability or usability improvements. For a more mature product you might spend 50% in the reliability bucket and only 10-20% on new features. Then how you actually implement that in your team can vary. In my teams we look at it at investment over time: we might be focusing on a boulder for a quarter, then go back and tackle a few rocks and pebbles. Some of the teams have a rotation where 1 engineer is focusing on pebbles each week. etc. But the important part is to have a strategy and make it a conscious choice, vs something that you react to every time you get a new request from a customer or stakeholder.