Part IThe Story
Every complaint goes to him. Not most of them, and not a weekly digest assembled by a chief of staff who has sanded off the sharp edges. Every one. When Datadog had about seven hundred employees in sixteen offices and was doubling in size each year, its chief executive sat on a stage at the first SaaStr Europa in Paris, the city whose engineering school had produced him, and told the room that he still saw every customer complaint and every support email that came into the company. Then he described what seeing meant. "I don't read most of them," Olivier Pomel said. "I delete most of them right away. I just scan them very quickly."
He never replies. He never fixes the problem in front of him or goes back to the customer, and he doesn't forward the email with a terse note to whoever owns the feature. He lets the stream pass across the screen until some shape in it repeats often enough to mean something, which he calls pattern matching. When it does, he sends the signal down through the managers and the chain of command and lets them decide whether anything needs to change. He is collecting texture, not information.
This is a slightly perverse habit for the man who runs the company that watches the cloud. Datadog's entire business is the opposite of scanning and deleting. It instruments machines, applications, and networks, and captures what Pomel once called "all of the exhausts from those machines and applications that tell you what they're doing." It turns that exhaust into dashboards, alerts, and graphs so that the engineers responsible for ten thousand servers can tell, at three in the morning, whether the credit cards are still clearing. At the time of the Paris talk the company was processing four or five trillion records a day for its customers. Asked which numbers he used to steer his own product, its founder said something that could have come from a novelist: "Curiously enough for a company that's gathering so many metrics, we are not incredibly metrics driven for the product." Metrics lag, he explained. They "don't convey all the nuance."
Years later, with roughly eight thousand people and twenty-odd products, Pomel was still at it. He was sampling support tickets, reading sales transcripts, reading every comment on every employee survey, and firing off Slack messages when something snagged. His stated reason was to avoid living on what he called "beautiful summaries." He sells the most sophisticated summary engine in enterprise software, and he does not trust summaries.
The paradox is less strange than it looks. The company was built on a single wager: that the hardest thing in any organization is getting people to look at the same reality, and that reality is usually bad news nobody wants to deliver. Pomel has spent sixteen years building machines that deliver it at scale. In all that time, as far as the public record shows, he has kept exactly one piece of bad news for himself.
By the Numbers
The Company That Watches the Cloud
$3.43BDatadog revenue, fiscal 2025
8,100Employees at end of 2025, up from about 700 at the Paris talk
$7B+Cisco's acquisition offer, rejected the day before the IPO
$648MRaised in the September 19, 2019 Nasdaq IPO, at an $8.7B valuation
~$25MCash burned on the way to more than $2B in revenue, per Index Ventures
750+Technology integrations supported as of October 2024
0Lines of code written in the company's first six months
The Severed Floor
In 1999, a computer science graduate from the École Centrale Paris arrived in New York for what was supposed to be a six-month internship with IBM's research group. The Frankfurter Allgemeine Zeitung later called it an Intermezzo, a short interlude meant to give a young engineer some professional experience before he went home. He did not go home. "I've actually lived in the US since 1999," Pomel told the Paris audience, apologizing for an accent that had stopped matching his address a long time ago.
The job did not begin in Manhattan. "My first job in the US (with Alexis Lê-Quôc)," Pomel wrote on LinkedIn decades later, "started in a Saarinen-designed research complex in upstate New York. I get flash-backs when I watch Severance." The reference is precise. Severance is a television show about office workers whose working selves and home selves have been surgically separated, so that each half goes through life unable to see what the other knows. It is a comedy of the partitioned mind set inside a midcentury corporate building of long corridors and sealed departments. When Datadog booked the show's star, Adam Scott, to speak at its DASH conference, Pomel added a line: "The cycle is complete!"
It is tempting to read too much into a joke on LinkedIn, so it is enough to note that the company he eventually built exists to keep one half of an organization from being blind to the other.
The man beside him in that upstate corridor came from the same school by a slightly different route. Alexis Lê-Quôc finished at Centrale with the class of 1997, two years ahead of Pomel's class of 1999. Their first meeting has become a small piece of company lore, compressed by Datadog's social media team into one teasing line: "Before they were co-founders, they were on opposite sides of a campus network investigation." The company has left the rest for its own video series, and I won't speculate about which side either of them was on. What matters is the arrangement. One student was poking at a system and the other was responsible for its integrity, and they met at the point where those two roles collide, as developers and operators always do.
Lê-Quôc became the operator in every sense. His later résumé reads like the career of a man who keeps the lights on: IBM Research, then Neomeo, then Orange, the French telecom. Datadog's own biography credits him with "a strong focus on technical elegance and operational efficiency," and describes him as "a member of the original 'devops' movement." That second phrase is the closest corporate prose comes to calling someone a founding member of a scene. He gave talks at Monitorama, DevOpsDays, Velocity, and PyCon, the conferences where the people who get paged at night compare scars.
Pomel built things. While still a student at Centrale he was one of the original authors of the VLC media player, the free, open-source program that grew out of a project at the school and went on to play video for an enormous share of the world's computers. VLC is defined by a promise of radical compatibility: whatever format you throw at it, it will play. That is a developer's ideal, the user who never has to think about what's underneath. Fifteen years later Pomel would build a company with the same promise aimed at infrastructure. Whatever you run and wherever you run it, Datadog would see it, and the integration count would climb past 750.
After IBM came the internet startups, Neomeo and Silicongo, and then the dot-com bust, which Datadog's official retelling lists between "college to IBM" and "eventually, Datadog" as casually as a layover. The two Frenchmen had worked together, by Pomel's count, at four companies before they started their own. When journalists later called Datadog the story of "two French engineers in New York," they were half right. It was the story of two French engineers who had spent a decade in New York learning what it is like to be on opposite sides of the same machine.
Datadog17
The longest of those collaborations took place at a company most people in enterprise software have never heard of. Wireless Generation built data systems for K-12 teachers: assessment tools, reporting software, the plumbing that let a third-grade teacher see which students were falling behind in reading. Pomel became its vice president of technology and grew the development team, as his official biography puts it, "from a handful of people to close to 100 of the best engineers in New York." Lê-Quôc became director of operations and built the team and infrastructure that served more than four million students in 49 states. The two of them spent roughly nine years there, on opposite sides of the deployment pipeline.
They did everything managers are told to do. "We knew each other very well," Pomel later recalled. "We hired everybody on our teams. We scaled them from scratch, basically. We had a strict no jerk policy." The friendship at the top was real. The hiring filter was real.
"And yet," he went on, "we ended up a couple of years down the road with developers that hated operations, operations that hated developers, finger pointing, all of the things that you can imagine."
In a later telling he made it even plainer: "Development hated operations, operations hated development. Everyone pointed fingers." Two friends had built two departments out of nice people, and the departments had come to despise each other anyway. The cause wasn't character. It was the structure of the work and the information each side could see. Developers wrote code and threw it over the wall. Operators kept it alive in production and absorbed the blame when it fell over. Each side worked from its own picture of the system and its own logs, and neither picture showed the other side's pain.
Somewhere in Wireless Generation's infrastructure sat the thing everyone was afraid of. "At our previous company," Pomel told an interviewer from RTP Global, "we had an Oracle database that everyone feared. That database was called Datadog17." When the two men started sketching what would become their company, they used the name of the feared database as a codename. "People remembered it. So, we dropped the 17 and kept Datadog." Asked about the obvious canine connection, he said: "Well neither of us had a dog!" (The company later acquired a mascot, a dog named Bits, which is the kind of thing that happens when a joke outlives the people who were in on it.)
The name records something the founders understood early. A monitoring company exists because of fear: of the database that might fall over, of the pager going off, of the moment developers and operators meet in a postmortem to argue about whose fault it was. Datadog17 was the thing nobody wanted to look at. Datadog would be the thing that made everyone look at it, and look at it together.
"The starting point," Pomel has said repeatedly, "was, why don't we give all of those teams the same viewpoint?" In the fuller version: "We wanted to build a system that brought the two sides of the house together, all of the data was available to both and they speak the same language and see the same reality." He has always been careful to say the cloud was not the idea. It was "more of a trigger," a new environment in which developers could provision infrastructure "without needing to ask for permission, or file a form," which forced the two tribes into the same room whether they liked it or not. "We didn't quite understand at the time that the cloud was going to be so big," he admitted. "We thought, hey, it looks cool."
In 2010, Wireless Generation was sold. On November 24, The New York Times ran the news under a headline that put the buyer's other recent hire ahead of the deal itself: "News Corp., After Hiring Klein, Buys Technology Partner in a City Schools Project." Joel Klein, the former New York City schools chancellor, had gone to work for Rupert Murdoch, and the edtech company that had helped run the city's student data went with him into the empire, for $360 million. By then Pomel and Lê-Quôc were already gone. Datadog had been founded that June.
They left one of the most conventional exits in New York tech, a respectable sale to a media conglomerate, to start something nobody in New York had any reason to believe they could pull off.
Louder and Slower
On July 27, 2010, five companies from a New York accelerator program called SeedStart presented to more than a hundred venture capitalists, angel investors, and corporate strategists who had come in from New York, Boston, Philadelphia, and Washington. A local blogger listed them in alphabetical order: DataDog, Introspectr, Lexeem, Reducify, Risktail. Only the first has a Wikipedia page. The others are now just names on an archived web page, which is the usual fate of a demo-day company and worth remembering when people tell this story as if it was always going to turn out this way.
It was not obvious. "From the beginning," Pomel said at Slush in 2024, Datadog "was not seen as an obvious winner." The seed round eventually came together from New York money and a few independent believers: NYC Seed, Contour Venture Partners, IA Ventures, the investor Jerry Neumann, and Alex Payne, among others. One analysis of the company says it went almost a year without real funding. The founders divided the work along familiar lines. Lê-Quôc talked to the DevOps community, his own people, and to potential customers. Pomel went looking for money.
His account of that search is one of the best scenes in his public record, and he tells it without much self-pity:
"I didn't live in the Bay Area so I'd fly in, take a taxi to meetings, drag my little rolling suitcase up and down Sand Hill Road. Sometimes it rained. Sometimes I'd be stuck in a sad Starbucks between pitches. At the end of the day, you get all the polite rejections: 'too early for us,' 'we love the idea, but not the right time,' and 'let's stay in touch.' It was horrible."
The objections came in three kinds, and only one was technical. The first was geography. "Whenever we pitched West Coast investors," he told Matt Turck at a Data Driven NYC event, "it was sort of seen as a form of mental deficiency to be based in New York and doing infrastructure." New York in 2010 was seen as a place for ad tech, media, and e-commerce, startups that sold things or sold attention. Infrastructure was the Valley's business. The second was credentials. "Neither my co-founder or myself came from Google or Facebook," Pomel said, "so we didn't have the credentials for helping companies run at very large scale." Investors asked how nine years at an education startup could possibly be relevant to data infrastructure, as though four million students in 49 states were not a scale problem.
The third was nationality, and here the record turns comic. Martin Mignot, a partner at Index Ventures, later summarized what Pomel had heard on the road: he "was asked if French people only think about taking vacations." Some American investors, Pomel joked, "would seemingly speak to him a bit more loudly and slowly, to make sure he fully understood." So here was a man who had lived in New York for a decade, co-written one of the most widely installed media players in the world, and run a hundred-person engineering organization, sitting in a conference room in Menlo Park while someone explained venture capital to him at reduced speed.
Y Combinator turned them down. Paul Graham, YC's co-founder and the most influential early-stage judge in the industry, questioned what the company's first product actually was. That was a fair point, as later sections will show. Pomel kept the rejection email. Years after that, Datadog received an invitation to a YC dinner in Paris, which was withdrawn once the organizers realized Datadog wasn't a YC company. "I got rejected twice by YC," Pomel says, deadpan.
— Olivier Pomel, SaaStr Europa, ParisWe didn't have millions and millions of dollars, we didn't have a whole suite of people that were repeating to us that we were geniuses so they could invest in our company. So we basically had to focus on the problem.
The company's legal records include an employment offer letter dated May 20, 2011, addressed to "Alexis Le-Quoc," with the home address redacted. It offers him the title of Chief Product Officer and "a starting salary at the rate of $90,000 per year," says he will report to the Chief Executive Officer, and notes that "your employment with the Company will be 'at will,' meaning that either you or the Company may terminate your employment at any time and for any reason." The two men had known each other since college and worked together at four companies, and here their partnership is formalized in boilerplate, at a salary below what either had presumably earned in Wireless Generation's senior ranks. It was later filed with the SEC as an exhibit to a quarterly report. The document is dry. It is also the most honest record we have of what Datadog was a year into its life: two friends paying themselves a modest wage while they figured out whether the thing they were building was the right thing.
Real institutional money arrived in November 2012: a $6.2 million Series A co-led by Index Ventures, with partner Shardul Shah leading the deal, and RTP Ventures. More than a decade later, sitting across from RTP partner Jules Schwerin at the firm's 2023 Founder Summit in New York, Pomel was still grateful in the specific way of someone who remembers the sad Starbucks. "We were very thankful for the investment from RTP," he said. "It put us on the map at the time."
He has also been plain about what the rejections did for him. "Failure to raise initially," he has said, "was what made us." All those constraints "are how we built the culture of the company, how we found certainty about what it is we're building and for whom." And once, more broadly: "Anything that is a constraint or a hardship actually is good for the future if you survive it." The qualifier in that sentence carries most of the weight. Pomel doesn't romanticize hardship. He just declines to waste it.
Six Months Without a Line of Code
Two engineers start a software company and spend the first six months not writing software. It sounds like the setup to a joke about French labor law. It was the most consequential decision they made.
"For the first six months of the company," Pomel recalled, "we didn't write a line of code, which took some restraint because when you start a company, you're super excited, you want to get it done, you're thinking of all those beautiful things you can build." What they did instead was talk to potential customers and users, at length and in detail, and they discovered something nearly every founder learns too late: the window for candor closes the moment you have something to sell.
"When you don't have anything to sell, everybody is super happy to talk to you. You'll get hours and hours of really fantastic people at fantastic companies and I spend all that time explaining to you what their problems are, what's working, what's not working for them. They'll be extremely candid." Once a product exists, "you're tainted, you have too much of a vested interest and you're trying to push something, so people won't open up so easily."
Behind the restraint was a particular fear. "Our biggest fear," Pomel said, "was not to manage to build software. It was to build the wrong thing." These were two people who had shipped software for a decade and were sure they could build. Their worry was about direction. Building was the easy part.
Even with six months of listening, they built the wrong thing at first, or a version of it. The first alpha took six months to build and went to a hand-picked group of customers, "the best people at the best companies," and it was "way too open-ended, way too general." It was a collaborative data platform, a beautiful idea of a shared space where everyone could see everything. Users loved it and did not pay for it. "There was confusion on how to charge for the product," as a summary of his Slush 2024 talk put it. People "struggled to justify its value to their superiors." This is where Paul Graham's objection lands: a platform with no first product is a platform nobody can buy.
Two corrections followed, and the company has been built on both ever since.
The first was about distribution. The hand-picked alpha had been, in Pomel's word, "limiting." Choosing the best customers meant guessing who would have bandwidth, urgency, and the right problem at the same moment, and the guesses were usually wrong. So they opened the product to a wide beta, and "all of a sudden everything changed." He spelled out the reasoning: "It's a lot easier for users to self-select and to go and start using your product than for you to understand for whom you're going to be the right thing at the right time." Founders cannot predict which customers will be ready at a given moment. Customers can.
The second correction was about language. Datadog stopped calling itself a new thing. "People love the idea of a new thing," Pomel explained years later, "but they have no idea what they should pay for." So the company described itself as infrastructure monitoring, a category every engineering manager already had a budget line for and every engineering manager's boss already understood. As far as anyone can tell from his telling, the product didn't change much. The noun did. When a podcast host posted the clip, one commenter summed it up: "Go after the existing budget item. So much easier. People kill themselves trying to become category creators."
Pomel has never believed in the minimum viable product, at least for what he builds. "I actually don't think, for what we do, an MVP exists," he told the Paris audience. A monitoring tool has to monitor everything, and none of its features is revolutionary on its own, "but you need to have them, otherwise, you cannot be minimally useful." He thinks of it as a continuum rather than a launch: "you keep adding those features, and then at a certain point you have enough of them that customers can start buying." The hard part is knowing where that point is, and the only way to know is to keep going back to the customer.
In this account the cloud was still something serious enterprises dismissed. "Cloud was still considered a 'toy,'" RTP wrote, recalling 2010, "and most VCs weren't buying in." Datadog bet on the toy and on the people who were playing with it. When the toy became the infrastructure of the global economy, the company was already holding the map of it.
The Right to Leave
For a long time Datadog would not let customers sign annual contracts.
Most young software companies would treat that as malpractice. The annual contract is how a startup turns hope into a number a venture capitalist can underwrite, and Datadog, which had trouble raising money, needed numbers more than most. Pomel refused anyway, and his reason was less about revenue than about hearing.
"For a very long time we didn't sell yearly deals. We only sold month to month, meaning that customers had the opportunity to churn all the time. Meaning, if something is wrong with a product or if we don't solve the right problem or it's not valuable enough, then we know right away." With annual terms, he said, "you're going to have the bad news, but a year later, and by then you've wasted a year."
In other words, the contract structure was a monitoring instrument pointed at the company itself. Monthly billing worked like a heartbeat check: every thirty days each customer reported, by staying or leaving, whether the system was healthy. An annual contract is a sensor that reports once a year, long after the failure. Pomel set up his revenue model the way his product told customers to set up their servers, with short polling intervals, no blind spots, and alerts that arrive while there is still time to act.
He has a phrase for this, and it comes up whenever he is asked for the one lesson. Look for the bad news.
"You're not going to learn anything if you go and meet your customers and you say you lead by saying, isn't it great? And they say, yeah, yeah, it's kind of great. And then you find out actually, it's not all that great, and they're going to churn." The bad news is the useful part. It is "a thread to pull," he said, "it tells you where you need to go."
The rest of the early operating model followed the same logic of exposure. Every engineer at Datadog went on a support rotation for one week a year. Professional support engineers handled most of the load, but every developer spent a week dealing with actual customers' actual problems. Pomel describes the emotional arc with a straight face: "Engineers are super happy when it starts and they're even happier when it ends." What they take away is empathy, and more specifically "the consequences of their choices." A feature that looked brilliant in the design review turns out, from the customer's side, to be "a bit confusing." In his telling, the engineer's conclusion goes: "It was a great idea. But we're going to change it."
Then there were the conferences. Datadog sold to companies moving to the cloud, and in the early 2010s everyone moving to the cloud was new to it, so everyone went to conferences. Datadog exhibited, and instead of staffing the booth only with salespeople, it rotated engineers through, so that each might spend a full day at one, two, three, maybe five shows giving demos and answering questions. Pomel knew what he was asking. "One thing that's not natural when you're an engineer, like me, actually, is going and seeking other people's attention." So the company split the job. Other team members drew people in and handed them to the engineers, who then never had to start a conversation, only continue one. "There's never a shortage of people talking to them." It's a small arrangement, and it shows him designing around his own temperament rather than demanding his people change theirs.
Datadog did not hire its first product manager until it had more than a hundred employees. When it finally built a product management function, Pomel defined the job against the usual idea. "Product managers are not here to invent product. They're here to spend the most time possible with the customers, so outside of the office." The PM's work is to understand the problem, the value of solving it, what customers use and don't use, what they pay for and why, and then to test the product against all of that. In the Datadog sense, a PM is a professional listener with a budget.
Pricing, which most companies treat as where the friendship with the customer ends, he treated as another instrument. "There are two kinds of products, two kinds of companies," he said. One is the low-end disruptor whose main feature is being cheaper, and whose customers will push forever to make it cheaper still. The other is the high-value product, where "the cycle with your customer is to deliver more and more." You have to agree with the customer on which kind you are, "because otherwise you're never going to meet." After that, price becomes a source of signal: "Asking for money and putting a dollar amount on the product really focuses the mind for the customers and it helps you get really, really good feedback."
Nearly a decade later, with the company valued at more than $55 billion, Pomel described another of these deliberate counterweights. The bottom half of Datadog's customers, by spend, accounts for about one percent of revenue. Any spreadsheet would say to drop them. He keeps them, because they force product discipline: they are the first to show when a feature has become too complicated for ordinary use. They are paying customers, and they also work as a canary.
Seven Billion Dollars and a Comment Letter
The middle years of the story read like a list of rounds, offices, and floors, and the list itself is revealing.
In 2014, Datadog extended its monitoring across Amazon Web Services, Microsoft Azure, Google Cloud Platform, and Red Hat OpenShift, and added OpenStack. It refused to bet on a single cloud vendor at a time when betting on one seemed sensible. That same year it raised a $15 million Series B led by OpenView Venture Partners. In January 2015, Index led a $31 million Series C, and the company integrated with almost a hundred commonly used technologies. In February 2015 it made its first acquisition, Mortar Data, which tied analytics to monitoring. In October 2015 it opened a research and development office in Paris. The intern who never went home sent part of the company back.
It opened 2016 with a $94.5 million Series D led by ICONIQ Capital, one of the largest venture rounds raised by any New York company that year. Eight years after it was said to show mental deficiency, a New York infrastructure startup was among the city's largest fundraisers. That spring Datadog moved its headquarters onto a full floor of the New York Times Building on Eighth Avenue, a team that doubled during the year settling into the newspaper's tower. It also released a beta of application performance monitoring, giving it for the first time what the industry called full-stack coverage: infrastructure and code, operations and development, the two halves of the old Wireless Generation fight shown on one screen. In 2017 it bought Logmatic.io, a Paris log-management company, and gained a third pillar.
By the summer of 2019, the SEC's comment letters on Datadog's draft registration statement were addressed to "Olivier Pomel, Chief Executive Officer, Datadog, Inc., 620 8th Avenue, 45th Floor, New York, NY." The letters are usually dull, and these are mostly dull too, with one exception that fits the company too neatly to skip. In a letter responding to the July 30, 2019 amendment, SEC staff questioned how Datadog calculated its monthly run-rate. The company had included "additional usage" in it. If a customer used more than its commitment in a given month, the staff asked, "please clarify whether that customer is contractually committed to similar levels of usage in the following months. If not, tell us how you determined that it is appropriate to annualize the revenue related to such additional usage."
Put plainly, the regulator was asking whether a customer who is free to leave counts as revenue. That was the same question Pomel had built into the business from the start and answered the opposite way from most of his peers. Monthly freedom, usage that rises and falls with the customer's own infrastructure, a relationship that has to be re-earned continually: the qualities that made the model a sensor also made it hard to fit into the categories of a securities filing. Another letter asked about the treatment of goodwill and deferred contract costs in a pro forma net tangible book value of $60.6 million. The accountants wanted the company to sit still long enough to be measured.
On September 18, 2019, the day before the IPO, Bloomberg reported that Cisco had offered more than $7 billion to buy Datadog outright. One later analysis put the figure at $8 billion, more than twenty times that year's revenue. For two French engineers who had dragged a suitcase up Sand Hill Road and been asked about their vacation habits, it was a vindication that would have been easy to accept. Pomel turned it down.
— Olivier PomelTo grow to a large company, one of the hardest thing to do is not to sell along the way.
The next morning Datadog listed on the Nasdaq under DDOG. It sold 24 million shares, raised $648 million, and was valued at $8.7 billion. The stock rose about 37 percent on its first day, and the company closed near $10 billion in market capitalization, more than Cisco had offered. Back in France, a recruiter posted congratulations on LinkedIn to the two CentraleSupélec graduates of the classes of '99 and '97, whose company was worth more than $10 billion after two days of trading. The comments filled with bravo and impressionnant and one woman excited to discover that she had attended the same high school in Besançon as Lê-Quôc.
Matt Turck, the FirstMark investor who organized the Data Driven NYC series, noticed something about the city a few months later. Of New York's three largest public technology companies, two, Datadog and MongoDB, were infrastructure software businesses. The supposed deficiency was now a local specialty. Turck also noted, accurately, that Pomel was "not one for gratuitous self-aggrandizing" and had "given surprisingly few interviews over the years." Three hundred and fifty people packed the room to hear him.
Don't Be Evil Is Not on the Wall
When Datadog had seven hundred people, it had no written list of values. When it had eight thousand, it still didn't, and Pomel had stopped apologizing for it.
At the Paris talk he sounded a little sheepish, saying it was "something we'll do in the future because I think, as we scale, we will need that so we can better train." By the time he was talking to Y Combinator's audience, the companion company to the accelerator that had rejected him, the sheepishness was gone. "If you need 'don't be evil' to be written on the wall," he said, "you probably shouldn't work here." In his view, culture is "really about who you hire, who you fire, who you promote. That's what shows the culture." Anything else is decoration.
His dislike of decoration extends to language. Pomel has said he is allergic to the word "strategic," which he mentally translates into a two-syllable barnyard noun. In product reviews he describes his role less as a decider than as an editor. "I think this is too much," he says, or, "If I don't understand it, how will customers?" That question carries some weight. It comes from a man who co-wrote a media player, ran a hundred-person engineering team, and was once condescended to across conference tables in Menlo Park. If he is confused, the problem is not his intelligence.
The support inbox, the survey comments, the sampled sales transcripts, the Slack messages are not micromanagement, at least by his own account. He insists he never acts directly. The effect he is after is subtler. Once managers know the CEO is reading the raw feedback, they stop looking up to please him and start looking down at their own domains. A beautiful summary is no longer enough, because the person it's meant for has read the unedited version. Martin Mignot of Index described it carefully: Pomel taps "into samples of customer feedback, team surveys, sales contracts" but "resisted the urge to jump in and solve things himself."
The same discipline governs how the company talks about customers, which Pomel watches as closely as anything:
— Olivier Pomel, SaaStr Europa, ParisWhy are they doing this stupid thing with our product? The way we see it is if they're doing a stupid thing with our product, it's our fault. We let them do that. We made them think they should be doing that or we didn't explain it right enough.
Then comes the part that turns the policy from a nicety into a matter of epistemics: "Even if you think they're wrong, the fact that they're thinking something wrong in and of itself is a fact. It exists and you have to deal with it." A customer's misunderstanding is data. To dismiss it is to delete the one email that mattered.
He applies the same treatment to competitors. "Whenever I was tempted to read more about competitors and look at what they're doing, instead, I'd go and spend time with customers or potential customers. This way, you avoid overfitting what you're doing to what you think the competition is doing." The word overfitting is a machine-learning term, a model that has memorized its training data and can't generalize, and it is how an engineer would describe strategy.
On the 20VC podcast in 2021, Pomel described the CEO's role with a phrase that suits a man who once had to keep two warring departments from destroying each other: the CEO, he said, has to be the "equaliser in chief." He meant it structurally. In his telling, sales drifts toward the next deal and engineering toward the elegant long-term solution, and "if you want to be customer-focused, you can't be either of those." The CEO is the counterweight that keeps either from tipping the company over. He has also admitted that the hardest part of the job is speed in both directions, hiring faster and firing faster, the decision founders know they should make and delay anyway. In a culture defined by who you hire, fire, and promote, slow firing amounts to rewriting the values without telling anyone.
The people around him reflect that temperament. David Obstler, Datadog's chief financial officer, came to a company with no written values after more than three decades in operational finance. He holds a BA from Yale and an MBA from Harvard, worked in investment banking at JPMorgan, Lehman Brothers, and Goldman Sachs, and served as CFO at MSCI, RiskMetrics Group, OpenLink Financial, and TravelClick. Several of those businesses existed to measure the risk others couldn't see. He has spent a career translating messy operational reality into numbers investors can trust, and he works for a man who distrusts tidy numbers about his own product. Done well, the two make one instrument: one reads the raw stream, and the other tells the market what it means.
The show notes for that 20VC episode list under "Olivier's Favourite Book," simply, "Kurt Vonnegut." Not a title, just the author. Vonnegut's most famous novel is about a man who has "come unstuck in time," moving among the moments of his life out of order. It may be nothing. It may be an executive who would rather name a sensibility than a single text.
Buying People, Rewriting the Dog
At the Paris talk, Datadog had made two acquisitions, and Pomel described them in a way that would hold for the next twenty. "Both were heavily tilted towards the tech platform, the product and the team. So we acquired companies to build upon. We didn't acquire revenue streams to add to our sales engine." The question that mattered was what happened after closing. "Are people going to stick around? How long are they going to stick around? Are we going to be able to build on top of that team and turn that team of 20 we just acquired into a department of a company that's going to have 200 and 500 people?"
What followed is a steady record of small purchases. Madumbo, an AI testing platform, in February 2019. Undefined Labs, for CI/CD visibility, in August 2020. Sqreen, application security, closed in April 2021. Timber, the high-performance data pipeline, and Ozcode, live debugging, in 2021. Hdiv, Seekret, CoScreen, and Cloudcraft in 2022. Codiga and Actiondesk in 2023. Augmend in 2024. Quickwit, the open-source log search engine, in January 2025, followed that spring by Metaplane and Eppo. Propolis, AI-driven QA testing, in January 2026. Adaptive ML, a reinforcement-learning operations platform, in June 2026.
There are no headline mergers in the list. It is a run of teams of a few dozen people, each bought because it could seed a product line. By Pomel's account at Slush in 2024, about a third of the company's recurring revenue traced back to acquired products and teams, though the companies themselves had been bought at modest prices. The approach was opportunistic rather than market-driven: Datadog kept a broad product roadmap and bought when it found a team that fit the culture and the map, not when a banker announced that a category was hot.
The product strategy followed a similar rule, which Pomel puts plainly: build what customers are already building around you. Many of the twenty-odd products Datadog now sells started as things customers were assembling on their own on top of its APIs. Index Ventures described a three-bucket intake process: customer requests, executive analysis, and engineering experiments, each tested cheaply before the company decides how hard to push it. By 2024, Pomel said, the company had more than twenty products and fifty billable SKUs. A single open-ended platform that nobody paid for had become a platform of specific products that people did pay for, which is the outcome Paul Graham's objection had implied.
One decision from these years reads like a parable about restraint. Datadog's agent, the small program customers install on every server to collect data, started as a fork of an open-source Python agent written in 2009 by David Mytton for Server Density, a monitoring company previously called Boxed Ice. Mytton, a British engineer, had built the original as part of a business very like Datadog's. In that sense Datadog began its life inside someone else's code, adapted from a fellow traveler's work.
On February 28, 2018, Datadog released version 6.0.0 of the agent, rewritten from scratch in Go. Asked about it by a French technology outlet, Pomel answered with a principle before giving a reason: "En général mauvaise idée tout réécrire quand ça marche." Generally speaking, it's a bad idea to rewrite everything when it works. Then the reason. Compared with the state of the art a decade earlier, modern instrumentation collects "cent fois, mille fois, dix mille fois plus de données," a hundred, a thousand, ten thousand times more data. The agent sits on every customer machine, so its CPU and memory footprint has to be as small as possible. The rewrite was about performance, and Go had become the lingua franca of systems programming. He broke his own rule only when the bad news, in this case the cost of running the old agent at modern scale, had become a fact he couldn't ignore.
Two Quarters
In July 2025, Datadog joined the S&P 500, replacing Juniper Networks after its takeover by HPE. On the morning of July 9, Pomel went on CNBC and described his company in terms a ten-year-old could follow: "We help engineers ensure systems are fast, reliable, and secure." He said AI was driving explosive growth in software, and everything AI produced would need to be observed.
It was a slightly ironic thing for him to say, given the record. When generative AI first took hold, Datadog was by Pomel's own account reluctant to put AI in its products. The reluctance came from a monitoring company's instincts: a tool that tells you your system is broken has to be right, and early models were often confidently wrong. On Matt Turck's MAD podcast in September 2024, he explained how the reluctance gave way. First came Watchdog, the anomaly-detection layer. Then Bits AI, an assistant. Then Toto, Datadog's own time-series foundation model, trained on the kind of data the company had spent fourteen years collecting.
His view of trust is precise. Talking to Pigment's Perspectives podcast, he said humans can be effective at around 70 percent accuracy, because colleagues correct and work around each other's mistakes, but automated systems need accuracy "well above 90%" to be adopted. A security agent needs consistency. An incident-response agent needs speed. He described the larger change in a line that will probably define this phase of his career: "We are spending a lot less time building and a lot more time running applications." Code that once took years to write now takes minutes, and it runs across thousands of machines and hundreds of services. The scarce work is no longer making software but knowing what the software is doing.
That brings the story back toward where it began. In 2010 Datadog spent six months writing no code because its founders were listening. In 2026, Pomel told a Y Combinator interviewer that the company's engineering organization had been given two quarters to stop writing code by hand. His reasoning: he had watched individual developers, working with AI agents, rebuild in days systems that had once taken teams of six people six months. The company that began by deliberately not coding is being pushed by its founder toward a second period without hand-written code, this time with the machines doing the work.
The rest of the 2025 numbers show a company still doing what Pomel says companies should do, reinvesting. Revenue was $3.43 billion. Net income was $108 million. Operating income was slightly negative at minus $44 million, the consequence of a company that, as one product writer noted back in 2021, spends more on research and development than on sales and marketing. Index's Mignot pointed to the efficiency underneath: in 2023 Datadog had more than $2 billion in revenue and a 28 percent free cash flow margin, having burned only about $25 million to get there. In the early years the sales organization was built bottom-up, and a management layer was added only after enough individual reps were working at full capacity. Constraint was not just a story about the early days. It became an operating system.
Pomel's angel investments show the same instincts outside the company. He backed Hugging Face, whose co-founder Clément Delangue later shared a stage with him at DASH. He backed Mistral, Pigment, Dust, Dataiku, and Replit, and in June 2026 a seed-stage company called Niteshift, founded by Datadog veterans building an AI coding product. A good share of these are French, or Paris-founded companies that went to New York. When European founders ask him whether to start at home or move to America, his advice is consistent: target the U.S. market immediately, and have at least one founder move there to build local credibility. He is offering the same deal he took in 1999, minus the six-month return ticket.
A suitor reportedly came back too. In the summer of 2026, after a quarter in which Datadog posted $1.12 billion in revenue, social media posts circulated reports that Palo Alto Networks had explored an acquisition. A company once worth $7 billion to Cisco was now, by those accounts, valued at around $85 billion. Whatever happened in those conversations, Datadog was still independent at the time of writing.
The Email on the Wall
Sixteen years in, the two men from opposite sides of the campus network investigation still run the company together: one as chief executive, the other as chief technology officer, a title that has replaced the "Chief Product Officer" in a 2011 offer letter paying $90,000 a year. A chapter of one of Pomel's recent interviews is titled "20 Years With a Co-Founder," and if anything that undercounts. Asked how the partnership survived the years when nobody would fund them, he gave the most unsentimental account of friendship available: "It's good to have a co-founder, because your lows are not completely aligned. So there's always someone to pull the other up."
In March 2026 the two sat down together for a special episode of Datadog's own video series. The chapters move from how they met in college, to the problem they wanted to solve, to founding the company in New York, to Toto. Its title is direct: "What Datadog means to Olivier and Alexis." Neither of them, it should be noted, ever got a dog.
Pomel still goes back to Paris. He has spoken at Startup School Paris, at the Datadog Summit Paris, and at the conference where he admitted to deleting most of what he reads. In those rooms he is the cautionary example in reverse: the intern who stayed, the founder who was asked about vacations, the man investors addressed more slowly. He tells the young founders the same few things. Don't write code until you understand the problem. Charge early. Sell month to month if you can bear it. Read the complaints. Run toward the bad news.
Messages still come into his inbox all day: support tickets, survey comments, sales notes, a complaint about a confusing feature, a churned account, a customer who did something stupid with the product, which by his rule is the company's fault. He scans them. He deletes almost all of them. He never replies.
On his wall in New York hangs a rejection email from Paul Graham.
Part IIThe Playbook
What follows is an attempt to extract an operating system from a man who has deliberately refused to write one down. Pomel has no published list of values, no doctrinal letters, no management manifesto. His philosophy has to be pieced together from decisions and from the few, usually understated, interviews in which he explained them. The principles below come from that record. Most of them are versions of one idea: build instruments that deliver bad news early, and then make sure someone actually reads them.
The principles
- 1.Listen while you have nothing to sell.
- 2.Rename the product before you rebuild it.
- 3.Let customers leave every month.
- 4.Run toward the bad news.
- 5.Be the equalizer between sales and engineering.
- 6.Read the raw stream; act through the chain.
- 7.Put the builders in front of the buyers.
- 8.Hire product managers to listen, not invent.
- 9.Treat scarcity as an operating system.
- 10.Refuse the vindication offer.
- 11.Acquire for what happens after closing.
- 12.Return to startup mode with every new product.
D
From Datadog17 to the S&P 500
The decisions that compounded, in order.
1999
Pomel arrives in New York for a planned six-month IBM Research internship; he never leaves.
2010
Datadog founded in June; presents at SeedStart Demo Day on July 27. Rejected by Y Combinator. Six months pass without code.
2011
May 20 offer letter makes Lê-Quôc Chief Product Officer at $90,000 a year.
2012
$6.2M Series A co-led by Index Ventures and RTP Ventures.
2014
Multi-cloud support across AWS, Azure, Google Cloud, OpenShift; $15M Series B led by OpenView.
2015
$31M Series C; first acquisition (Mortar Data); Paris R&D office opens.
2016
$94.5M Series D led by ICONIQ; full floor of the New York Times Building; APM beta delivers full-stack visibility.
2018
Agent rewritten from scratch in Go (v6.0.0, February 28).
2019
Cisco offers more than $7B on September 18; IPO on September 19 raises $648M at an $8.7B valuation.
2025
$3.43B revenue; Datadog joins the S&P 500 in July.
2026
Acquires Propolis and Adaptive ML; engineering given two quarters to stop hand-writing code.
Principle 1
Listen while you have nothing to sell
Pomel's most counterintuitive observation from Datadog's founding is that customer candor is a resource that runs out, and its expiration date is your first sales pitch. Before Datadog had a product, prospective customers gave the founders "hours and hours" of frank conversation about their problems. Afterward, Pomel said, "you're tainted," and people "won't open up so easily."
So the six months without code were not hesitation. They were an attempt to use up a resource before it disappeared. Two engineers who could certainly have built something chose to spend the most valuable phase of the company's life learning what to build. The fear driving it was specific: not that they would fail to ship, but that they would "build the wrong thing."
Even so, the first alpha missed. Six months of listening didn't produce a correct product; it produced a correctable one. The founders had enough context to recognize that the open-ended platform was too general and to understand why. Listening didn't guarantee they'd be right. It made it possible for them to learn quickly when they were wrong.
Tactic: Before writing your first line of production code, schedule thirty customer conversations and forbid yourself from describing your solution in any of them.
Principle 2
Rename the product before you rebuild it
Datadog's first product was something users loved and nobody could buy. The problem wasn't the functionality but its classification: people "love the idea of a new thing but they have no idea what they should pay for." A buyer can't take "collaborative data platform for DevOps" to a boss who controls a budget. "Infrastructure monitoring" already has a line in the spreadsheet.
The move took discipline because founders like being category creators. A new category flatters the founder and confuses the buyer. Pomel picked the opposite: a familiar noun for an unfamiliar product, letting the differentiation show up after purchase instead of requiring it to be understood beforehand.
The lesson goes beyond positioning. Paul Graham's objection, that a platform needs a first product, and Pomel's answer, call the platform monitoring, are the same idea from two directions. Platforms are built by companies that first won in a category someone was already paying for.
Tactic: Find the existing budget line your product could replace or extend, and describe your company in that line's vocabulary until customers ask you what else you do.
Principle 3
Let customers leave every month
Month-to-month contracts are usually seen as a weakness, a lack of commitment that venture investors discount. Pomel used them as a sensor. "If something is wrong with a product or if we don't solve the right problem or it's not valuable enough, then we know right away." Annual contracts deliver the same bad news "a year later, and by then you've wasted a year."
The insight is that a contract structure is also an information structure. Lock-in doesn't stop dissatisfaction; it delays your awareness of it. Every month a dissatisfied customer stays locked in, the company keeps building in the wrong direction while its revenue figures look healthy.
Datadog moved to annual and multi-year terms as it grew into enterprise sales, and the SEC's 2019 questions about annualizing usage show how unusual the underlying model still looked. The principle survived the change. The company kept usage-based expansion as its main signal: customers deploying Datadog into more places, which Pomel said was the metric he actually watched.
M
Contract Terms as Feedback Latency
How long until bad news reaches the company?
| Structure | Signal latency | What the company learns | Verdict |
|---|---|---|---|
| Multi-year upfront | Years | That the customer was unhappy, after it is too late | Risk |
| Annual term | 12 months | Renewal or churn, with a year of drift baked in | Try |
| Month-to-month | 30 days | Whether the product is earning its keep right now | Scale |
| Usage-based expansion | Continuous | Where customers are deploying you, and where they are not | Scale |
Tactic: For any new product, offer a commitment-free tier for the first year, and treat every cancellation as a mandatory exit interview.
Principle 4
Run toward the bad news
Ask Pomel for one piece of advice and this is what he gives, every time and in nearly the same words. The phrase is simple; the hard part is practicing it. "The wrong thing to do is, you know there's some issue in some account and then you show up in front of the customer and you hope they're not going to bring it up." The right thing is to raise it first.
He frames bad news as a resource rather than an injury: "It gives you a thread to pull, it tells you where you need to go." For new products this means asking customers directly, "are you going to pay for it?", and preferring a clear no to a polite stall. "It's actually much better to hear from them, no, I'm not going to buy, and this is why, than to kick the ball down the road."
This is the core of Datadog's culture, and it is also what Datadog sells. An observability platform is a machine for surfacing bad news before it becomes a disaster. Pomel runs his company on the same principle he packages for customers, which may be why the culture has held without being written down: everyone at Datadog spends the day building the thing the culture describes.
Tactic: At your next renewal meeting or customer review, open with the most serious known problem in the account before saying anything good.
Principle 5
Be the equalizer between sales and engineering
"Everybody wants to be customer-focused," Pomel says. "The problem is that most companies end up being either sales-driven or engineering-driven." Each failure has a recognizable shape. Sales-driven companies become short-term, optimizing for "the next deal," which is "very often" not "what you want to do for the long run for your customers." Engineering-driven companies drift the other way, focusing "way too much on their solutions" and "the long term," until the gap back to the customer becomes hard to cross.
Pomel's answer is not a structure but a continuous correction: "It's a struggle of every day." He described the CEO as "equaliser in chief," the person whose job is to keep either function from capturing the company. This echoes the founding problem. Datadog began because two well-run, well-staffed departments at Wireless Generation had become enemies, and the founders concluded that the cause was structural.
Customer focus is not a third department. It is the absence of capture by either of the other two, and it needs active, constant maintenance from the top.
Tactic: Each quarter, list your five largest product decisions and mark whether each was driven by a deal, by engineering preference, or by observed customer behavior; if the third column isn't the largest, intervene.
Principle 6
Read the raw stream; act through the chain
Pomel's support-inbox habit is often cited and seldom understood. The striking part is not that he reads complaints; it's what he refuses to do with them. "I never actually act on them directly. I never go back to the customer, I never reply, I don't try to solve the problem." Instead he sends patterns down through management.
The method has two parts, and both are needed. The raw stream protects him from "beautiful summaries," the gradual polishing that happens as information moves up a hierarchy. Acting through the chain protects the organization from an executive who jumps in and solves things himself, which teaches managers that their job is to anticipate the CEO rather than understand their own areas. Done right, Pomel's attention changes managers' incentives. Knowing he has seen the unfiltered version, they stop managing up and start looking down.
The same reasoning explains why he never wrote down the values. A written list is a summary. The raw stream of hires, fires, and promotions is the culture itself, and that is what people actually see.
Tactic: Subscribe yourself to one unfiltered feed (support tickets, churn notes, or survey comments), spend fifteen minutes a day scanning it, and route every pattern you notice through the responsible manager rather than to the person involved.
Principle 7
Put the builders in front of the buyers
Every Datadog engineer spent one week a year on support rotation. Engineers worked the conference booths. The aim in both cases was to make developers see "the consequences of their choices" in the customer's own environment.
What sets Pomel's version apart is that he designed around engineers' temperaments instead of pretending they didn't matter. He knew from experience that "going and seeking other people's attention" doesn't come naturally to many engineers, so he split the booth job between people who drew visitors in and engineers who gave the demos. The goal was exposure, not a change of personality.
This is the founding idea of Datadog applied inside the company. Developers and operators hated each other because they saw different realities. Developers who never met customers would build for a different reality than the one customers lived in. The cure in both cases was a shared view.
Tactic: Put every engineer on at least one week of frontline customer support per year, and staff it so they handle problems rather than drum up conversations.
Principle 8
Hire product managers to listen, not invent
Datadog went past a hundred employees before hiring its first product manager. That restraint is instructive, and so is the job Pomel eventually defined: "Product managers are not here to invent product. They're here to spend the most time possible with the customers, so outside of the office."
In many software companies, product management has turned into internal coordination, a role spent in roadmap meetings and spec documents. Pomel moved it outside: the PM's main duty is to understand the problem, what solving it is worth, and why customers use or ignore what they already have. When product teams at Datadog decide what to build next, they draw from customer requests, executive analysis, and engineering experiments, and many products come from watching what customers were already building on Datadog's APIs.
The founders did this work themselves until it stopped scaling. Product management at Datadog is a way to replicate the founders' listening, not to replace their judgment.
Tactic: Rewrite your PM job description so that at least half of the role's measured time is spent outside the building with customers, and judge PMs on problem clarity rather than feature output.
Principle 9
Treat scarcity as an operating system
"Failure to raise initially was what made us," Pomel says. The rejections from Sand Hill Road, the snub from Y Combinator, and the year of underfunding forced the founders to focus on the problem because they had no money to spend on anything else. "Anything that is a constraint or a hardship actually is good for the future if you survive it."
What makes this a principle rather than a comforting story is that Datadog kept the habits after scarcity ended. By Index Ventures' account, the company reached more than $2 billion in revenue having burned about $25 million in total. Its sales organization added management layers only after individual reps were at capacity. It spends more on R&D than on sales and marketing. Constraint became the default, not an emergency setting.
The trap Pomel avoided is common: founders who struggle to raise money often overcorrect as soon as money arrives, as if trying to catch up with better-funded rivals. Datadog took a different lesson from its early years. The scarcity had been a good teacher.
P
Conventional Wisdom vs. the Pomel Approach
Where Datadog's operating choices diverged from the standard playbook.
| Conventional wisdom | Pomel's approach | Mechanism |
|---|---|---|
| Ship an MVP fast. | Six months of listening, no code. | Scan |
| Create a new category. | Sell into "monitoring," a known budget line. | Scale |
| Lock in annual contracts. | Month-to-month for years. | Scan |
| Hand-pick beta customers. | Open wide and let users self-select. | Try |
| Write down company values. | Culture is hiring, firing, promotion. | Scale |
| Take the strategic exit. | Refuse $7B+ from Cisco a day before IPO. | Risk |
| Acquire revenue. | Acquire teams that can become departments. | Scale |
Tactic: When a funding round closes, keep the hiring and spending rules you used before it, and require an explicit, written exception for every departure from them.
Principle 10
Refuse the vindication offer
The hardest offer to turn down is the one that proves you were right. For two French engineers who had been told New York infrastructure was a "mental deficiency," Cisco's bid of more than $7 billion the day before their IPO was not just money. It was the industry admitting it had been wrong. Pomel said no.
His explanation is short: "To grow to a large company, one of the hardest thing to do is not to sell along the way." The difficulty isn't financial; it's psychological. Every acquisition offer feels like a verdict, and founders who were long underestimated are especially tempted by a verdict that finally goes their way.
The decision fits the rest of the playbook. A company built on finding out the truth about its own value can't let a single offer stand in for that truth. Datadog closed its first trading day worth more than Cisco had offered, and joined the S&P 500 six years later.
Tactic: Before engaging any acquisition offer, write down the company you believe you can build independently in five years, and evaluate the bid against that rather than against the doubters you'd like to answer.
Principle 11
Acquire for what happens after closing
Pomel's acquisition rule hasn't changed since Datadog's first deals: buy "companies to build upon," not "revenue streams to add to our sales engine." The deciding question is about the future: will this team of twenty grow into a department of 200 or 500?
That explains the pattern in the acquisitions table: many small teams, few big revenue purchases, and a steady record of acquired groups becoming product lines. By Pomel's account, about a third of Datadog's recurring revenue now traces to acquired products and teams. The approach is opportunistic within a broad roadmap, which lets the company choose on cultural fit rather than being pushed into a category by market timing.
The key insight is that the deal model is the wrong place to optimize. The value of an acquisition is created in the years after it closes, by people who could leave. Structuring deals around retention and motivation is not a soft priority. It's where the value comes from.
Tactic: For every acquisition target, name the specific people who must still be there in three years for the deal to work, and structure the terms around keeping them.
Principle 12
Return to startup mode with every new product
Pomel names a specific risk that comes with success: "It's easy to basically get used to winning and then you expect everything's just going to work this way." When a mature company launches a new product or enters a new market, "all of a sudden you're back into startup mode and it's easy to not realize that."
His prescription is to make new-product teams behave "exactly as you would have when you were a tiny company": wide betas, fast feedback, and asking early whether customers will pay. He admits the tension. A large company doesn't want rough new products damaging the reputation of polished existing ones. He chooses the wide release anyway, because a new product kept behind velvet ropes learns nothing, just as Datadog's own hand-picked alpha taught them little in 2011.
The AI period is the biggest test yet. A company that once refused to ship AI because it couldn't be trusted now holds AI to an accuracy standard "well above 90%," and has given its own engineers two quarters to stop writing code by hand. That isn't recklessness; it's the 2010 method applied again. Go back to the customer, ship widely, ask whether they will pay, and run toward whatever bad news comes back.
Tactic: Give every new product line its own churn metric and its own explicit "will you pay for this?" milestone within ninety days of beta, separate from the core business's numbers.
Free playbook
Get The Business Model Playbook
58 business models, one visual page each: how the money flows, the metrics that matter, and who runs it. Free when you join the Faster Than Normal email.
Free. No spam. Unsubscribe anytime.
Part IIIQuotes and Maxims
In their words
— Olivier Pomel, SaaStr Europa, ParisMy piece of advice is really, look for the bad news. You run towards the bad news.
— Olivier Pomel, Data Driven NYC, in conversation with Matt TurckWhenever we pitched West Coast investors it was sort of seen as a form of mental deficiency to be based in New York and doing infrastructure.
— Olivier Pomel, SaaStr Europa, ParisWhen you don't have anything to sell, everybody is super happy to talk to you.
— Olivier PomelIt's good to have a co-founder, because your lows are not completely aligned. So there's always someone to pull the other up.
— Olivier Pomel, SaaStr Europa, ParisAsking for money and putting a dollar amount on the product really focuses the mind for the customers and it helps you get really, really good feedback on where you need to go.
Maxims
- Candor runs out. Customers tell you the truth only until you have something to sell them, so get it all before then.
- Speak the budget's language. A familiar category name will close deals a brilliant new category never will.
- Shorten the polling interval. Every contract structure is also a feedback structure; choose the one that brings bad news soonest.
- Bad news is a thread. A customer's clear no tells you where to go next; a polite maybe tells you nothing.
- The customer's mistake is your fact. If customers use your product wrong, you let them, and their confusion is data to act on, not dismiss.
- Read raw, act through the chain. Scan the unfiltered stream yourself, but route every fix through the people who own it.
- Culture is who you hire, fire, and promote. Values on a wall are a summary, and summaries hide things.
- Scarcity is a teacher, not a phase. Keep the habits hardship forced on you after the money arrives.
- The hardest offer to refuse is the one that proves you right. Don't let vindication substitute for ambition.
- Buy the team, not the revenue. The value of an acquisition is created by the people who stay after closing.
