How Olivier Pomel Built Datadog By Refusing Every Shortcut

This video features Datadog CEO Olivier Pomel recounting how his 2010 Y Combinator rejection—with Paul Graham's warning that "a platform is only as successful as its first product"—became "a little revenge on life" when the company went on to prove the platform vision viable. His durable partnership with co-founder Alexi, rooted in over a decade of friendship that began when Pomel enforced Alexi's campus ban for hacking, is sustained by rituals like agenda-free lunches and rigorous alignment on existential decisions, including turning down acquisition offers of $200 million and then ten times that amount before going public at roughly $7.5 billion—a company now worth around $80 billion. Lacking industry pedigree and passed over by nearly every VC, Datadog converted its outsider status into an advantage by attacking the fundamental disconnect between developers and operations rather than incrementally improving existing tools, though Pomel candidly credits much of the success to the lucky timing of the cloud adoption wave. His management philosophy centers on forcing employees to "look down instead of up," which he models personally by reading support requests, sales conversations, and every product brief daily to keep his sense of reality grounded. Decision-making at Datadog is governed by reversibility rather than importance, enabling rapid iteration that B2B customers make easy to test in contained ways before broader rollout. Finally, product expansion follows three paths—customer-driven incrementalism, top-down strategic bets, and bottom-up innovation—as illustrated when customers building their own APM and security tooling atop the platform signaled clear demand for native offerings.

• ▶ 0:33 Datadog co-founder and CEO Olivier Pomel applied to Y Combinator with his co-founder Alexi in 2010, made it to the interview stage, but was rejected — PG's core feedback being that a platform is "only as successful as its first product."
• ▶ 0:52 Pomel describes Datadog's eventual success as "a little revenge on life," since the team ultimately proved the platform vision viable despite the initial rejection.
• ▶ 1:19 The early fear of failure shaped Datadog's lasting culture — an obsession with building something valuable and a self-sustainable business that didn't depend on raising money — principles that remain central to the company to this day.
• ▶ 2:04 Pomel came to New York for what was meant to be a six-month IBM research internship after college, but ended up staying permanently — roughly 26 years — arriving just as the dot-com bubble was peaking.
• ▶ 3:02 He lived through both the dot-com boom and its full crash, including a period afterward when it seemed technology might never be valuable again; he stayed in New York throughout, as did his future co-founder Alexi.
• ▶ 3:43 He first met Alexi in school in France, where Alexi was caught hacking the campus network and Pomel personally enforced his "sentenced" disconnection — yet they later reconnected at IBM and went on to work together at four or five companies, with one always bringing the other along within a few months (▶ 4:12).
• ▶ 4:38 Pomel attributes the durability of his Datadog co-founding partnership to over ten years of shared history and genuine friendship established before the company existed, plus deliberate rituals like agenda-free lunches to keep open conversation flowing.
• ▶ 5:12 The co-founders treat alignment on existential decisions — especially "sell or not sell" — as critical, engaging seriously with real offers while spending extensive time ensuring neither founder is hiding their true preferences.
• ▶ 6:43 Datadog refused offers of $200 million, then ten times that amount, before going public at roughly a $7.5 billion valuation; the stock roller-coastered up to $10–12 billion and now sits around $80 billion, with each rejection guided by a checklist asking whether there's still 5x–10x more upside worth pursuing.
• ▶ 8:23 Datadog's founders lacked industry pedigree (no systems management or hyperscaler background), and nearly every VC besides Y Combinator passed on the company — a "very humbling experience," though some investors later joined in later rounds.
• ▶ 9:02 Being outsiders became an advantage: instead of improving an existing product category, they attacked the core problem that Dev and Ops don't communicate, aiming to unify them on one platform — a perspective the rest of the market wasn't seeing.
• ▶ 10:33 Pomel credits much of Datadog's success to luck, especially timing: in 2010 AWS was dismissed as a toy, but the broad and deep cloud adoption wave happened to place Datadog at its center — something no amount of hard work could have predicted.
• ▶ 11:43 To counteract employees' natural tendency to "manage up" with polished stories, Pomel designs every system at Datadog to force people to look down instead of up—at what's actually happening with customers, the product, and the technology.
• ▶ 12:01 He personally samples ground-level reality daily by reading support requests, sales conversations, and all product briefs and releases; these occasional direct questions jolt local managers into genuinely understanding their areas, while giving him a sampled sense of reality to check whether top-level stories match what's actually occurring.
• ▶ 13:41 His involvement doesn't create a bottleneck because executive feedback is possible but never required—teams ship freely through regular communication points, with pre-approval reserved only for narrow, hard-to-reverse, high-impact changes like pricing and packaging.
• ▶ 14:25 Pomel's decision-making heuristic: low-stakes decisions require no deliberation ("nobody's waiting for anyone"), and even high-stakes decisions move fast if they're "easy to reverse" — so reversibility, not importance, is the real bottleneck.
• ▶ 14:37 The B2B model structurally enables fast iteration because it's "very easy to test some products with some customers" — changes can be scoped to individual customers, containing mistakes and delivering quick feedback.
• ▶ 14:45 This incremental approach is how Datadog builds in general: start small with a limited set of customers, validate whether the change improves things significantly, then roll it out more broadly.
• ▶ 15:14 New product decisions at Datadog are driven by observing actual customer behavior — the situations they put the product in and the extensions, scripting, and custom tooling they build around it — rather than internal brainstorming.
• ▶ 15:29 The expansion into APM was validated in advance: before Datadog built native tracing, customers were already constructing simple versions of APM themselves on top of the platform, making it "pretty clear" they wanted that functionality natively.
• ▶ 15:44 The same pattern repeated with security — customers were using Datadog for security automation and instrumenting their own security posture, signaling an unmet need and paving the way for Datadog's security offerings.
• ▶ 16:14 Customer feedback drives the majority of Datadog's product changes — it's generally reliable with enough customers, but tends to be very incremental (e.g., fixing buttons, adding integrations).
• ▶ 16:46 A second source is top-down "strategic" bets into new areas beyond the core business, justified by conviction and proof points; Pomel admits you'll be wrong more often this way, but believes it's necessary.
• ▶ 17:07 The third source is complete bottom-up innovation, where engineering or product teams build solutions to problems they themselves encountered and believed needed solving.
• ▶ 17:18 Datadog's products emerge through diverse bottom-up paths — internal needs, customer-inspired experiments, or simply an employee wanting to build something — often without originating from leadership or being backed by customer demand data.
• ▶ 17:39 Employees who experience a problem firsthand and solve it themselves have generated their own validation, since the builder was also the user.
• ▶ 17:43 All three innovation channels — customer feedback, deliberate strategic direction, and protected space for internal experimentation — must be preserved simultaneously so none crowds out the others.
• ▶ 17:56 Datadog operates at significant scale with about 25 commercially available products and roughly 8,000 employees — a size the interviewer notes makes it "a complex organization to get right."
• ▶ 18:11 Pomel explains that customers adopt Datadog's products much like they adopt services from a cloud provider such as AWS — through one unified relationship rather than separate evaluations — and accepts the interviewer's claim that Datadog executes this model "even better than AWS."
• ▶ 18:25 The strategic payoff is that products land within a single integrated platform, eliminating the need for 25 separate sales processes or distinct sales teams, which keeps the organization "much simpler" despite its breadth.
• ▶ 19:04 Datadog's IPO lockup expired the very day pandemic lockdowns began, coinciding with a massive market crash — an unlucky baptism into public-market volatility that Pomel says ultimately lasted longer than anyone anticipated.
• ▶ 19:28 Pomel distinguishes stock turbulence from business health, describing the period as a "seesaw" of crashes and quick rebounds; such gyrations are simply part of being a public company.
• ▶ 19:45 Datadog deliberately educates employees about volatility through quarterly all-hands meetings, closing each with the "voting machine vs. weighing machine" slide to keep focus on execution fundamentals rather than short-term stock moves.
• ▶ 20:27 Despite technically leading the observability market, Datadog holds only ~13% of it, and Pomel argues that because the market is growing fast, "the bulk of the opportunities is still ahead" — incumbency is strength, not saturation.
• ▶ 20:52 Pomel argues AI-driven business model disruption largely doesn't threaten Datadog: its usage-based pricing insulates it, whereas seat-based competitors must rethink their economics as AI automation erodes per-seat value.
• ▶ 21:25 Datadog pursues a dual AI strategy — "Datadog for AI" (serving both AI-native builders and AI adopters) and "AI for Datadog" (embedding AI into its own product for automation).
• ▶ 21:36 Datadog flipped the disruption narrative: once feared as "the incumbent that was going to be disrupted by AI," it instead became a winner of the AI change, since AI brings more customers, bigger customers, and new products.
• ▶ 21:45 Olivier Pomel confirms there is now far more demand — more software, more infrastructure, "more everything" that keeps growing — with much of Datadog's business coming directly from AI companies.
• ▶ 22:02 Datadog's AI-native footprint is elite: it serves the top 10 AI companies in the world (all growing fast) plus a good fraction of the many other startups building and scaling rapidly in the AI era.
• ▶ 22:19 Olivier Pomel explicitly confirms that Datadog saw a "substantial reacceleration of the business," marking a meaningful inflection point in its growth trajectory rather than a marginal uptick.
• ▶ 22:23 Pomel frames this reacceleration against the broader industry context, urging listeners to "zoom out" and consider the massive wave of complexity the AI era is imposing on companies.
• ▶ 22:29 He identifies two layers of that complexity: the enormous GPU infrastructure buildout underpinning AI workloads, and the models themselves, which are "really hard to manage and monitor and secure" — challenges closely aligned with observability needs.
• ▶ 22:45 AI-generated code is exploding in volume, but engineers don't actually know or understand the code being produced — "you can't even read the code anymore" — since they haven't written it themselves.
• ▶ 22:55 The strategic consequence is that economic and technical value is shifting away from writing code toward validating, running, and operating it.
• ▶ 23:00 This shift plays directly to Datadog's strengths because it already lives in production and in contact with the real world — a big opportunity ("tons of opportunity") that nonetheless requires substantial product investment to shepherd code and applications through every stage of their lifecycle.
• ▶ 23:29 AI represents a completely new space of potential products to build for Datadog, though it comes with acknowledged competitive risk.
• ▶ 23:46 The observability/infrastructure monitoring industry has marketed "AI" features for 20–30 years, but Pomel is blunt that it never quite worked until now.
• ▶ 23:52 Today marks an inflection point where real automation is finally achievable — the long-sought vision being that instead of waking a human engineer at 2:00 a.m., the machine fixes the problem overnight and reports back in the morning.
• ▶ 24:20 AI's biggest internal impact at Datadog is on the development side of the business — a pattern Pomel says holds across most software companies — with everyone in the company using AI coding tools.
• ▶ 24:33 Engineering offers far more productivity upside than functions like sales: AI may only modestly improve sales productivity, while in some parts of engineering it could multiply it by two, three, four, or even ten times.
• ▶ 24:58 Illustrating how advanced AI-assisted development has become, Pomel's co-founder publicly predicted that within two quarters Datadog would mostly not be writing code at all.
• ▶ 25:14 Datadog experienced an industry-wide "epiphany in December," shared by almost all tech companies, triggered by a step-change jump in new AI model capabilities combined with holiday downtime that let engineers discover how powerful the models had become.
• ▶ 25:40 Despite the enthusiasm, major open questions remain: what actually works with AI development, how it performs long-term (since nobody has historical precedent), and what appropriate team structures should look like in terms of PMs, designers, engineers, and security engineers.
• ▶ 26:06 Pomel believes distinct roles are converging, which directly shapes Datadog's strategy of building one unified platform to bring those roles together — framing AI development as an acceleration of familiar cloud-era consolidation trends rather than something entirely new.
• ▶ 26:36 All of Datadog's hardships ultimately helped the company survive and grow stronger — early rejections shaped its culture, and a serious security breach completely transformed its approach to security.
• ▶ 27:15 Pomel's second piece of advice to his younger self is to "always decide faster," applying to both product and people decisions — though product decisions get easier with experience while people decisions stay hard.
• ▶ 27:35 On hiring and firing specifically, leaders are generally too slow; his rule of thumb is that "whatever your instinct is, is too slow," so act even sooner than your gut suggests.
• ▶ 27:56 Datadog deliberately kept hiring conservative — even while "growing like crazy," it never expanded its engineering team by more than 2x (doubling) in any given year, a cap Pomel acknowledges was very conservative relative to the company's aggressive business goals.
• ▶ 28:17 The philosophy was shaped by cautionary examples among Datadog's own customers, whose engineering teams became "pretty dysfunctional" after scaling headcount roughly 10x within just 18 months.
• ▶ 28:30 A key cost of hypergrowth hiring is the collapse of short-term productivity: nearly all organizational energy gets consumed by recruiting and ramping new hires, leaving little capacity for actual output.
• ▶ 28:41 Scaling engineering headcount too aggressively risks catastrophic failure — Pomel warns that if Datadog grew too fast, "things are going to blow up."
• ▶ 28:43 The cost of such a breakdown is asymmetric and long-lasting: an overextended team that fails gets "set back for another year or two," making hypergrowth a net negative despite short-term gains.
• ▶ 28:47 Datadog knowingly accepted the trade-off of leaving business on the table by staying undermanned, prioritizing stability and execution quality over maximum revenue capture — culminating in the deliberate decision to cap engineering team growth at ▶ 28:52.
• ▶ 28:55 The excerpt opens mid-sentence ("engineering team"), wrapping up the prior discussion of team composition and headcount as a natural handoff into the new topic.
• ▶ 28:56 A speaker asserts that keeping team size fixed is "probably even more fruitful today" — framing it not as a mere constraint but as potentially more advantageous now than in the past.
• ▶ 28:58 The supporting reasoning begins immediately — "you can do so much more with the same..." — implying per-person productivity has risen sharply, so the same headcount can deliver far greater results than it historically could.
• ▶ 29:02 The host warmly thanks Olivier Pomel by name for joining the interview, expressing that it was "awesome to have you."
• ▶ 29:05 Pomel graciously reciprocates, thanking the host for having him and closing with a final word of gratitude.
• ▶ 29:00–▶ 29:07 The brief, friendly exchange serves as the formal endpoint of the interview, introducing no new topics but conveying mutual appreciation between host and guest.

• ▶ 0:05 Introduction & the 2010 YC Rejection (0:05 - 1:47) - Olivier Pomel is introduced as Datadog's CEO, and he recounts his 2010 YC application, rejection, and how proving doubters wrong shaped the company's culture.
• ▶ 1:48 From France to New York: Dot-Com Era and Meeting Alexi (1:48 - 4:19) - Pomel's journey from France to New York via IBM Research, living through the dot-com boom and crash, and how he met co-founder Alexi, beginning a long-running partnership.
• ▶ 4:20 Founding Datadog: Big Decisions, Refused Offers, and the IPO (4:20 - 7:32) - Co-founding Datadog, the secret to a lasting partnership, how the duo handles major decisions like turning down acquisition offers, going public, and the commitment founding demands.
• ▶ 7:33 Early Struggles and Riding the Cloud Wave (7:33 - 10:48) - The difficult pre-cloud early days, widespread investor rejections, their outsider perspective as an advantage, timing the decade-long cloud adoption wave versus today's fast AI adoption, and the mix of luck and hard work.
• ▶ 10:50 Founder-Led Culture and Staying Close to the Ground (10:50 - 14:25) - Pomel's hands-on involvement—reading support requests, sales conversations, and product briefs daily—to fight "managing up," jolt teams into reality, and his framework for which decisions truly require his feedback.
• ▶ 14:28 Product Strategy, Scale, and Life as a Public Company (14:28 - 29:09) - How Datadog decides what to build—from infrastructure monitoring to observability—drawing on three sources of product direction, its roughly 25 products, 8,000 employees, simplified sales model, and public-market volatility since the IPO.