Orbiter
Winter 2020InactiveML models automatically monitor business performance
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
ML models automatically monitor business performance
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Orbiter was a San Francisco-based startup that built ML-powered business metric monitoring for non-engineers. Founded by Mark Wai, Victor Zhang, and Winston Zhang — operators with backgrounds at Tesla, DoorDash, and Facebook — the company participated in Y Combinator's Winter 2020 batch and launched publicly in March 2020. Its core product automatically detected anomalies in business metrics and delivered Slack alerts without requiring any engineering involvement, positioning itself as the "PagerDuty for the business side."
The company wound down quietly around early 2022, roughly two years after launch. The primary cause was structural: Orbiter entered a capital-intensive market segment — data observability — at the same moment that much better-funded competitors were raising nine-figure rounds. A three-person team on a pre-seed budget could not build the enterprise integrations, sales infrastructure, or ML differentiation needed to compete against Monte Carlo's $236M war chest.[26]
No acquisition or acqui-hire was announced. The founders dispersed entirely: Victor Zhang and Winston Zhang co-founded Cega, a DeFi exotic options protocol, while Mark Wai began co-founding Inkly as early as January 2021 — while Orbiter was still nominally active.[20][22] The pivot to DeFi, a domain entirely orthogonal to data monitoring, suggests the founders concluded the market had consolidated around larger players before they could establish a defensible position.
The founding story of Orbiter is autobiographical in the clearest sense: the three co-founders built the product they had personally needed and failed to find.
Mark Wai, Victor Zhang, and Winston Zhang each came from operator roles at companies where data monitoring was a daily operational concern. Wai held senior positions as both a Staff Product Manager and Staff Data Scientist at Tesla, and previously worked on Messenger Growth and Product Analytics at Facebook.[5] Victor Zhang's background spanned product management roles at Tesla, Moat, Tumblr, Yahoo, and Google.[6] Winston Zhang came from a Product Manager role at DoorDash and prior management consulting work at Oliver Wyman.[7]
The founding insight was direct. In their own words, posted to Hacker News on launch day: "Before Orbiter, we were product managers and data scientists at Tesla, DoorDash, and Facebook. Even with tools like Amplitude, Tableau, and Google Data Studio, we would still catch real issues late by days or weeks."[4][25]
That experience — watching real business problems compound because no one had been alerted in time — was the founding thesis. The founders had seen firsthand that engineering teams were well-served by monitoring tools like PagerDuty, Sentry, and DataDog, but that the product and business side of organizations had no equivalent. A metric could drop 30% over three days and no one would notice until a weekly review meeting.[14]
The company's legal entity was incorporated as "InGoodCompany Inc," with "Orbiter" serving as the product brand.[3] The entity name suggests the founders may have had a broader company vision beyond the initial product, though no public record of what that vision entailed has surfaced.
The founding date is ambiguous. Crunchbase lists 2019, while YC's own database lists 2020.[0] The getorbiter.com domain was registered on January 22, 2020,[27] and Mark Wai's LinkedIn tenure begins in January 2020,[19] suggesting the most likely interpretation is that the founders were ideating or doing early customer discovery in 2019 and formally incorporated for the YC application cycle in late 2019 or early 2020.
How the three co-founders met is not documented in any public source. Given the overlapping Tesla tenures of Wai and Victor Zhang, a prior working relationship there is plausible — but this is inference, not confirmed fact.
Orbiter's core product was a metric anomaly detection system designed explicitly for people who could not — or should not have to — write code to monitor their business.
The setup flow was deliberately minimal. A user connected their existing analytics database using credentials, the same way they might connect Looker or a SQL editor.[10] No engineering involvement was required. The company claimed onboarding in under five minutes.[9] Once connected, Orbiter's ML models began learning the normal behavior of the company's key metrics — revenue, conversion rates, user engagement figures, or whatever the business tracked.
When a metric deviated from its expected range — a spike, a drop, an unusual pattern — Orbiter sent an alert directly to a connected Slack workspace.[11] The alert was immediate, not batched into a weekly report. This was the core value proposition: catching the problem on Tuesday afternoon rather than discovering it at Friday's review meeting.
The product included a pragmatic fallback for early-stage customers whose datasets were too sparse for ML to operate reliably. These users could set manual "guardrail" thresholds — for example, "alert me if this metric falls below 70%" or "alert me if this metric drops more than 5% week over week."[13] As the dataset grew, the ML layer would take over from the manual rules. This was a sensible onboarding wedge: it gave customers immediate value before the ML had enough data to be useful, reducing the cold-start problem that plagues many ML products.
Beyond alerting, Orbiter also automated root cause analytics workflows and incorporated user feedback to continuously improve model accuracy over time.[12] The feedback loop — where users could confirm or dismiss alerts — was a standard approach to supervised anomaly detection, allowing the model to learn which deviations were genuinely actionable versus noise.
The Slack integration was a deliberate distribution choice. In 2020, Slack had become the operational hub for most tech-forward companies. Delivering alerts inside Slack meant the product lived where decisions were already being made, rather than requiring users to log into a separate dashboard. TechCrunch described the product at Demo Day as "an automatic real-time monitoring and alert system integrated with Slack to ensure better customer service and revenue management."[15]
What made Orbiter distinct from existing tools was its explicit positioning against the engineering-side monitoring stack. PagerDuty, Sentry, and DataDog were built for software engineers monitoring system uptime and error rates.[14] Orbiter's target user was a product manager or business analyst who understood what the metrics meant but had no way to be automatically notified when they went wrong. The product did not require SQL knowledge, dashboard-building skills, or engineering support to configure.
No archived product screenshots, pricing tiers, or feature changelog are publicly available. The specific databases and data warehouses supported — whether Snowflake, BigQuery, Redshift, or others — were not disclosed in any public source. Whether the ML models were proprietary or built on top of open-source anomaly detection libraries is similarly unknown.
Orbiter's stated target was non-engineering business users at companies that already had analytics infrastructure — specifically product managers, growth teams, and business analysts who owned metrics but lacked the tooling to monitor them automatically.[9][14]
The implicit customer profile was a mid-size technology company: large enough to have meaningful metric volume and an existing analytics database, but not so large that it had already built internal monitoring tooling. Early-stage startups with sparse data were partially addressed by the manual guardrail feature,[13] but the ML value proposition required a baseline of historical data to function.
No public information is available on whether Orbiter's actual customer base skewed toward SMBs, mid-market, or enterprise. The self-serve onboarding model — five minutes, no engineering required — suggests the initial go-to-market was bottom-up, targeting individual teams rather than enterprise procurement processes.
The data observability and monitoring market was real and growing at the time of Orbiter's launch. Soma Capital's framing captured the commercial stakes: "companies are losing millions of dollars because they are too slow to detect real problems that impact their revenue and customer experience."[29]
The broader data quality and observability market was attracting significant venture capital by 2020–2022, with analysts citing the explosion of data infrastructure spending as companies moved to cloud data warehouses. The fact that Monte Carlo reached a $1.6B valuation by May 2022[26] and Bigeye raised $66M by September 2021 confirms that investors believed the market was large. Orbiter's specific sub-segment — business-side metric monitoring for non-engineers — was a narrower slice of that market, though the addressable universe of non-engineering business users at data-driven companies is substantial.
Orbiter's competitive position was more precarious than the market size alone would suggest, for reasons that were structural rather than merely executional.
The engineering-side incumbents (PagerDuty, Datadog, Sentry) were not direct competitors in 2020 — they were built for engineers monitoring infrastructure, not business analysts monitoring conversion rates. This gap was Orbiter's founding thesis. But it also meant these incumbents had strong incentives to expand into the business-side monitoring layer as their own products matured.
The analytics platform incumbents (Amplitude, Looker, Tableau, Google Data Studio) were the tools Orbiter's founders had personally used and found insufficient.[25] These platforms had the distribution advantage: they already sat on top of the same databases Orbiter wanted to connect to, and they had existing relationships with the exact users Orbiter was targeting. Adding anomaly alerting as a native feature was a natural product extension for any of them — and several did exactly that in the years following Orbiter's launch.
The data observability pure-plays — Monte Carlo, Bigeye, and Anomalo — were the most direct competitive threat. Monte Carlo, founded in 2019 (the same year as Orbiter), raised $236M and reached a $1.6B valuation by May 2022.[26] These companies were competing on the same axis as Orbiter — automated monitoring of data and metrics — but with orders of magnitude more capital for enterprise sales, integrations, and ML infrastructure.
Orbiter's position on the competitive map was caught between two dynamics simultaneously: the risk that analytics incumbents would absorb the feature natively (eliminating the need for a standalone product), and the risk that better-funded pure-plays would out-execute on the same problem. A three-person pre-seed team had no structural advantage on either dimension — not distribution, not data volume, not sales capacity.
Orbiter never publicly disclosed its revenue model, pricing tiers, or any financial metrics. The absence of any pricing page in archived sources, combined with the lack of any fundraising announcements beyond the initial YC pre-seed, means the business model must be inferred from context.
The self-serve onboarding model — five minutes, no engineering required — is consistent with a SaaS subscription structure, likely priced per seat, per connected data source, or per monitored metric. This is the standard model for analytics tooling targeting business users. A freemium or trial tier would have been consistent with the bottom-up go-to-market implied by the Slack integration and HN launch.
Directional unit economics (inferred, not confirmed): YC's standard deal at the time of the W20 batch was $125K for 7% equity. Soma Capital's investment amount is undisclosed. Assuming total funding in the range of $150K–$300K — a reasonable estimate for a pre-seed company with no disclosed follow-on round — and a team of three in San Francisco, annual burn would have been approximately $400K–$600K at market salaries, implying a runway of six to eighteen months from the March 2020 close. This is consistent with the company appearing to wind down around early 2022, roughly two years after launch, if the founders operated lean or supplemented runway with early revenue.
The unverified Latka figure of $490K in 2021 revenue,[24] if taken as a directional signal rather than a confirmed fact, would suggest the company achieved meaningful early traction — but not enough to justify or attract a Series A in a market where competitors were raising tens of millions. Latka's methodology is opaque and its figures for small private companies are frequently unreliable; this number should be treated as a rough order-of-magnitude signal only.
No Series A, seed extension, or follow-on financing was ever announced or found in any database.
The only quantitative traction signal available is the unverified Latka figure of $490K in revenue for 2021.[24] Latka aggregates self-reported or scraped revenue data from private SaaS companies, and its figures for small, early-stage companies are frequently imprecise. Whether this represents ARR, total annual revenue, or a projection is unknown. It should be treated as a rough directional indicator — the company was generating some revenue — rather than a confirmed metric.
The Launch HN post in March 2020 generated public visibility and a substantive discussion thread, which is consistent with a product that resonated with the Hacker News audience of technical operators.[28] However, no follow-up traction posts, customer announcements, case studies, or press coverage of growth milestones appeared in any subsequent period.
The absence of a public growth narrative is itself informative. Companies that achieve breakout traction in the YC ecosystem typically announce it — through blog posts, fundraising announcements, or founder tweets. Orbiter produced none of these after its March 2020 launch. This pattern is consistent with a company that achieved early revenue but did not reach the growth rate that would have warranted a Series A or public celebration.
No verified customer count, churn rate, NPS data, or sales motion details are available from any primary source.
Orbiter wound down quietly around early 2022, with no public announcement, no acquisition, and no founder post-mortem.[18] The failure was not a single catastrophic event but a slow erosion of competitive position in a market that was simultaneously validating the problem and consolidating around better-capitalized players.
The most important factor in Orbiter's failure was structural, not executional. The data observability market was real — Orbiter's investors, its competitors' investors, and the market's subsequent development all confirm this. But the market rewarded scale: enterprise integrations, a broad connector library, a sales team capable of navigating procurement, and enough ML infrastructure to handle large data volumes. These capabilities require capital.
Monte Carlo, founded in the same year as Orbiter, raised $236M and reached a $1.6B valuation by May 2022.[26] Bigeye raised $66M by September 2021. Orbiter's total disclosed funding was a YC pre-seed (standard deal: approximately $125K at the time) and an undisclosed amount from Soma Capital.[16][17] No follow-on round was ever announced.
The team attempted to address this disadvantage through product simplicity — the five-minute onboarding, the no-engineering-required positioning — which lowered the cost of customer acquisition and reduced the need for a large sales team.[9] But product simplicity is not a durable moat in a market where competitors can replicate the same simplicity with more resources. The attempt to compete on ease-of-use while better-funded competitors were building enterprise-grade platforms was a holding strategy, not a winning one.
Orbiter's product occupied a gap between two existing categories: the engineering monitoring stack (Datadog, PagerDuty) and the business analytics stack (Amplitude, Looker, Tableau).[14] The founders correctly identified this gap in 2020. The problem is that gaps between incumbent categories are inherently unstable — incumbents on both sides have strong incentives to close them.
Amplitude, which already had the user relationships and the underlying metric data, added anomaly detection and alerting features to its platform in subsequent years. Looker (acquired by Google in 2020) similarly expanded its alerting capabilities. When an incumbent adds a feature natively, the standalone product serving that function faces an existential question: why pay for a separate tool when the platform you already use does it for free?
Orbiter's Slack-first delivery mechanism was smart for 2020 but did not create a durable moat. Slack itself added workflow automation and alerting integrations that reduced the friction of building similar functionality without a dedicated product. The distribution advantage Orbiter gained by meeting users in Slack was available to any competitor willing to build the same integration.
No public record exists of Orbiter attempting to respond to this dynamic — for example, by deepening integrations with specific data warehouses (Snowflake, BigQuery) where incumbents were slower to move, or by focusing on a specific vertical where the monitoring problem was acute and underserved.
A telling signal in the timeline: Mark Wai began co-founding Inkly in January 2021,[22] while Orbiter was still nominally active. His LinkedIn tenure at Orbiter did not end until approximately January 2022 — meaning there was a period of roughly twelve months during which the CEO was simultaneously building a new company.
This is not necessarily evidence of negligence; founders sometimes run parallel projects, and the circumstances are not fully documented. But it is a signal that conviction in Orbiter's trajectory had declined before the company formally wound down. A CEO who has begun a new venture is unlikely to be pursuing the aggressive fundraising, enterprise sales, or product differentiation that would have been required to compete in the data observability market.
Victor Zhang and Winston Zhang's subsequent pivot to Cega — a DeFi exotic options protocol — is the sharpest signal of all.[20] DeFi and data monitoring share no customers, no technology, and no distribution. This was not an adjacent pivot or a narrowing of focus within the same market. It suggests the founders concluded that the data observability space had become structurally unfavorable for a small team — that the market had consolidated around larger players and that their position was not worth defending.
The data observability category followed a pattern common in B2B infrastructure markets: an initial period of fragmentation, followed by rapid consolidation around a small number of well-capitalized players. The window for a small team to establish a defensible position — through a specific vertical focus, a proprietary data advantage, or a unique integration — was narrow.
Orbiter launched in March 2020, at the beginning of this consolidation window. By the time Monte Carlo reached unicorn status in May 2022, the window had effectively closed for undercapitalized entrants. A company with $150K–$300K in total funding and a team of three had no realistic path to competing for enterprise contracts against a company with $236M and a full sales organization.
The founders never disclosed whether they attempted to raise a seed round after YC and failed, or chose not to pursue additional capital. Either scenario is consistent with the outcome: a failed raise would have confirmed the market's judgment that Orbiter's position was not fundable at competitive terms; a decision not to raise would suggest the founders themselves had concluded the opportunity was not worth the dilution.
Entering a market simultaneously with better-funded competitors is a structural problem that product quality cannot solve. Orbiter and Monte Carlo identified the same problem in the same year. Monte Carlo raised $236M; Orbiter raised a YC pre-seed. The data observability market rewarded scale — enterprise integrations, sales capacity, ML infrastructure — and Orbiter's product simplicity was a positioning choice, not a moat. When the market consolidates around capital-intensive players, a three-person pre-seed team needs a structural advantage (a specific vertical, a proprietary data source, a distribution partnership) that Orbiter never publicly articulated or demonstrated.
Positioning a product "for non-engineers" lowers the barrier to adoption but also lowers the barrier to feature absorption. Orbiter's no-code, five-minute onboarding was its clearest differentiator in March 2020. But the same simplicity that made the product easy to adopt made it easy for Amplitude, Looker, and others to replicate as a native feature. Products that live in the gap between incumbent platforms are inherently vulnerable to being absorbed; the durable version of Orbiter's strategy would have required going deeper into a specific workflow or data environment where incumbents were structurally slower to move.
CEO attention is a leading indicator of company trajectory. Mark Wai began co-founding Inkly in January 2021, twelve months before his LinkedIn tenure at Orbiter ended. In a three-person company, the CEO's attention is not a marginal resource — it is the primary resource. The period during which Wai was simultaneously building a new company was also the period during which Orbiter would have needed to raise a Series A, close enterprise customers, or differentiate against Monte Carlo. The absence of any public milestone from this period is consistent with a company that had effectively stopped competing before it formally shut down.
A quiet shutdown with no post-mortem, no acquisition, and no customer communication is itself a data point about the company's final state. Orbiter's YC page lists "Inactive" with no further detail.[18] No blog post, no tweet, no press release. Companies that achieve meaningful traction and then fail typically generate some public record of the failure — a founder essay, an acqui-hire announcement, a pivot explanation. The complete absence of any such record suggests the company wound down before it had built the kind of customer relationships or investor expectations that would have required a formal explanation.