Profile / Vol. XII / Founders
Satyen Sangani
The economist who turned the dullest problem in enterprise software - a search box for your data - into a $1.7 billion category.
The economist who turned the dullest problem in enterprise software - a search box for your data - into a $1.7 billion category.
Satyen Sangani runs Alation from a glass-walled office in Redwood City, where the company he co-founded in 2012 sells the unglamorous, indispensable thing every large enterprise discovers it needs around its third reorg: a place to find the data, trust the data, and remember who owned the data before the last layoff. He calls it data intelligence. Analysts call it the data catalog. Customers call it whatever it takes to get the budget approved.
The company has roughly 550 employees, around $160 million in revenue, and a valuation north of $1.7 billion as of its last priced round. None of that is the interesting part. The interesting part is how a man trained to write economics papers about development policy ended up building software for chief data officers, and why he keeps insisting, on his own podcast, that the point of all this is to give power away.
His vision for the company comes in seven words: enable a curious and rational world. It is the kind of mission statement that sounds like a Stoic affirmation and reads, on close inspection, like a working theory of how organizations should behave. If you can find the number, you can ask the question. If you can ask the question, you can change the answer. If you can change the answer, you have power. Sangani would like more people to have it.
The arc of his career reads like a man trying on careers the way other people try on shoes, and then walking the same direction in every pair. Morgan Stanley analyst out of Columbia. Associate at Texas Pacific Group. A short stint at a friend's startup during the dot-com bubble - where he discovered, with mild surprise, that he preferred building things to analyzing them. Then Oxford. Then nearly a decade at Oracle, running the financial services data warehousing and performance management business. By 2012 he had watched enough bankers fail to find their own numbers to suspect the problem was structural. He was right.
Approximate, drawn from public reporting. Sangani's preferred unit of measurement is the customer who finally finds the right table on the first try.
Sangani spent close to ten years inside Oracle, helping banks pull intelligence out of warehouses they had already paid millions to build. Most of the work was the same shape. A vice president would ask a question. Three weeks later an analyst would deliver a number. The vice president would ask whether the number was right. The analyst would shrug. Somewhere in the building, on a wiki nobody read, the right answer existed - belonging to a person who had since changed teams.
He left, eventually, with two convictions. The first: the problem was not a missing dashboard. The problem was missing context. The second: machine learning could read the metadata - the log of who queried what, when, and why - and surface that context automatically. The catalog would not replace the analyst. It would tell the analyst who to ask.
He brought the conviction to a first meeting with Aaron Kalb, then fresh off building parts of Siri at Apple. Kalb's instinct was crowdsourcing: humans annotating data the way humans annotate everything else they care about. Sangani's instinct was machine learning. They agreed to do both. Feng Niu and Venky Ganti completed the founding team. The category they invented did not yet have a name.
The pitch in the early days was patient and slightly embarrassing. Nobody in 2013 woke up wanting to buy a data catalog. They woke up wanting to fire the consulting firm that was supposed to have built one. Alation sold to that frustration. By 2019 it had a hundred million dollars in ARR. By 2021 it was a unicorn. By 2022 a centaur. By the time generative AI arrived in earnest, the catalog had stopped being a nice-to-have. Models needed grounding. Grounding needed metadata. Metadata needed a system of record. Alation had spent a decade quietly building one.
Patience compounds. So does metadata.
Grew up in Orinda, California. One of three or four South Asian students in his Miramonte High class. Self-describes, with no apparent regret, as an introvert and a nerd.
His father immigrated from India in 1958, started electroplating factories, then bought hotels and supermarkets. The arc rhymes with his son's: technical training, operator's instinct, long time horizon.
Met his now-wife at two separate weddings where they happened to sit next to each other. She said yes only after multiple requests. The pre-Oxford phone call lasted 18 hours and rerouted the next two decades.
Two sons, Karan and Rohan. Sangani calls the household one of "stubborn, willful, very argumentative people." Not peace, exactly. Productive friction.
Yoga and gym for the head. Skis with the younger son for the lungs. Aspires to get back to rock climbing, the hobby that keeps not winning his calendar.
Hosts the Data Radicals podcast - long conversations with the chief data officers and academics he finds most useful. The show is, in part, his job. It is also, transparently, his therapy.
The pitch for a data catalog used to be: you have many tables, you want to find them. The pitch now is harder to summarize and more important. A model that hallucinates is a model without grounding. Grounding is metadata. Metadata is provenance, lineage, definitions, owners, freshness, sensitivity. It is the work of a hundred half-finished wiki pages, written by a thousand people who have since left. Alation's bet, made a decade ago and quietly compounding, is that the company that becomes the system of record for that work also becomes the substrate every other enterprise AI tool sits on top of.
Sangani argues the bet in a particular way. He is not interested, in interviews, in the language of disruption. He talks about curiosity and rationality the way other founders talk about TAM. He uses the word "distribute" more than the word "scale." He notes, often, that giving someone access to a number is the same as giving them a vote. He is, by temperament and training, an economist who happens to ship software - which makes him slightly unusual in a category increasingly populated by ex-database engineers and ex-McKinsey consultants in roughly equal measure.
It also makes him patient. Alation took years to look obvious. The decade between founding and unicorn status was not a victory lap. It was a slow argument with chief information officers about whether the catalog was infrastructure or a feature. Sangani won the argument by outlasting it. The same patience now shapes his commentary on AI - less breathless, more skeptical, more interested in what gets written into the catalog than what gets generated by the model. The catalog, in his telling, matters more in an LLM-saturated world, not less. The model needs to know what it is reading. The catalog tells it.