Supabase sends a lot of emails to customers who are running into trouble. When a product signal fires, such as an account scaling up fast, hitting usage limits, throwing errors or reaching a certain level of spending, the company reaches out. Customers who write back get a call from a salesperson. Those who never reply become something rarely found in a sales department: a control group.
Supabase described this setup on an episode of Latent Space, the AI engineering podcast, published October 28, 2026. Co-founders Paul Copplestone and Ant Wilson spoke with host swyx about how AI agents that write software are changing what a backend has to do. Supabase sells the behind-the-scenes parts of an app, including a database built on Postgres, a widely used open-source database, along with user sign-in and file storage, as its announcement of Supabase for Platforms lists.
Sales built on product signals
Supabase calls itself a product-led company: developers find it, sign up and grow on their own, without talking to a salesperson. In the conversation, the company claimed that 60% of every batch of Y Combinator, the startup accelerator, uses Supabase. Its stated role model is Stripe, the payments company, described as "one of the only developer tools that we've ever seen that has withstood more than a decade of moving our market and still keeping that developer love."
On top of that self-serve growth, Supabase added what it calls product-led sales. Instead of chasing cold leads, it watches for triggers. One is a "logo," a customer whose name the company considers important. "More importantly," the triggers include accounts that are scaling up, hitting limits, having errors or paying a certain amount.
Then, as Supabase put it, "AI came along." The company walked through the method with round numbers offered as an example ("let's say"): a thousand product triggers a week lead to a thousand emails to people having problems, and perhaps 500 reply. The customers who don't respond become the control group, and Supabase measures how much they grow. That figure, the company said, turned out to be 25% month over month.
The customers who do respond go to the sales team with a narrow instruction: get on a call and help. Because Supabase can see each account's traffic patterns, it already knows what the problem is, and AI prepares a script for the call. According to Supabase, accounts that get this help grow 125%.
Salespeople are paid on the difference, the roughly 100-point "incremental uplift" between the two groups, rather than on everything the account spends. Supabase contrasted this with the traditional approach: "get on a call with these people and just sell them shit and then you get the full quota." Its reasoning is that this group of customers already grows about 25% a month without any call, so paying on the whole account would hand salespeople credit for growth that was going to happen anyway.
Winning on day one
The same logic shapes how Supabase approaches large companies. Enterprises usually come to it because they want to launch many databases, Supabase said, and "almost never" to transfer an existing production database. Moving a live database from MySQL to Postgres, or from one scalable database to another, "almost never happens."
That was Supabase's core thesis from the start. The company said every Postgres or database company it had studied that tried to win by migrating customers had failed over time. "You just can't convince people to migrate even if you build a bigger, better, more scalable Postgres." Instead, Supabase aims to land with a customer on day one and keep it to "day 1,000, day 10,000," focusing on greenfield work: new projects with no old system to replace.
Wilson described a change in who brings Supabase into a company. In the past, the company sold to developers, who then championed it internally. Now, he said, there are CPOs saying, "wow, I didn't code for the last 10 years," who are building internal tools themselves with Lovable, an AI tool that builds apps from plain-language descriptions, and Supabase. Those executives have become "champions at the C-level" without Supabase having to build a big sales effort around them, Wilson said.
Why Oracle stays put
The conversation turned to a bigger prize: whether coding agents could finally let companies "rip out Oracle," the long-dominant commercial database. The hope voiced in the discussion was that agents would let businesses migrate off legacy software entirely and "cut over everything."
Supabase cited a statistic it had heard about a year earlier, and which might be out of date now that Oracle is trying to move everything to the cloud: only about 5% of Oracle workloads by revenue ran in the cloud. The company said it would "of course" try to win some of that business, but that it is "incredibly difficult" and the first focus should be landing inside the enterprise.
The discussion then turned to what makes Oracle sticky: code that lives inside the database. Companies write "million line stored procedures," programs stored and run by the database itself, and rewriting those is "probably" a lot easier these days. But, the conversation continued, "most of the problem around migrations is actually not rewriting it."
Moving an analytics warehouse is comparatively easy. The hard case is the OLTP database, short for online transaction processing: the system that records live activity such as orders and payments and usually powers the application that brings in the most revenue. A minute of downtime means lost money, so companies run the old and new systems in parallel. That creates the risk of "split brain," in which some data lands on one side and some on the other, and reconciling the two is "a very hard problem." Many companies end up splitting off workloads instead, starting new ones on the new system and hoping to move the rest later, which, as the discussion noted, usually leaves them with two problems.
Amazon Web Services' guidance on Oracle-to-PostgreSQL migration also treats rewriting database code as only one hurdle. It asks whether the applications around the database can be changed at all, noting that commercial software without PostgreSQL support can rule out a migration.
A fleet view for open-source users
Supabase is both open-source software that anyone can run on their own servers and a hosted cloud platform. The company deliberately puts no tracking in the open-source version, so it does not know how widely that version is used. It learns about big deployments only when customers tell it. Large pharmaceutical companies, for example, came to Supabase saying they were running a hundred instances of the open-source project and asked for a product to manage them.
The answer is Control Tower, which lets an enterprise monitor a fleet of databases. At the time of the conversation it was not on Supabase's website, and its launch timing was still being settled. Supabase described it as a repackaging of the fleet view in Supabase for Platforms, a product launched in December 2025 that lets other companies create and manage Supabase backends for their own users through an API and later hand ownership of those projects over. Supabase said AI companies build on Platforms and that it launches "millions of databases" for them. Enterprises want the same thing internally: to spin up many Supabase instances for their own staff, then see and administer all of them from one place.