The SaaS that did not care about its customers
How investing in the product turned support cost into margin, and why we now run their technology too.
+200 → <20
tables in the database
10 → 1
support interventions
much+
supplier for product, technology and operations
Context
A mid-sized multinational with a regulatory compliance SaaS for companies operating in several countries. The product worked and had loyal customers. It also had two problems the board saw as one: “IT is slow”.
What it looked like
A usability problem. The platform was so hard to use that the customer support team logged into it on behalf of customers to get things done. Every task that should have been one customer click was a phone call, a ticket and a person.
What it actually was
Two problems feeding each other. The first was product: the platform had been designed around internal operations, not around the customer, so the cost of the bad experience didn't show up in the product but in the support and operations payroll. The second was technology: a small IT team spent all its time keeping alive a monolith that had accumulated patches for years, up to a database of more than two hundred tables nobody dared touch. There was no capacity to improve the product because all the capacity went into holding it up. And the business underneath was paying for it: a low single-digit operating margin in a company that sold software.
What we did
The business first. A short diagnostic exercise that put a simple calculation in front of the owner: every euro invested in reducing technical debt and improving the customer experience came back multiplied in support and operations cost. That conversation is what made everything else possible.
Then the product. A new platform designed around the customer's task: get in, do the thing, leave. Support only steps in for incidents, not for normal use.
Then the technology. Rebuilt on a modern stack (Next.js), with a microservices backend inside a monorepo (Turborepo) and a severe simplification of the data model: from more than two hundred tables to fewer than twenty. That is not a vanity number: it is what lets a small team evolve the product instead of propping it up.
And the operation. We took on not just development but running the platform: separate environments, continuous integration, end-to-end and regression tests. That is the frame that lets you code with AI as the engine without losing quality, because the guarantee comes from the tests, not from the programmer.
Where we are now
The new platform is on the market taking in new customers. Existing customers are migrating in waves, as performance settles and new functionality lands. In two years the company has gone from an almost non-existent operating margin to a double-digit one, and the board has asked us to take responsibility for all of their technology, not just the service platform. The first thing we will do is a plan to reduce systems: when a complete system can be built in days with agents, many corporate tools that used to be essential no longer make sense.
What comes next
The product opens up a new business model: not only the company that needs to solve compliance across several countries, but the professionals who complete the service where the administration is not yet digitised. Expanding into other regulatory areas and other geographies is in next year's plan.
What we took away
The cost of a bad customer experience almost never shows up in the product. It shows up in the support headcount, which is why nobody sees it. And technical debt is not an IT problem: it is a margin problem.
