What we learned shipping $10M of software in 2025
Six lessons from twelve months, four offices, forty product launches, and one quiet bet that paid off bigger than we expected.
We don't usually write retrospectives. We did this one because some of these took us by surprise, and at least three would have saved us a quarter of pain if a peer studio had written them down a year earlier.
1. The fastest 'no' is a discovery sprint
Six 2-week paid discovery sprints saved us four engagements we would have regretted. The format: $25k flat, two senior engineers + a designer, a single tight problem statement. We finish with a 'go / no-go' recommendation. Clients pay because the deliverable is decision-grade. We benefit because we don't lose 18 months to the wrong fit.
2. AI feature requests double; AI feature shipments don't
We received 3.4× more AI-flavored RFPs in 2025 vs 2024. We shipped only 1.6× more. The delta is mostly the three-filter test from earlier in this blog — most AI feature requests don't survive contact with unit economics.
3. Boring tech is a feature
Postgres, Next.js, Flutter, Sentry. We chose the slightly older option in five different categories this year and shipped two months faster than the brief average. Every 'next-gen' choice has a tax in onboarding, ecosystem, and weird bugs at the edges.
4. The 4-office structure is a moat
We didn't plan it. It started as 'hire who we know, where they are.' Then we noticed clients were closing deals partly because they wanted MENA + EU coverage from one studio. The economics work, the culture holds, and the engineering bar stays uniform because the design system enforces it across all four offices.
5. Defaults are the design
Every Innoveev-built product now ships with: dark mode, RTL support, WCAG-AA contrast, OpenTelemetry traces, error budget dashboards, and a feature-flag system. None of those are extras. The default is the design.
6. The quiet bet
In 2024 we set aside 10% of senior engineering time as a 'house bet' — internal R&D on tooling we thought would matter to clients in 12 months. Most of those bets fizzled. One — a Flutter design-system generator that produces a typed token library from any Figma file — became the most cited reason new clients sign with us. It saves about 6 weeks of project setup. We're keeping the 10% rule.