Fintech & Real Estate

Gridky.

40,000 properties, 12 tax regimes, banking-grade security. When the machine has to be reliable before it can be smart.

80%

fewer production bugs, approximate estimate

What was at stake.

Gridky manages over 40,000 real estate properties across 12 tax regimes, with banking-grade security requirements. When I arrived, the rush to market had left significant technical debt.

Every morning, I watched hundreds of exceptions pour into Slack. The same fault appeared in several messages; I grouped them by hand. Nobody else read the channel any more. About 10 production bugs a week, stressful twice-weekly releases: to me, we had reached technical bankruptcy.

The product needed to evolve: financial simulator, advanced search engine, OpenAPI standardization. But none of that was possible until the foundations were solid.

The reduction of around 80% is a retrospective estimate. Reported frequencies — about 10 bugs per week, then 1 to 2 per month after six months — are separate approximate figures, not the series used to calculate that percentage. The six-month period includes work after stabilisation; it is not the feature pause, which lasted a month and a half.

What needed to change

About 10 production bugs per week. Slack alerts had become background noise. Bi-weekly deployments were stressful events.

The team was chasing bugs instead of building the product. Tech debt accumulated during the rush to market was blocking all progress.

The choices I had to argue for.

  1. 1

    I personally took over bug tracking and spent 2 hours a day on fixes. If I expected the team to take quality seriously, I had to start there. Separating incidents from feature work was no longer enough: the same underlying causes remained.

  2. 2

    Our CTO secured the CEO’s agreement to a full month without new features, including for the priority client. The Drynuary eventually lasted a month and a half. The whole team worked on stabilisation; every code change needed tests. Developers chose their work according to the severity and frequency of errors.

  3. 3

    Stabilisation let us resume product work: standardising APIs with OpenAPI, the financial simulator and advanced search. I also optimised the infrastructure to cut costs by 15%.

What changed.

  1. Reported frequencies, approximate figures

    BeforeAbout 10 bugs/week

    AfterAbout 1–2/month after 6 months

  2. Deploy frequency

    BeforeRisky bi-weekly deploys

    AfterMultiple reliable releases/day

  3. Test coverage

    Before10%

    After50% after the Drynuary → 85% afterward

  4. Infrastructure costs

    AfterNew-15%

  • x5Deploy frequency

In their words.

Yanis transformed the way we work. In 6 months, we went from chaos to a reliable machine.

MathieuCEO, Gridky

The takeaway.

I still consider that feature freeze the most profitable decision we made that year. We had to accept having nothing new to show while clients waited. Continuing to build on those foundations was already consuming the team’s time in fixes.

Latest articles

Feature freeze: why I still stand by Gridky’s Drynuary

Hundreds of Slack exceptions every morning, a team split in two, a priority client waiting. Then a month without a single new feature, negotiated with the CEO. I still stand by it: in my view, the most profitable decision of the year.

Blog