Case study

Moidex

A B2B SaaS CRM for service networks — car washes and tyre-fitting stations. Late 2022 to early 2026. Founded, designed and led by me (backend, product, architecture), then scaled with a team of six.

Python backendMulti-tenantPaymentsServerlessRoadmap & P&L
The problem

Service-network operators — a chain of car washes, say — ran bookings on paper, spreadsheets and phone calls.

No one offered a CRM that both handled their own operations and plugged them into the marketplaces where demand actually came from. Moidex did both — and that combination is what made it grow.

Four generations, no lost service
How the product evolved
2022 — Telegram order bot

The fastest way to test demand: a bot taking car-wash bookings for a handful of stations. It proved operators would actually use something.

2023 — Flask admin panel

Operators needed a real back office — schedules, price grids, staff roles. Server-rendered because it shipped fastest. First stations moved over by hand.

2024 — React CRM, free

The version that scaled: multi-tenant, self-service onboarding, role-based access. Free, to grow connected stations fast. The entire client base migrated with no downtime.

2025 — Paid tier + mobile

A paid tier for advanced tooling, and a React Native app that kept the product reachable through local internet and Telegram restrictions. Launched December 2025.

Infrastructure
Four stages, forced by load — not fashion
Stage 1
Single VPS

Everything on one box. Enough to find product-market fit.

Stage 2
3 VPS + NLB

Horizontal split behind a network load balancer as traffic grew.

Stage 3
Kubernetes + ALB

Real orchestration, rolling deploys, an application load balancer.

Stage 4
Serverless

10 microservices in one monorepo. ~70% lower infra cost, 99.9% uptime held.

Releases shipped as a 1:9 canary on a pre-stable environment, with the go/no-go call made from live data. CI/CD on GitHub Actions; monitoring on Grafana and Prometheus; on-call was me.

Decisions that mattered
Multi-tenancy

Tenant-isolated data on a document store with a Redis cache layer. Every query scoped by tenant; separate role-based interfaces for owners, managers and floor staff.

Payments

Card acquiring with holds and commission retention, plus recurring subscription billing, across two providers. Idempotent write paths so a retried webhook never double-charges.

Async processing

FIFO and standard queues for job orchestration and partner-event processing. The integrator's order events — most of total volume — flowed through here.

The growth engine
Most order volume came through one integrator partnership

It started with a single marketplace surface and, over two years, expanded to cover several of the partner's consumer apps. Monetised on per-order commission — which gave the business a near-zero customer acquisition cost. I negotiated it and I built the integration.

Beyond the code
Product & business
Roadmap & discovery

Owned the roadmap and the P&L. Ran continuous discovery — interviews with station owners — and turned it into a prioritised backlog.

Pricing & go-to-market

Designed the freemium boundary: partner bookings free, own-order management capped, advanced tooling paid. Priced from competitive analysis. The launch reached ~80% trial-to-paid.

Applied ML

Shipped a demand-forecasting model on weather and calendar features, so stations could plan staffing and supplies ahead of demand.

Support, and saying no

Built support from nothing: personal tickets → duty rotations and answer reviews → a key-account role. One feature I'd built — a simplified floor-staff UI — saw little uptake and was deprioritised rather than defended.

The team

I started the backend and the product myself. As it grew, I hired and led a cross-functional team of six — two backend, two frontend, sales, design — running sprint planning, reviews and retros twice a week. I kept owning the backend the whole time.

2,000+
connected stations
500–600
daily active
~85%
retention
99.9%
uptime
~70%
infra cost cut
~80%
trial-to-paid
10
microservices
4
product generations
What I took from it

I owned every architectural decision and its consequences. I learned where to spend engineering effort and where to ship the boring thing, how to size an integration before it hits the roadmap, and how to kill a feature I'd already built.

Want this kind of ownership on your product?