Michael Ratnikov
All case studies

YEZAutomotive

Car repair marketplace that stayed fast when suppliers were slow

An online marketplace for car repair. Car owners pick the job, see the price for parts and labour, and book a repair shop. As CTO, I hired the team and built the platform from zero.

10+ external APIs integrated6 web apps on one coreMVP in 6 months, pivot includedInstant quotes from our own price data
YEZ car repair marketplace on laptop and phone
The challenge

Part prices came from outside suppliers. Their systems were slow, limited how often we could call them, and were sometimes down. A customer still had to see a price while they waited, as easily as booking a hotel. And no single source said which part fits which car, so we had to build that data ourselves.

I started with no team and no product.

A product that depends on other companies' systems can pass every demo and still fail with real users. In a Review, I check what happens to your product when those systems are slow or down.

Outside systems we depended on
  • Parts suppliers

    APIs, a CSV price list, an emailed Excel file

    5
  • Payment providers

    Card payments for repair orders

    2
  • Receipt services

    A receipt for every payment, required by law

    2
  • CRM, SMS, maps

    Customer records, text messages, addresses

    3+

10+ external APIs. Any of them could be slow or down.

What kept it working

The same questions I ask about your product

Every outside system failed at some point. Customers did not notice. Here is how, and under each one: what I look for in your product.

When a supplier is slow or down

Supplier prices were copied into our own database in the background. A quote never called a supplier, so a slow or broken supplier could not stop a sale.

In your product I check

What your product does when an API it depends on is slow, down or out of credit.

A quote never waits on a supplier
  • Supplier API
  • Supplier API
  • Supplier API
  • CSV price list
  • Emailed Excel

5 parts suppliers, each with its own format and limits

Background sync

Uses only 70% of each supplier's daily limit. Oldest prices are refreshed first.

Our own price database

PostgreSQL, with a Redis cache in front

Customer sees the price at once

Supplier down? Its prices get older. The quote page still works.

Staying inside API limits

Suppliers allowed 2,000 to 5,000 calls a day. A shared limiter spent only 70% of each daily limit, by design.

In your product I check

Whether one busy day can use up an API limit, or run up your AI bill.

Adding or switching a provider

Each supplier, payment and receipt service sat behind its own small interface. A new supplier meant one new adapter, not a change to the core.

In your product I check

How tied your product is to one vendor, and what it would cost to switch.

Data you can trust

Every price kept its update time. A quote left out prices older than the limit and logged why.

In your product I check

Whether old or wrong data can reach a customer without anyone noticing.

Old prices never reach a customer

Brake pads, supplier B

Every price row keeps its update time

2 days old

Checked before every quote

Older than the limit (7 days by default), out of stock or too slow to deliver? It stays out.

Passes: shown in the quote

Fails: kept out, reason logged

Too oldLow stockSlow delivery

Who can see which orders

One login for 6 apps and 10 roles. Each request was checked for the user's role and for who owns the record.

In your product I check

Whether a user can open someone else's order by changing a number in the address bar.

Errors someone hears about

Errors in the logs went out by email. A receipt that still failed after 24 hours of retries reached staff.

In your product I check

Whether a failed background job is retried, and whether anyone hears about it.

Payments and receipts

Every order followed 7 states, with each change logged. A lock stopped double payments. Pending payments were re-checked with the bank, and receipts were retried.

In your product I check

What happens when a payment callback is missed, or a customer clicks Pay twice.

No payment lost, none taken twice

Every order, 7 states

  1. 1Draft
  2. 2Sent to shop
  3. 3Sent to customer
  4. 4Confirmed
  5. 5Work started
  6. 6Work finished
  7. 7Closed
  8. or Canceled

Every change logged: who and when

A lock stops a double click paying twice

Payment

Bank callback missed? Every pending payment is re-checked with the bank every 30 minutes.

Receipt

Receipt service down? Retried for up to 24 hours, then staff get an email.

Nothing fails without someone hearing about it

Releases and backups

One release pipeline over 4 environments, with database changes as their own step. Daily database backups, and scripts to roll back and restore.

In your product I check

Whether there is a test copy of your product, and how a bad release is undone.

How does your product do on this list?

Book a 30-minute Risk Assessment Call. You get your 3 biggest risks in writing after the call.

How the work ran

From no team to six apps in 1.5 years

6 months

to MVP, a pivot included

4 + 1

engineers I hired, plus a designer

483

database changes on the order database

4

environments on one release pipeline

A team from zero

I started with no team and no product. I hired 2 full-stack, 1 back-end and 1 QA engineer, plus a designer, and stayed hands-on in the code.

Plans from the business

Vision and requirements came from the business team. I turned them into tasks, sprints and a release plan, and talked to the suppliers' technical teams myself.

Standards and QA

I set the coding standards, the release process and a QA step. Six apps for four kinds of users ran on one shared core and one login.

One core, one login, clear limits between the parts. That is the structure I look for in a Review, because it lets a small team keep shipping.

The product

What car owners, repair shops and the team got

For car owners

  • Pick the car, down to the exact model and engine
  • Choose a service and see the price for parts and labour
  • Compare repair shops on a map, with ratings and prices
  • Book a repair shop online, or order by phone
  • Pay by card, with the receipt the law requires

For repair shops

  • Their own app for orders, services and prices
  • A text and an email when a customer books
  • Extra work found during a repair goes to the customer as its own order
  • Every order's history: what changed, who changed it, when

For the YEZ team

  • A CRM for orders, customers and repair shops
  • Spare-parts management: suppliers, prices and replacement parts
  • Data tools: which part and which job fit which car, and the standard hours
  • Car data loaded from 6 reference sources
  • A public site with its own CMS, and promo codes

Technologies

Application

.NET CoreASP.NET CoreAngularTypeScriptEntity Framework Core

Data

PostgreSQLRedisIdentityServer4

Delivery

DockernginxJenkinsS3 backups

Integrations

5 parts suppliersCard paymentsFiscal receiptsExternal CRMSMSGeocodingEmail
Review

What the client said

“Very stress-resistant, always even-tempered, responsive - perfect fit for a rapidly evolving and uncertain startup environment.”

“Michael was a Chief Software Developer at an e-commerce startup YEZ developing a multifaceted platform from scratch. As our product engineering lead he was architecting the entire tech platform staying at the same time 100% hands-on since the team at that time used to be very small. We worked closely on multiple tasks and Michael turned out to be not only a diligent full-stack software engineer with strong background and consistent track record, but a very nice guy to communicate with both during work and off the clock. Very stress-resistant, always even-tempered, responsive - perfect fit for a rapidly evolving and uncertain startup environment. Would highly recommend Michael for every team trying to tackle difficult and challenging task. He is a real pro.”
Roman Lebedev

Roman Lebedev

CFO, YEZ

Is a deadlinecoming?

A round, an audit, a big customer, a vendor contract. Book a 30‑minute Risk Assessment Call. We go through who built your product, what must be ready before that date, and what breaks today.

Free. No preparation. You leave with your 3 biggest risks in writing, and an honest answer on whether a review is worth it now.