How it started
It started with a simple question I could not answer: when was the last oil change? I knew it had been done. The receipt was somewhere, the mileage was in a note on my phone, and the insurance date was in an email I would have to search for. None of it was hard information. It was just in five places.
Around the same time I was spending my evenings comparing motorcycles, the way riders do even when they are not buying anything. That meant a row of browser tabs, each with a spec sheet in a different format, some in horsepower and some in kilowatts, half of them missing the one number I wanted.
At some point I noticed these were the same problem. The bike you own and the bike you want are both a motorcycle with data attached. One needs a place to keep it, the other needs a fair way to compare it. I build software for a living, so I decided to build that place.
What I built
RPMVault has two sides that share one catalog:
- A garage. Each motorcycle gets its own record: purchase, insurance and kasko dates, plate and VIN, mileage, and quick actions to log fuel or a service. Monthly analytics split spending into maintenance, fuel and other costs, and show when the next service is due.
- A catalog. More than 18,000 models, each with its own page. Pick any two and compare engine, power and performance side by side, with the better value highlighted.
The two sides feed each other. Research a bike, buy it, and it moves into your garage with its specs already filled in. Over time the service log turns into an ownership timeline you can share, which is handy when it is your turn to sell.
The decisions that shaped it
Web first, built for search
Most people start choosing a motorcycle in a search engine, not an app store. So the web platform came first, in Next.js, with an architecture that gives every model a fast page of its own. My day job is front-end performance on a large e-commerce site, and the same habits apply here: light pages, structured data and no waiting.
A short list of tools
The phone apps are React Native, for iOS and Android, and everything talks to one Node.js API. Working alone, every extra technology is one more thing I have to keep running, so I kept the list short on purpose.
AI only where it answers a real question
The assistant answers questions about a specific bike, like whether it suits a new rider or what it costs to run. It is a helper for research, not a mechanic: anything to do with safety or a repair still belongs with a qualified workshop.
Research that leads somewhere
Offer requests let someone who has finished comparing ask for a price. That closes the gap between reading about a bike and riding one, and gives the project a way to pay for itself without filling the pages with ads.
What I learned
Data is the real product. Getting thousands of models into one consistent shape, with the same units and no duplicates, took longer than any screen. A comparison is only as good as the numbers behind it, so most of the quiet work went there.
The store is part of the product. Screenshots, privacy answers, account deletion and review notes are not paperwork you do at the end. They decide whether people ever see the app, so I learned to plan them from the start.
Launch is the beginning. The first support emails taught me more than my own testing did. Small fixes in the weeks after release are where a product becomes something people keep using.
RPMVault also showed me what I wanted to build next: the same idea, with everything I had learned, for cars. That became RPMBase.