I spent a year on a job aggregator. Here are 5 mistakes I made
About a year ago, I started building a job aggregator.
The idea came from a personal pain point: freelance -> full-time -> freelance again. Each time, job search meant dozens of tabs — DOU, Djinni, LinkedIn, Work.ua, and others.
I decided to collect everything in one place.
Early Results
At the beginning, I:
- connected several job sources;
- implemented basic data normalization;
- set up analytics (Amplitude, Google Search Console);
- optimized pages for SEO.
Without ads, the site was getting 5–6 new users per day. Around 500 visitors and 2,000+ page views in the first month.
Reasonable for a side project.
Tech Stack
- Frontend — Next.js.
- Database — Supabase.
- Hosting — Vercel.
- Parsing — n8n automations + custom scripts.
The architecture worked but wasn’t stable for growth.
Domain Issue and SEO Drop
At some point, a domain verification issue led to temporary suspension.
Consequences:
- pages dropped from index;
- organic traffic declined;
- growth stopped.
If your growth depends entirely on SEO, it’s fragile.
Infrastructure Limits
Later I faced:
- Vercel limits;
- bot traffic consuming resources;
- unstable parsing from some sources.
I was optimizing code instead of validating demand.
Architecture Rewrite
After a pause, I rebuilt the system:
- separated backend into a standalone API;
- cron-based ingestion modules;
- Redis caching;
- removed unnecessary admin tools;
- kept 4 stable sources.
Technically improved. User growth didn’t change.
The Core Mistake
I tried to build a better version of DOU and Djinni.
Better filters. Cleaner structure.
But users don’t switch platforms because of cleaner UI. They stay because of trust and network effects.
What I Learned
- Technology is not the main barrier anymore.
- Rewriting code doesn’t fix lack of demand.
- Single-channel traffic is risky.
- Marketing can’t be postponed.
- Without distribution, it’s just a side project.
What’s Next
Now the focus is not new features, but validating demand and building audience.
Building is easier than ever. Being needed is not.