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.

Other posts

© 2026 Vadym Mirvald
GMT+1 : --:--