Automating an eBay Business, End to End
I run a resale business that, right now, lives mostly on eBay. The economics are simple: the money is made in sourcing and pricing. Everything between those two (listing, bookkeeping, buyer messages, customer service) is overhead.
So over the past year I've been systematically automating the overhead. Not the judgment, the overhead. A picture of where it landed:
- Photos get shot in a lightbox, run through a pipeline that strips their metadata (your phone tags every photo with GPS coordinates, which means a listing photo taken at home quietly broadcasts where you live), renamed to a consistent scheme, and pushed to cloud storage.
- Listings are drafted with AI reading the actual product photos and barcodes, then created and published in batches through eBay's developer API instead of one painful form at a time.
- The books sync from eBay every night into a small database that tells me what each item really netted after fees, shipping, and the label corrections I didn't even know were bleeding me.
- Buyer messages get forwarded to me on Telegram so I can answer from anywhere.
The headline result: a listing used to take me 20 to 30 minutes. A batch of 10 to 20 now takes about that same 20 to 30 minutes. What was a 3-hour evening is now half an hour, and the listings come out more consistent than anything I did by hand.
What I learned, in three lines
- Automate the pipeline, not the judgment. Software does the photos, the listings, and the books. Sourcing and pricing decisions stay with me; that's where the money is, and where a human still wins.
- Instrument before you optimize. A per-item profit-and-loss view changed my pricing more than any listing tweak ever did. You can't fix what you can't see.
- Privacy is a process, not a setting. Metadata scrubbing is a permanent stage in my pipeline now, not something I remember to do. I keep my data on my own systems and network on purpose.
The full series
That's the summary. If you want the gory details (the 30 failed iterations, the money I leaked, the day I found my home GPS coordinates on the public internet), I broke the whole build into a series. Start wherever the pain sounds familiar:
- The manual grind: what listing looked like before I automated anything
- Photos, and the metadata leak that scared me straight
- Teaching AI to see my inventory
- The CSV wilderness
- The API unlock
- Finding the leak in the books
- Messages and offers
- What's still manual (and what's next)