How we went from 20 listings a day to 250+

The situation

We had a good operation. Stock was coming in. Sales were strong. Marketing was working.

The ceiling was listing.

Every item followed the same process. Photograph it on your phone. Upload to Google Drive. Download on desktop. Log in to eBay. Type out the title. Write the description. Pick a category. Fill in the item specifics. Set the condition. Upload the photos again.

Every item. Every day.

At 20 items a day, it was manageable. Slow, but manageable. As volume grew, it became the thing that capped everything else. More stock arrived than we could list. Items sat in boxes. Revenue that should have been live on eBay wasn't.

We weren't short of inventory. We weren't short of buyers. We were short of listing bandwidth.

What we tried

We looked at what was available. The tools that existed were either built for solo sellers doing low volume, or they were generic AI writing tools that had been pointed at eBay without really understanding how eBay works.

The problems with them were consistent:

The AI would fabricate details it wasn't sure about rather than ask. A listing would come back with a brand confidently stated that wasn't visible in any of the photos. Item specifics would be plausible but wrong. Condition assessments would be based on one image when there were six.

There was no workflow. Generate a listing, fine. But then what? How does it get reviewed? How does it get into a batch? How does scheduling work? How do you track which items are in which boxes?

None of them had been built by people who actually ran a listing floor. You could tell.

Building the first version

Oliver built an internal tool. Not a product. A solution to a specific problem we had in a specific operation.

The first version was simple. Upload photos, get a listing. But even that first version had one principle baked in that everything since has been built around: the AI should ask, not guess.

When it wasn't certain about something, brand, model, measurements, condition detail, it asked. A targeted question. We'd answer it, and the listing came back accurate.

The results were immediate. We went from 20 listings a day to 150+ overnight. Same team. Same hours. The manual process was the bottleneck, not the people.

What happened when we scaled

150 listings a day created problems we hadn't had at 20.

Boxes. At 20 items a day, you know where things are. At 150+, with stock coming in and going out, boxes multiply. A buyer messages about an item and nobody knows which box it's in. Dispatch gets slow. Mistakes happen.

Post days. Without scheduling, all the listings go live whenever someone has time to publish them. That creates chaotic days where the entire team drops what they're doing to push listings live. No rhythm. No predictability.

Quality drift. At higher volume, consistency gets harder. Different team members listing differently. No standard review process. Things going live that shouldn't.

No visibility. Who listed what? How many items did each person process today? Where are we against our weekly target? No data, no management.

Each of these problems became a feature. Not a product roadmap feature. A feature born on the listing floor because we hit the problem and needed to solve it.

Containers and SKU lookup solved the box problem. Scheduling solved post days. The QC workflow solved quality drift. Performance tracking solved visibility.

We moved into a bigger operation and pushed to 250+ listings a day. The problems that appeared at that scale became the next round of features.

What the operation looks like now

Capture. Staff photograph items on their phones or through connected cameras. Photos flow straight into the platform. No Drive folders, no desktop transfers. SKU scanned or entered. Generation guidance added if needed. Box ID assigned.

Generate. The AI agent processes every image, not just the first one. When it's uncertain about a detail, it generates a question. The team answers it. The listing comes back complete: title, description, category, condition, item specifics. Matched to our writing style. Nothing fabricated.

Review. Every listing enters a QC queue. A reviewer approves it, requests specific changes with comments, or rejects it. Nothing touches eBay without sign-off. At 250+ items a day, quality is consistent because the process enforces it.

Publish. Approved listings go into a batch. Scheduling is set, spread across the week, specific times, whatever the operation needs. Listings go live on schedule. No chaotic post days.

Manage. Orders sync and link back to their captures. Buyer messages come in one place. Performance data shows output by person, by day, by week. The operation runs on data, not gut feel.