I usually try to blog twice a month, but if you’ve been following me on LinkedIn, you know December has been… intense.
I have been 100% heads-down relaunching my vintage business, VintageReveries. I’ve been busting my ass, quite frankly, and the silence on this blog is simply because I’ve been too busy executing to write about it.
But now that we are halfway through the month (and I have a moment to breathe), I want to share what happens when you apply enterprise-level systems and AI automation to a small business.
The Metrics (Black Friday to Now)
December isn’t even over, and I’ve sold 56 items.
With the exception of two items, every single sale has been a vintage fur coat.
Processing furs is infinitely harder than processing t-shirts. You can’t estimate the weight (a heavy coat can vary wildly from a lighter pelt), and the inspection process for leather health and hidden damage is time-consuming. But, by applying strict processes, I’ve been able to move over 50 of them in just a few weeks.
(Platform Note: Depop is currently crushing it for me, followed by eBay—where my 2006-era account seems to get algorithmic preference. Poshmark’s 20% fee is hurting, and Vinted traffic is nearly non-existent.)
The “Why” Behind the Inventory
There is a personal story behind why I had 50+ fur coats to sell. It involves the winter of 2009, being nearly homeless, and hoarding coats as a form of “armor” against poverty.
I wrote a deeply personal post about the psychology of this inventory release over on the Vintage Reveries blog. If you want the human story behind the business moves, you can read that here: Letting It Go, Listing It Right: Def Jam, Mouse Pee, and the Fur Coat Era I’m Releasing.
The $300 Sale (AI vs. Decision Fatigue)
While the furs were volume sellers, my biggest single win this month was a 1990 WCW Starrcade wrestling crew t-shirt.
It was Friday night. I was exhausted. The shirt had no direct sales comps and looked rough. Normally, I would have tossed it back in the “later” pile. Instead, I leaned on my tech stack.
I used a custom AI assistant (built on Gemini) to identify the provenance, validate a high-ticket pricing strategy despite the lack of comps, and write the SEO copy.
- 8:00 PM: Listed.
- 9:00 PM: Negotiated a counter-offer.
- 9:30 PM: Sold for $300.
The AI didn’t replace my knowledge; it unlocked it when I was too tired to access it myself.
The Tech Deep Dive: Digitizing a 1930s Fashion Dictionary
While I’ve been moving units, I’ve also been coding.
I wanted to process a public domain fashion dictionary I scanned back in 2012. I didn’t want to just OCR the text; I wanted to create a dataset for future computer vision training.
I used Google Gemini to help me write a Google Apps Script that connects to the Gemini API.
- It pulls high-quality JPEGs from my Google Drive.
- It extracts the term and the original definition.
- The Kicker: It uses AI to generate a “Visual Cue”—a detailed description of what the item looks like, which wasn’t in the original text.
For example, for the term “Witch Hat,” the AI extrapolated: “A tall conical hat with a wide brim typically made of felt… The cone shape has a pointed or slightly rounded tip.”
The script processed 5,098 terms in about four hours. Even better? Gemini helped me write the throttling logic so I stayed entirely within the free limits of the API.
I now have a proprietary dataset of 5,000+ fashion terms with visual training data, and I didn’t type a single line of data entry.
Looking Ahead
This month has been about cleaning databases, refreshing themes, fixing checkout bugs, and proving that my processes work.
If you need me, I’ll be over here fulfilling orders and working on my AI listing system (to help and sell to other fashion retailers).
0 Comments