How we turned 2,400 unique physical products into a production-ready online store in eight weeks.
Duration
8 weeks
Year
2024
Services
Products published
2,400+
Build duration
8 weeks
Time per product
< 2 min
Description writing reduction
85%
The challenge
The client ran a specialist physical retail shop with over 2,400 products — every one of them unique. No two items were identical. The products ranged from small decorative objects to large furniture pieces, each with individual dimensions, conditions, provenance notes and pricing. They had tried Shopify twice before but abandoned it both times: manually photographing, editing and writing descriptions for thousands of unique items was an operational problem no template-based approach could solve. They had a shop full of inventory they couldn't sell beyond their local walk-in clientele.
The fundamental constraint was that traditional ecommerce assumes repeatable inventory. Every tool in the ecosystem — from product upload workflows to photo editing software — is designed for products that exist in multiples. For a business where every item sells once and is never restocked, the entire category of standard tooling was wrong. We needed to build something that could move at the speed of a physical shop, not at the speed of a manual data entry team.
The solution
We built a two-part system. The first part was a mobile photography app used by the client's staff in the shop. Staff photographed each product against a portable white card background — front, back, detail shot, and label or maker's mark where present. The app captured these in sequence, guided operators through consistent angles, and bundled each product's images with a minimal data entry form: category, condition, dimensions, and price. No copywriting required at point of capture. A product could be photographed and submitted in under two minutes.
The second part was a processing pipeline that ran on our servers. Incoming image bundles were passed through a background removal and enhancement model that standardised every photograph to a clean white background with consistent lighting, eliminating the inconsistencies of the in-shop environment. A fine-tuned image captioning model generated a structured product description from the visual content — identifying materials, style period, condition details and notable features visible in the images. These descriptions were reviewed and spot-edited by a single team member rather than written from scratch. The processed products were pushed automatically into a Next.js storefront with Stripe integration, live within 15 minutes of photography. The client went from 0 to 2,400 live products in the first six weeks of using the system, and has been adding 30–60 new products per week since.
Key features
A purpose-built iOS web app guided staff through a consistent four-shot photography sequence per product, with built-in lighting check and angle guides. Data entry was reduced to five fields: category, condition (five-point scale), dimensions, maker/origin notes, and price.
Each image batch was processed through a background removal pipeline and a lighting normalisation pass that made variable in-shop conditions invisible in the final images. Output was a clean white-background product image set consistent with premium ecommerce standards.
A vision-language model analysed each image set and generated a structured product description identifying material, period, condition, and key visual features. Human review focused on accuracy and tone rather than writing from scratch — cutting description time by approximately 85%.
Processed products moved from photography to live listing in under 15 minutes via a webhook-driven pipeline. No manual upload steps, no template filling, no copy-paste from spreadsheets.
A bespoke storefront built on Next.js with static generation for maximum page speed, Stripe for payments, and a custom filtering system built around condition, category, price range and dimensions — the actual parameters buyers of unique goods use to search.
Outcome
The client went live with 2,400 products in the first six weeks — more than they had listed across all previous attempts combined. The photography-to-live workflow now runs without any dedicated web or data team involvement: shop staff operate it as a routine part of their working day. Average time-on-site for buyers visiting from organic search is significantly higher than industry benchmarks for similar ecommerce categories, which we attribute to the quality and consistency of the imagery and descriptions. The client has since expanded into selling on their own platform exclusively, removing their dependency on third-party marketplaces and their associated commission structures.
More case studies
A platform that lets anyone create original music, own it outright, and distribute it to streaming services — without giving away rights.
A restaurant replaced its static booking grid with a live table occupancy system driven by camera feeds and predictive ML.
A private members club replaced a PDF waiting list and email newsletter with a full membership platform — events, access tiers, and community in one place.
We build custom when off-the-shelf doesn't fit. Tell us what you need.