Most companies already have a product hierarchy: Category → Sub-Category → Product Type → SKU. But once your SKU count crosses five digits, that structure starts to strain.
- Sub-categories become catch-alls
- Similar products end up in different groups
- Teams interpret the same data differently
And decisions take weeks when they should take hours.
I’ve lived this SKU challenge twice: first as a consultant, trying to make sense of an expanding product range for an FMCG client; then again while managing a scaled eCommerce assortment across multiple channels. This time, with AI, I used a different approach.
I’m not a developer. But AI helped me build a Python script — line by line. No data left my system. No AI hallucinations. Just structured logic and AI guidance, running in my own environment.
This wasn’t just a commercial project. It required cross-functional collaboration — marketing, ops, product and merchandising. Without structure, it gets stuck. With AI, it gets unlocked.
Why bring AI into the process?
Because your current hierarchy — while useful — is often built for internal logic, not reflective of customer behavior, and inconsistent across teams. Here’s the four-step framework I followed, built through a script that AI helped me write, even though I don’t code myself.
Step 1 — Group SKUs based on usage
The script reads SKU titles and descriptions, matches similar words or themes, and assigns them to intuitive clusters like “Sprinkles & Toppings” or “Cake Decorating Tools.” This gave us a more customer-centric lens — without replacing the internal hierarchy.
Step 2 — Flag underperformers
The script tags SKUs as “Keep,” “Review” or “Drop” using sales velocity, revenue contribution and transaction count. The thresholds are adjustable — and it’s all repeatable. Suddenly, SKU-level decision-making became structured, not subjective.
Step 3 — Spot redundancies
The script uses fuzzy-matching logic to find SKUs that are nearly identical in function, just packaged or named differently — confusing customers or cannibalizing each other. We found dozens of clean-up opportunities.
Step 4 — Propose a leaner assortment
The script groups decisions by cluster and shows which categories to expand, which to trim, and where whitespace exists for innovation. We could simulate future assortment shape — and align fast.
Final thought
Back then, SKU rationalization took weeks. Now it takes a few days. I didn’t write a single line of code myself — AI did. It’s not just about cutting SKUs; it’s about building a tighter, cleaner, more customer-ready assortment.
You don’t need to upload data to ChatGPT. You don’t need to be a developer. You just need a clear question — and a bit of AI in your corner. Let’s simplify to scale.