Merge Three Catalogs Without Deleting the One Part Someone's Safety Depends On

📥 Download the practice file: case-file.txt — three distributor catalogs, exactly as they'd land on your desk, ready to paste into any AI tool (Copilot, ChatGPT, or Claude). You need this file to run the exercises below.

---

Atlas calls it a "Hex Bolt." Ironclad calls it a "Hex Cap Screw." Summit doesn't list the grade. Three catalogs, three naming conventions for the same half-inch of zinc-plated steel — and your VP of Operations wants them merged into one comparison table so he can cut two of these three distributors and stop bleeding tail spend.

So you have three price files open side by side. Different column layouts. Different SKU systems. And — the part that will bite you — different units. One quotes per each. One quotes per box of a hundred. One quotes per box of fifty, with the "fifty" hiding in a footnote. Line them up by the number in the price column and you will rank them exactly backwards.

That's the arithmetic risk. Here's the one that keeps category managers up at night. Two catalog lines can read "nitrile gloves, blue, large" and mean two different products — one rated for a chemical-handling station, one not. Merge them as a "duplicate," phase out the distributor that carried the safety-rated one, and the consolidation quietly deletes a part that a solvent crew's hands depend on. Nobody notices until the wrong gloves show up in the supply closet and EHS starts asking questions.

The recommendation this case produces: consolidate to Summit (cheapest on 8 of 12 comparable lines after per-piece normalization), retain Ironclad for two safety-critical items only it carries, and phase out Atlas — the distributor whose printed prices look lowest but is cheapest on zero lines once the units are converted.

You've done this merge by hand before: four hours of INDEX/MATCH, calling a rep to ask what "BX" means in their catalog, re-reading footnotes for pack quantities. An AI can do it in minutes. The question this lesson answers is not "can AI merge three catalogs" — it can, faster than you'd expect. The question is: which parts of the merge do you still have to check yourself, and how do you brief the AI so those parts are checkable in ten seconds instead of re-derivable from scratch?

We'll do it in three moves: feed the catalogs so units come out comparable, cross-check the matches so a spec difference doesn't get merged away, and compress the result into the one page your VP will actually sign.

---