We Stopped Renting Our Product Data: Why ADH Built Its Own PIM Instead of Paying by the SKU

Somebody has been charging us, by the SKU, to describe products we already sell.

In May we wrote about loading thousands of new SKUs in an afternoon instead of two weeks. Several distributors wrote back with the same question: fine, but where does that data live afterward, and who owns it?

That's the right question, and it's a bigger one than it looks.

The meter that never stops

The standard answer in distribution is to rent your catalog. A product information vendor charges a fee per SKU, maintains the descriptions and attributes and images, and you pull from them. It works, and there are decent vendors doing it.

But look at what's actually being bought. The meter runs forever, and it runs whether or not the catalog got better this year. It prices you out of your own long tail, so slow-moving items — frequently the best margin in the building — stay half-described. And the data isn't yours, which becomes clear the day somebody wants to do something the vendor didn't anticipate.

The deeper cost is that rented product data can't be used as infrastructure. Search results, category navigation, spec filters, cross-sell, barcode labels, a website's ability to answer a contractor's question at 9pm — all of it sits downstream of clean product data. When the catalog is a subscription you query, everything downstream is capped by what the subscription happens to include.

So we built our own, and we own it outright.

What that looks like in practice

AdvantageOS now has a real product information system at its center: one master item record, a shared taxonomy, structured attributes, and links out to each company's ERP. It's the spine. The storefront reads from it, search reads from it, the filters read from it, the labels read from it.

cfm's catalog was the proving run — roughly three thousand live products classified, attributed, imaged and linked to their NetSuite records. The enrichment pipeline we described in May did the first pass. Everything after it is the part worth telling:

• We host the images ourselves. Pointing at a vendor's image server means product pages break the day that vendor reorganizes their site, which they will, on a Friday.
• Attributes are standardized, not free text. "3 ton," "36,000 BTU" and "36 MBH" are one fact. As three different strings, they make the filters quietly lie to customers.
• The people who know the catalog fix it themselves. cfm's product team went through the category tree branch by branch and told us where we had it wrong — residential and commercial condensers mixed together, coils that needed their own tab, refrigeration in the wrong place. Their corrections went in as data, and the site navigation rebuilt itself. No release, no vendor request, no ticket queue.

That last one is the one we'd underline. The people who know a catalog best already work at the company selling it. Software's job is to let them act on that without asking permission from a company in another state.

Rent versus ownership

Because the system isn't specific to one company, the work cfm's team did improves what CTC uses, and the reverse. Two distributors, one spine, every correction shared.

Three years of per-SKU fees buys three years of per-SKU fees. Stop paying and nothing remains — not the descriptions, not the attributes, not the leverage.

We understand perfectly well why the industry rents. Building the alternative is slow and expensive, and against one distributor's volume the math genuinely doesn't close, which is why almost nobody has done it. Across a portfolio it closes comfortably, and then it keeps closing, because the second catalog costs a fraction of the first.

One of those is a cost that grows with the catalog. The other is an asset that grows with it.