Most manufacturers today have moved well past the printed spec sheet. Product data lives in spreadsheets, PIMs, websites, and digital catalogs. The problem isn’t that the information is stuck on paper — it’s that it still breaks down on its way to where it needs to go. That same product data is expected to end up on dealer websites, inside search filters, in quoting software, in freight calculators — and it needs to stay consistent across all of them every time anything changes.
Whether a dealer is requesting spreadsheets, hand-entering details, or pulling from a data feed, getting the information from the manufacturer’s system into theirs — accurately and completely — is where the pain lives.
This is the first in a short series walking through the life of a foodservice product — from the manufacturer all the way to the customer hitting reorder. Where does the work pile up, why, and what are we actually doing about it? I’ve been in product data in foodservice since 2015, and this isn’t a sales pitch. It’s me thinking out loud about problems I’ve been close to for a long time, so folks in the industry can tell me where we’ve got it right or where we’re missing something.
Everyone has the “real” data. That’s the problem.
Ask ten people in foodservice where the official record for a product lives, and you’ll get ten honest answers. The manufacturer points to their website. The rep points to whatever the manufacturer sent them. The dealer points to their ERP. The buying group points to their master catalog. The customer points to whatever showed up first on Google. None of them are wrong. They’re just all looking at slightly different versions of the same product, and those small differences add up fast.
Say a manufacturer updates a dimension on a range. That one change now has to reach the rep, the dealer, the buying group, the dealer’s website, the dealer’s quoting tool, and the freight calculator — without anyone re-typing it. That’s a lot of systems that were never really designed to talk to each other.
Three things make this genuinely hard:
The same information gets written a dozen different ways. 36 inches, 36″, and 36in are all the same width, but a website filter looking for an exact match will miss two of them. Multiply that small mismatch across every spec on every product, and the problem compounds quickly.
The format isn’t the real issue. Some manufacturers send PDFs. Some send spreadsheets. A few give direct access through an API. Cleaner format doesn’t mean cleaner data — even pulling directly from a large manufacturer through a structured pipeline, we find model numbers that don’t distinguish variants, accessories tagged as full products, and core attributes like voltage or color missing entirely. What matters more is whether the information is consistent, complete, and lined up with how the rest of the industry describes the same things.
The scale sneaks up on you. Hundreds of thousands of active products across hundreds of manufacturers, with dozens of details on each one. It’s why searching a dealer site for a specific configuration sometimes returns the base model plus five options that aren’t actually available together, or why two products that should show up side by side don’t. Every glitch traces back to a small mismatch upstream.
What we’re actually doing with AI (it’s not just reading PDFs)
A lot of the AI talk in our industry makes it sound like AI means “we read PDFs.” That’s a small slice, and honestly the least interesting one. The more valuable work happens after the data lands. Here’s what we’re building into our Catalog Manager:
- Pulling dimensions into the right fields so length, width, height, and weight aren’t buried in a description — they’re structured values that actually power filtering, freight, and search.
- Matching up categories between a manufacturer’s category tree and a dealer’s or buying group’s, so nobody has to translate one product at a time.
- Filling in missing descriptions grounded in actual specs — not marketing fluff, just clear, useful copy that meets what a dealer website needs.
- Auto-assigning filters so when a new product lands, the right filters on the dealer site light up on their own.
- Spotting configurable products. A long list of similar SKUs is often one product with different options — same base model, but one is 208V and one is 240V, one is red and one is stainless. We identify the shared parent and pull out the options that actually differ, so customers aren’t scrolling through twelve near-identical listings.
- Catching problems as the data arrives — missing fields, wrong freight class, short descriptions — flagged right away instead of three months later.
- Pointing human attention where it matters most, surfacing which products and which missing details are hurting sales right now.
None of this replaces people who know the category. It takes the slow, repetitive part of catalog work off their plates so they can spend their time on the decisions that actually need a human.
Where manufacturers come in
After a decade of this, the honest thing I’ve come to believe is that the biggest wins happen upstream, at the manufacturer. A manufacturer who shares clean, organized product data — in whatever format — saves every dealer, rep, buying group, and platform downstream an enormous amount of repeat work. The manufacturers already doing this are quietly winning more online sales than they realize. The ones who aren’t are leaving their brand in the hands of third-party scrapers and marketplaces that rebuild the data their own way, without the manufacturer in the loop.
The good news: that conversation has genuinely shifted in the last two or three years. More manufacturers are asking us what good data looks like and what their dealers actually need. We’ve started working directly with manufacturers to help clean up and distribute their product data — not just to us, but to the whole channel. Every dealer, rep, and buying group benefits when that happens, regardless of what platforms they’re using.
What we’re building right now
A new product database built for how foodservice actually sells. We’re rebuilding our PIM from the ground up to handle configurable products, parts, and options the way they exist in the real world — not flattened into generic rows that lose the relationships between a base model, its variants, and the accessories that go with them.
Expanding how our dealers sell. Whether a dealer closes business through their website, a full quote, or a phone call with a longtime customer, they should be pulling from the same clean product record. One source of truth, powering all of it.
Hands-on cleanup with manufacturers. A dedicated team is working directly through manufacturer data right now — structuring configurable products correctly, filling missing attributes, normalizing names and specs, and making sure accessories tie to the right parent models. The unglamorous work that makes everything downstream possible.
Closing the loop. Early work on a view that lets manufacturers see how their data is actually rendering on dealer sites and inside quotes — so the feedback between source and shelf gets short enough to matter.
Part two picks up where this leaves off, following the product data into the place it gets used hardest: the dealer’s quote. That’s where margin quietly disappears in a typical foodservice deal, and where we’ve been spending a lot of our time. If any of this sounds familiar — or doesn’t — I’d really like to hear from you. These posts will be more useful if they turn into a conversation.
— Kari
