What Happened When We Audited Real Product Data Live

Most companies feel the cost of bad product data before they can name it. The work shows up as rework, missed launch dates, retailer deductions, returns, and product records that look complete until the moment they have to sell.

It starts with a distinction, not a tool

Our CEO, John Abrams, opened with a story from his time running supply chain at Cardinal Health: 10,000 suppliers, 650,000 active SKUs, and product data moving across the globe. The lesson from that scale was not just cost reduction. It was visibility.

Disconnected systems and inconsistent data quietly cap revenue when no one can see the whole product record from source to channel.

Complete and sellable are not the same thing. The gap between them is where the cost hides.

Putting a number on it

We walked through four places this cost actually lives: team and process, launch and availability, returns and fulfillment, and fees and compliance.

The numbers are sobering on their own. But the live moment that mattered was simpler: we opened the free cost calculator and started plugging in basic audience numbers. Even with only a small team-cost estimate entered, the running total already showed $364,000 a year in manual product-data effort before launch delays, returns, fees, or compliance exposure were added.

Asking AI what to buy

The moment that seemed to land hardest was the AI shopping test. We asked a genuinely specific question: What are the best earbuds for classical music?

A plain Google search returned the familiar sprawl of opinions and products. AI assistants narrowed the field quickly. Gemini returned three specific models. Perplexity narrowed to one. ChatGPT also returned three, but one of its recommendations did not appear in the other tools' answers at all.

That product did not win because it was the most popular. It won because its data gave the AI enough confidence to recommend it for a specific use case.

This is the bigger lesson: AI discovery depends on whether product data is structured, specific, and trustworthy enough to be chosen.

Then we did it live, on real data

To show what this looks like in practice, we ran a live audit on a real sample of product data: a line of protective work gloves exported from an ERP. Twenty SKUs were checked in real time against rules that matter across Amazon, Walmart, GS1 data pool standards, category requirements, and AI-readiness signals.

The result: only 8 of the 20 SKUs were actually ready to ship to Amazon as-is.

When we drilled into one glove record, the reason was specific and fixable. Amazon expects a product title to start with the brand name, and that record did not. Small, precise, easy to correct, and invisible until someone checks.

It is not about replacing what you have

This does not require ripping out your existing systems. Whether you are running ERP and spreadsheets, a PIM you already trust, or something in between, the point is not replacement. It is making the data you already have sellable everywhere it needs to go.

nutrād can serve as your PIM if you do not have one, work alongside the PIM you already invested in, or eventually replace it if that is where your roadmap goes. The starting point is wherever your product data operation is today.

Three ways to go deeper

Watch the recording to see the full live audit, the AI shopping test in detail, and the Q&A that followed, including how fast this can move when the work is approached as product data orchestration rather than another multi-quarter IT project.

You can also run the same cost calculator from the session using your own team size, launch cadence, return rates, and channel complexity. Or you can skip straight to your own data by sending nutrād a 100-SKU sample for a free Product Data Audit.