What makes product data "good"

Good product data isn't about having more of it. It's about having data a channel will actually accept, and that performs once it's live. Those are two different bars, and clearing both is the real goal of all the transforming you do. So what does "good" actually look like? It comes down to a handful of qualities.

Every field the channel requires is present. A missing identifier, price, or image is one of the most common reasons a product gets rejected outright, and a channel won't list what it can't fully describe. Complete doesn't mean every field, just every field that matters for where the product is going.

The values are true. The price in your feed matches the price on your site, the stock count reflects reality, the image is actually the product. Inaccurate data is worse than missing data, because it gets your products shown, then punished: wrong prices trigger channel penalties, and mismatches erode the trust shoppers and platforms place in your listings.

The same fact is expressed the same way across your whole catalog. If color is Slate / Cyan on one product and slate-cyan on the next, automated systems can't group, filter, or trust your data. Consistency is what lets a channel's algorithms make sense of thousands of products at once, which leans directly on the attributes and identifiers underneath.

Values match the channel's expected format, units, and structure. That's currency written the way the channel wants, dates in its format, sizes split or combined the way it expects, titles within its character limits. This is exactly the shape that transformation produces, and what each channel's requirements spell out.

A good catalog has a reliable unique item identifier for every product. It's the thread that ties a product together across your shop, Productsup, and every channel, so updates land on the right item and the same product is never counted twice. When identifiers wobble, everything built on top of them gets shaky.

Passing the minimum gets you listed. Going beyond it gets you found. Detailed titles, full descriptions, proper categories, and high-quality images all feed the search and ranking systems on the other side, so the richer and more descriptive your data, the better your products surface against everyone else's. Good data doesn't just clear the bar, it competes.

Here's the subtle part. Two channels can ask for the same field and still want it filled differently, because they reach different people in different ways. A product description written for Meta, where shoppers scroll a polished feed, won't necessarily land on TikTok, where the tone is faster and more native to short video. The data is "complete and accurate" in both places, but the version that performs is the one written for that channel's audience.

So good data isn't strictly one-size-fits-all. The strongest setups tune the same attribute per channel: a tighter, punchier title here, a keyword-rich one there, an image styled for one platform's look over another's. Productsup is built for exactly this, since each export gets its own stage where you can shape a shared field to fit one channel without touching the rest. It's the same reason channels decide what they want so differently in the first place.

Some of the biggest wins come from changes nothing forces you to make. The classic example is the title on Google Shopping. The spec only asks for a title, so a plain product name passes every check. But Google matches shopper searches against that title, so a title built as brand plus product plus color plus size, like Northpeak Trail Runner GTX Running Shoes, Slate, Size 42, gets matched to far more searches than Trail Runner GTX alone. Same product, much better reach.

These optimizations are easy to miss because they never look like a problem. Your readiness score stays green either way, since nothing is missing or malformed. They're about taking facts you already have, like brand and color sitting in their own attributes, and combining them into the fields that drive discovery. You build these with rule boxes in Dataflow, and getting them right is often what separates a feed that merely works from one that sells.

The hard part isn't building the title, it's knowing what order and which fields work best for your category. A good title for running shoes looks nothing like a good title for laptops. A quick way to get there is to ask an LLM, since these models have effectively read the channels' own guidance and millions of live listings:

The result is a title like Northpeak Trail Runner GTX Running Shoes, Men's, Slate, Size 42 instead of a bare Trail Runner GTX. Run the same exercise per category, since the ideal structure changes from one to the next, and you've got a repeatable recipe for titles that compete.

The catch with these optimizations is that you can't always tell in advance which version wins. Is the keyword-heavy title actually better, or does the shorter one convert more? Rather than guess, you can measure it with Feed Experimentation. It splits a subset of your products into two variant groups, lets you apply different rule logic to each (one title style versus another, say), and tracks how each performs so you can keep the version that genuinely does better. It's the difference between believing a change helped and knowing it did.

Even perfect data goes stale. Prices change, products sell out, new items arrive. Data that was good last week can be wrong today, so keeping it current is part of keeping it good. That's a topic of its own in keeping data up to date.

You don't have to eyeball all this. The platform measures it for you. Analyzer tests and the readiness score check your data against each channel's rules and give you a clear green-or-not answer on completeness and formatting, which is the measurable side of "good" covered in why data has to be transformed. To make sure it stays good run after run, Monitor watches your feeds and flags problems like a sudden drop in items or unmapped values before they reach a channel.