Catalog Debt: Measuring What Incomplete Data Costs You

Every engineering team knows technical debt: the shortcut you took to ship faster, which now costs more to maintain than it would have cost to do right the first time. Product catalogs carry the same kind of debt, and almost nobody tracks it. A missing attribute, a spec buried in a marketing paragraph, a compatibility relationship nobody bothered to enter — each one is a small loan against future sales. The bill doesn't come due immediately. It comes due the first time an AI agent, an AI search result, or a shopper with a specific question hits that gap and walks away.

What does bad product data cost you?

Bad product data costs you in three distinct places, and most merchants only see one of them. The first is direct: a human shopper can't find the fact they need — dimensions, material, compatibility — and either emails support (cost: time) or buys from a competitor whose listing answered the question (cost: the sale). The second is newer and less visible: an AI agent, whether it's a Brand Agent on your own site or an off-site assistant like ChatGPT or Perplexity, hits the same gap and either declines to answer, hedges with a vague non-answer, or — worse — guesses. A guess that's wrong is a return, a bad review, or a chargeback dispute. The third cost is the quietest: you simply never show up. If your data doesn't state a fact in a structured, extractable way, an AI system can't cite you for it, can't recommend you for it, and can't include you in a comparison where that fact was the deciding factor.

None of these three costs show up on a single line in your P&L. They're distributed across conversion rate, AI-referred traffic, support tickets, and return rate — which is exactly why catalog debt is so easy to carry for years without anyone noticing the balance.

Catalog debt

The accumulated gap between the facts your product data actually contains, in structured form, and the facts a buyer or an AI agent needs to complete a confident purchase decision. Like technical debt, it's invisible until something tries to use the missing piece and fails.

The debt analogy, taken seriously

The tech-debt metaphor isn't just a catchy label — it holds up structurally, and taking it seriously changes how you prioritize the fix.

It has a principal

The principal is the enrichment work itself: the hours it takes to fill in missing attributes, extract prose-trapped specs into fields, and build out compatibility relationships. Like any principal, it's a known, bounded, one-time cost. You can scope it, quote it, and schedule it.

It has an interest rate

The interest is every transaction that touches a gap and fails — the agent that declines to answer, the AI Overview that recommends a competitor because their data was cleaner, the shopper who bounces. Unlike the principal, interest is ongoing and recurring. It's charged every day the gap exists, on every relevant query, whether you're watching or not.

It compounds

This is the part most merchants miss. In 2024, a thin catalog mostly cost you organic search rankings and a slightly worse on-site search experience. By mid-2026, the same thin catalog costs you across on-site AI search, Brand Agents, Copilot Checkout recommendations, and every off-site assistant a shopper might consult before they ever land on your site. You didn't add any new debt — the number of surfaces reading your existing debt just multiplied. That's compounding: the same unfilled field now fails five tests instead of one.

How do I measure catalog data quality?

You don't need a precise industry benchmark to know whether you're carrying too much catalog debt — you need a rough, honest audit of your own catalog against your own shoppers' questions. Here's the field version we run before quoting any enrichment work.

1. 1

   List the questions, not the fields

   For your top two or three categories, write down every question a shopper actually asks — in reviews, support tickets, and pre-sale chat. Don't start from your existing field list; start from real questions. This surfaces gaps your schema doesn't even have a column for.

2. 2

   Score coverage per question

   For each question, check what percentage of SKUs in that category have a structured, populated answer — not "is it mentioned somewhere," but "is it a field." Score it roughly: mostly covered, partially covered, or mostly empty.

3. 3

   Separate 'missing' from 'prose-trapped'

   A fact buried in a description paragraph is a cheaper fix than a fact that was never captured anywhere. Tag each gap by type — it changes the cost of paying it down.

4. 4

   Weight by traffic and revenue

   A gap on your best-selling category is expensive debt. The same gap on a discontinued SKU is nearly free to carry. Weight your findings by revenue, not by SKU count.

5. 5

   Convert coverage into a debt estimate

   You now have a qualitative picture: which categories are debt-heavy, which questions go unanswered most often, and where the revenue is concentrated. That's enough to prioritize — you don't need a fabricated precision score to act on it.

Is catalog enrichment worth the investment?

Enrichment is worth it exactly when the interest you're currently paying — the visible, attributable lost sales from unanswered questions — is larger than the one-time principal cost of fixing the data, sequenced sensibly. That's a genuinely different question from "should I enrich my whole catalog," and it's the one worth answering first.

In practice, the math almost always favors enrichment for hero SKUs and high-traffic categories, and almost never favors a uniform, catalog-wide push on day one. The debt isn't evenly distributed — it's concentrated wherever prose was written fastest and governance was weakest, which is usually your newest and your longest-tail products. Pay down the debt where the interest is highest first: the categories with the most traffic, the most agent-mediated queries, and the thinnest existing data.

The version of this that goes wrong is the one where a merchant enriches everything evenly, in SKU order, with no weighting by revenue or traffic. That's the catalog equivalent of paying down your lowest-interest debt first — technically progress, but not the progress that moves the P&L. Sequence hero SKUs and high-traffic categories first; that's where the interest rate on the debt is highest. Our Catalog Enrichment for AI service and the deeper enrichment playbook both walk through that sequencing in more detail.

Why the debt compounds faster in 2026 than it did before

Shopify's Spring '26 edition shipped Brand Agents and Copilot Checkout on June 17, 2026 — both native Shopify features, currently available on Shopify Plus. Between an on-site Brand Agent, a checkout-embedded assistant, and the growing share of pre-purchase research happening inside ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews before a shopper ever lands on your domain, the number of systems reading your catalog data has multiplied in the space of about a year. None of this required you to add new debt. It just multiplied the number of places your existing debt gets tested — which is precisely why a catalog that felt "good enough" in 2024 can feel visibly broken in mid-2026 without a single field having changed.

This is also why the fix compounds the other direction. Enrichment work done once — a filled material field, a structured compatibility relationship, a dimension pulled out of a paragraph — pays off on-site search, the Brand Agent, and off-site GEO citations simultaneously. You're not enriching for one channel; you're paying down a balance that every channel reads from.

A simple way to talk about your own catalog debt

You don't need to present a fabricated industry statistic to get budget approved for enrichment. You need three numbers you can actually defend: which categories carry the most debt (from your own field audit), roughly how much revenue flows through those categories, and a short list of the specific questions currently going unanswered. That's a more convincing case internally than any borrowed benchmark, because it's your data and your shoppers' actual questions.

• Start with your top 3 categories by revenue. That's where the interest rate on catalog debt is highest.
• Pull real questions from support tickets and reviews, not your existing attribute list — the gaps you don't know about are the expensive ones.
• Tag each gap as missing or prose-trapped. Prose-trapped facts are cheaper principal to pay down; they already exist, they just need extracting.
• Sequence the fix by revenue, not SKU count. See the enrichment playbook for the full prioritization method, and enriching a 10,000-SKU catalog if you're working at that scale.
• Lock the schema after the paydown, or the debt starts accruing again on every new SKU added without governance.

What a well-structured attribute schema does to your debt balance

The single highest-leverage move against catalog debt isn't enriching more SKUs faster — it's designing the attribute schema correctly per category before you fill it in. A good product attribute schema does two things at once: it tells you exactly what "fully paid down" looks like for that category, and it becomes the governance layer that stops new debt from accruing every time a product is added. Without that schema, enrichment is a one-time cleanup that starts decaying the moment it's finished. With it, enrichment is a paydown that sticks.

Think of the schema as refinancing: it doesn't erase the debt you already have, but it locks in a lower ongoing interest rate by making it structurally harder for new gaps to enter the catalog unnoticed.