Your Pricing Engine Is Only as Smart as Your Item Master

Every mid-cap distributor has a pricing-project graveyard. A rebate-optimization rollout from 2021, a CPQ instance nobody trusts, a consultant's segmentation deck still sitting in a shared drive. The trade press keeps diagnosing this as a nerve problem — not enough governance, not enough executive spine to hold the line on discounting. We think that's one layer too shallow. Pricing science runs on product data, and most mid-cap item masters cannot feed it what it needs. Fix that first and the software stops being the hard part.

The habit story doesn't explain the repeat failure

Distribution Strategy Group has argued that what's keeping distributors from modernizing pricing is mostly organizational — no owner, no analytics discipline, reps conditioned to concede. Their companion piece on mid-cap distributors cites Revify research finding that roughly one in ten mid-market companies consistently uses analytics to drive pricing, with 100 to 300 basis points of margin quietly leaking every year. Their habits piece offers the fix: better governance, a named pricing owner, disciplined review cadence.

All of that is true and none of it explains why the same distributors buy a second pricing tool eighteen months after the first one flops. Governance failures are fixable with a memo and a steering committee. If governance were the whole story, the second attempt — with an executive sponsor, a budget line, and institutional memory of the first failure — should work more often than it does. It usually doesn't. Something more structural is eating these projects, and it sits underneath the pricing tool, not inside the sales org.

What a pricing engine actually has to read

Modern price optimization doesn't set one number per SKU. It builds peer groups — SKUs that compete for the same purchase decision — then sets matrix logic across customer segment, order size, and product tier within each group. Zilliant's own description of the method is candid about the inputs: segmentation runs on "customer, product, and order data," arranged in a tree, with statistical smoothing for nodes where transaction volume is thin. That's a reasonable description of how the math works once the tree exists. It quietly assumes the tree can be built — that the underlying product data is clean enough to sort SKUs into the right branches in the first place.

That assumption is where mid-cap distributors break. To build a peer group, the engine needs to know that a 3/8-inch stainless hex bolt from Vendor A and a functionally identical one from Vendor B belong in the same competitive set — which means matching category, material spec, size, and pack UOM across two vendor catalogs with two different naming conventions. To flag a substitute, it needs cross-reference data that usually lives in a rep's head, not a field. To hold matrix logic, it needs every SKU classified into a stable category taxonomy, not a legacy department code a buyer typed in fifteen years ago.

Most mid-cap item masters cannot do this. They accumulated through decades of acquisition, ERP migration, and vendor-fed catalog dumps, and nobody normalized the merge. Descriptions are free text. UOM is inconsistent between branches. The same physical part sits under three SKUs because three legacy systems never got reconciled. Feed that into a pricing engine and it will still produce an output — a recommended price, a discount ceiling, a rebate tier — because the math doesn't know the inputs are garbage. It just looks like garbage to the rep who has priced that SKU by hand for a decade and knows something the model doesn't: the model is comparing it to the wrong peer group.

Why reps override the model, correctly

This is the part the governance framing misses. When DSG's habits piece describes reps who override pricing guidance out of habit, the read is usually cultural resistance to discipline. Sometimes it is. But we'd bet a meaningful share of those overrides are reps correctly distrusting a peer-group assignment built on bad category data — a specialty fastener mis-bucketed with a commodity one, a kitted assembly priced against its loose components, a private-label SKU compared to a national brand it doesn't actually compete with. Override that, enough times, and the sales floor stops trusting the tool entirely. That's not a courage problem. That's an accuracy problem the software inherited from the catalog underneath it.

Gartner's data-quality research puts a number on the general pattern: poor data quality costs the average organization an estimated $12.9 million a year and contributes to roughly 40% of failed business initiatives. Pricing modernization is a business initiative sitting directly on top of the messiest dataset most distributors own. It shouldn't surprise anyone that it fails at above-average rates.

The sequencing case

So here's the contrarian order: don't buy the pricing tool first. Normalize the product master — consistent category taxonomy, populated attributes, resolved cross-references, clean UOM — and then evaluate pricing software. Vendors will tell you their platform handles messy data through statistical workarounds; some of that is true for thin transaction volume, none of it fixes a peer group built on the wrong category. A clean item master doesn't just make the pricing engine work better. It makes the pricing decision auditable, which is what actually earns rep trust — not a mandate to stop overriding, but a segmentation reps can see is right.

This is catalog mechanics, and it's the layer we spend our time on. Our Top Distributors 2026 index and its Digital Readiness Index methodology measure exactly this gap across 200-plus distributors — attribute completeness and catalog normalization, not storefront polish. The distributors who score well there tend to be the ones whose next pricing project has a real shot. Your PIM stores the data; whether that data is normalized enough to segment on is a separate, earlier question, and it's the one worth answering before the next pricing RFP goes out.