How to Improve the Reliability of Product Matching in its own catalog?
Product matching is not just a matter of competitive intelligence: internally, it prevents the need to manage prices, inventory, and product assortments based on a catalog riddled with duplicates across brands, channels, or systems.
A reliable product master data set (golden record) is a prerequisite for any serious data analysis: without it, price elasticity, cross-selling, and inventory tracking are calculated based on inaccurate data.
For additional information on price matching against competitors (comparing your prices to the right products offered by competitors), see our article on competitive monitoring.
Does the same item appear under three different SKUs in your ERP, your e-commerce site, and your marketplace? This isn’t just a technical detail—it’s a catalog that’s misleading your management tools.
Unlike competitive product matching, which compares your products to those of a third party, internal product matching involves ensuring the accuracy of your own product database: identifying and merging duplicate entries that accumulate across systems, retail chains, or as a result of a catalog merger.
This article explains why these duplicates occur, what they actually cost, and how to build a clean product database—the essential foundation for any analysis of price, margin, or product assortment.
Internal product matching: A distinct challenge from competitive intelligence
When we talk about “product matching” in retail, we often think of comparing products with competitors’ SKUs to compare prices. This is a real issue, which we cover in our article on competitive monitoring.
But before you even compare yourself to external sources, your own catalog must be consistent. That’s the role of internal product matching: to identify and merge records that actually refer to the same item within your own information system.
Why Catalogs Are Becoming Fragmented
Most retailers aren't starting from scratch: their catalog has been built up over time through the accumulation of systems and processes.
Multiple sources (ERP, e-commerce, marketplaces)
The same product is often entered separately in the ERP system for inventory management, in the e-commerce CMS for the product listing, and in the marketplace feed for external distribution. Each system has its own naming conventions and required fields, and no one verifies that they are indeed the same item.
Mergers, Acquisitions, and Multi-Brand Operations
When a company acquires or merges with another, it brings together two catalogs that were developed independently, each with its own naming conventions. Without a dedicated merger project, duplicates quietly accumulate over the years.
Decentralized Manual Data Entry
When multiple teams or stores can create a new product listing without checking to see if one already exists, there’s a strong temptation to recreate it rather than search for it, especially when under time pressure. The catalog grows faster than its quality improves.
The Cost of a Poorly Deduplicated Catalog
A duplicate reference is not just an extra row in a spreadsheet: it skews every decision based on the data it produces.
Biased price and margin analyses
If sales of the same item are spread across two or three separate SKUs, your calculationof price elasticity or contribution margin will automatically underestimate the product's actual importance. The pricing decisions that result from this are based on flawed data.
Fragmented inventory management
Duplicate entries often mean that inventory levels for the same item are tracked separately, creating a risk that one SKU could run out of stock while another remains overstocked without anyone noticing, due to the lack of a consolidated view.
A distorted product mix and cross-selling
Cross-selling recommendations and shopping cart analyses are based on purchase history by SKU: duplicates weaken this signal and make the recommendations less relevant, just as poor competitive matching skews a price comparison.
According to Gartner (2020), this is the average annual cost incurred by an organization due to poor-quality data. A product repository riddled with duplicates is a key contributor to this.
Methodology for Improving the Reliability of Your Product Repository
The logic is similar to that of competitive matching, but this time applied to your own data.
1. Define the scope
Don't process the entire catalog all at once: prioritize the best-selling categories or those most at risk (post-merger, multi-brand).
2. Building the Golden Record
Define the key attributes for each product: EAN, brand, model, and dimensions. This is the single reference record to which everything else must be linked.
3. Dedup across multiple signals
As with external matching, don't rely solely on the EAN code, which is often missing or entered incorrectly: cross-reference the title, brand, and technical attributes to identify duplicates with a confidence score.
4. Merge and trace
Merge duplicate records into the golden record while maintaining a history—specifying which records were merged with which others and why—so you can revert the changes if an error occurs.
5. Lock in future creation
The real challenge is to prevent new duplicates from being created: mandatory searches before creating a record, centralized validation, and naming conventions that are consistent across all systems.
Who should lead this project?
A clean product repository isn't just an IT issue—it's a matter of shared governance.
The category manager knows the products and can resolve ambiguous cases. The data/IT team ensures the technical reliability of the deduplication process. The product or pricing department, for its part, stands to gain the most from a clean catalog: it is this department that sets prices and margins based on this data.
Competitive matching compares your products to those of a third party to set your price in line with the market. Internal product matching, on the other hand, matches existing product records in your own systems (ERP, e-commerce, marketplace) to eliminate duplicates of the same item. Both rely on similar techniques (multi-signal confidence scoring) but serve different purposes.
The "golden record" is the product record considered to be the single, reliable reference for a given item, toward which all duplicates identified across your various systems are consolidated. It serves as the foundation upon which pricing, inventory management, and product assortment analysis are subsequently based.
One-time deduplication is not enough if nothing prevents new duplicates from being created. The creation of duplicates must be prevented at the source: a mandatory search of the repository before any new entry is created, centralized validation, and naming conventions common to all systems that feed into the catalog.
For the external aspect of this topic—comparing your prices to those of quality products offered by competitors—check out our article on competitive monitoring.
A PIM can store, enrich, and distribute a price. It generally cannot determine whether that price is the right one: price elasticity, competition, cannibalization, and impact simulation fall outside its scope.
The two tools complement each other rather than replace one another: PIM ensures the reliability of product data, while a dedicated pricing solution transforms that data into simulated and governed pricing decisions.
Retail promotion management must rely on rigorous data analysis to ensure profitability. By mastering uplift and cannibalization, retailers can transform a high-risk lever into a tool for healthy growth. Precise monitoring is vital, as six out of ten promotions today prove to be unprofitable.
The success of a retail pricing strategy relies on moving away from outdated spreadsheets in favor of (semi-)automated execution driven by AI. This technological pivot allows retailers to delicately balance profitability with commercial attractiveness.
This is essential for building customer loyalty, given that 62% of shoppers are willing to switch retailers for a better price.