SKU Mapping Before Price Benchmarking: A B2B Marketplace Engagement

The Problem: Everyone Wants Phase Two

Price benchmarking is an easy request to make and a hard one to fulfil, because the request contains a hidden assumption: that you already know which products are the same product.

On a B2B wholesale marketplace, you do not. The same item appears as:

• Different product titles, written by different sellers, with different word order
• Different pack configurations — a case of 24, a case of 12, a single unit, a "combo"
• Different unit conventions — per piece, per kilogram, per case, per dozen
• Different category placements on each platform
• Different attribute completeness, with the specification that distinguishes two variants sometimes present and sometimes absent

Compare a price across platforms without resolving this, and you get a benchmark that says you are 30% expensive when you are actually comparing a case of 12 against a case of 24. The number is arithmetically correct and commercially misleading — and because it looks like a real figure, someone will act on it.

The client's phrasing recognised this. The brief specified mapping and standardisation as a distinct phase: identify and align products across platforms and cities by SKU, category and attribute, then benchmark.

Why B2B Is Harder Than Consumer Marketplaces

Three differences make B2B mapping more difficult than the equivalent work on a consumer marketplace.

• No shared identifier. Consumer marketplaces often expose an EAN, UPC or platform-wide product ID that anchors matching. B2B listings frequently carry none — the seller lists a description, a pack size and a price.
• Pack configuration is the product. In consumer retail, pack size is an attribute of a product. In wholesale, the pack configuration is often what is being bought, and the same underlying item at three pack levels is three commercially distinct offers with non-proportional pricing. Normalising to per-unit price is necessary but not sufficient — a per-unit comparison between a single and a case ignores that they are different markets.
• City-level price divergence is normal, not exceptional. B2B pricing reflects local supply, local competition and local logistics. The same seller may price differently by serviceable city. So the mapping has to hold across cities while the benchmark is computed per city.

What Made It Work

• Selling the mapping phase. The commercially hard part of this engagement was not technical. It was agreeing that phase one had value before any price number appeared. Clients want the benchmark; the mapping feels like overhead. It is the entire foundation, and a benchmark built without it will eventually produce a wrong decision that costs more than the mapping did.
• Confidence scores, not binary matches. A three-state model — accepted, review queue, rejected — makes the uncertainty visible and reviewable. Binary matching hides it and pushes bad matches silently into the benchmark.
• Unmatched is a finding, not a failure. Reporting comparison coverage per category is what keeps the benchmark honest.
• Treat the map as an asset. The mapping table is the compounding value in this kind of work. Rebuilt each time, it is a cost. Maintained, it is why month six costs less than month one and covers more.

Where This Applies

Any multi-platform price benchmarking programme needs this phase, whether or not it is named: B2B wholesale and distribution, consumer marketplace benchmarking without shared identifiers, cross-border comparison where catalogues diverge, and private-label benchmarking against branded equivalents.

Related mapping work in our current footprint includes cross-platform product mapping across Meesho and Snapdeal, brand-versus-marketplace comparison programmes, and SKU standardisation across quick-commerce platforms.

Typical first phase: one category, two platforms, one city — mapping only. The deliverable is a match rate and a coverage report. That tells you whether a benchmark on this category is even feasible before you pay for one.

FAQ

Why is SKU mapping needed before price benchmarking?

Because a price comparison is only meaningful between the same product. Without mapping, comparisons run on name similarity and routinely compare different pack configurations, producing precise-looking figures that are commercially wrong.

How are products matched without a shared identifier like EAN or UPC?

By extracting structured attributes from listing text — brand, pack configuration, unit of measure, specifications — normalising them, then scoring candidate matches on attribute agreement. Matches are classified as accepted, needing review, or rejected rather than forced into a binary decision.

What happens to products that exist on only one platform?

They are recorded as unmatched and excluded from benchmarking. On B2B platforms, exclusive listings are common; forcing a match to the nearest similar product manufactures a comparison the market does not support.

Can prices be benchmarked per city?

Yes, and on B2B platforms it is usually necessary. Wholesale pricing reflects local supply and logistics, so a national average obscures the city-level gaps that matter. The product map holds across cities while the benchmark is computed per city.

How is pack-size variation handled?

Through per-unit normalisation, with the original pack configuration retained. Both are needed — per-unit price enables comparison, and pack context prevents comparing a single unit against a case as if they were the same offer.

How long does the mapping phase take?

It depends on category breadth and platform count, and it is scoped separately from benchmarking. A single-category, two-platform pilot produces a match rate and coverage report quickly, which is the input to scoping the full programme.