Build a Supersession Table from Price-List Changes

Most manufacturers do not send a structured supersession feed. They send a new price list. Somewhere in the delta between the January file and the July file, part numbers disappeared and new ones appeared. Those changes are the raw signal. Turning them into a trustworthy supersession table is the work.

Prerequisites

• Two or more successive price lists from the same manufacturer, in a structured format (CSV, Excel, or parsed from PDF)
• A consistent product-group or category field in each list (so disappeared and appeared parts can be paired within the same family)
• Access to the manufacturer’s product documentation or a product specialist who can verify candidate pairs

The workflow

1. Normalize both price lists

   Ensure part numbers are in the same format (uppercase, stripped of leading zeros, hyphens normalized). Align column names so the two files can be compared field by field. Common normalization issues: trailing spaces, inconsistent hyphenation (AB-1001 vs AB1001), mixed case.

2. Diff the two lists

   Produce three sets:

   • Disappeared: part numbers in the old list but not in the new list
   • Appeared: part numbers in the new list but not in the old list
   • Unchanged: part numbers in both (these may still have price or description changes, but are not supersession candidates)
3. Filter noise

   Not every disappeared part is a supersession. Filter out:

   • Parts removed due to a product-line exit (the entire group disappeared, not just one number)
   • Parts removed due to data cleanup (duplicates, test items)
   • Appeared parts that are genuinely new products, not replacements for old ones

   Signals that help: if a part disappeared and nothing new appeared in the same product group, it is more likely a simple discontinuation than a supersession.

4. Pair candidates by product group

   Within each product group, match disappeared parts to appeared parts. The strongest pairing signals:

   • Same or very similar description (after normalization)
   • Matching key specifications (dimensions, ratings, material)
   • Similar price point (a successor is often priced within a narrow band of the predecessor)
   • Part-number pattern similarity (a suffix change like AB-1001 → AB-1001A, or a series shift like AB-1xxx → AB-2xxx)
5. Score and rank candidates

   Assign a confidence score to each candidate pair based on how many signals match. A pair with matching description, specifications and price is high confidence. A pair matched only by product group is low confidence and needs manual review.

6. Verify candidate pairs

   Before loading any pair into the catalog, verify it:

   • Check the manufacturer’s website or documentation for a supersession notice
   • Ask a product specialist: is this the same part under a new number, or a different product?
   • Check the quantity: did the pack size or unit of measure change?
   • Confirm the direction: is the new part a drop-in replacement, or does it have a different specification?
7. Record the supersession

   For each verified pair, create a supersession record with:

   • Old part ID and new part ID
   • Direction (one-way unless explicitly bidirectional)
   • Effective date (the publication date of the new price list, or the manufacturer’s stated date if available)
   • Quantity ratio (1:1 unless pack size changed)
   • Source: “Price-list diff, [manufacturer] [old-list date] → [new-list date]”
   • Status: “Active” for verified pairs, “Pending review” for unverified candidates
8. Check for chain extensions

   Look up each old part ID in the existing supersession table. If it already has an incoming REPLACED_BY edge (it was itself a successor to an even older part), the new pair extends the chain. Verify that the chain resolves correctly to the new terminal part.

What the output looks like

Common pitfalls

• Treating every disappeared part as superseded — some are simply discontinued with no replacement.
• Matching only by part-number pattern — suffix changes can indicate a different product, not a replacement.
• Ignoring quantity-ratio changes — a pack-size change without a ratio creates wrong-quantity orders.
• Loading unverified candidates as active supersessions — low-confidence pairs should enter as “pending review,” not “active.”
• Not checking for chain extensions — a new pair may extend an existing chain, and the full chain must resolve correctly.

Where Claro fits

Claro automates the diff, pairing and scoring steps. When a new price list arrives, Claro compares it against the previous version, identifies candidate pairs using description similarity, specification overlap and part-number pattern analysis, and presents the results as a scored queue for the catalog team to verify. Verified pairs are added to the Product Identity & Evidence Graph as sourced REPLACED_BY edges, and existing chains are extended automatically.

Related

FAQ