Part Supersession Chain Intelligence
Recovering $1.1M in Dead Inventory and Eliminating 4,200 Annual Lost Sales from Broken Part Number Chains
Industry
Automotive & Heavy Equipment
Scale
$285M Revenue
Duration
20 Weeks
Location
Indianapolis, Indiana
Engagement
AI Consulting
Executive Summary
The Director of Inventory at a 16-branch automotive and heavy equipment parts distributor in Indianapolis knew the catalog was rotting from the inside. Manufacturers superseded part numbers constantly — discontinuing old numbers, issuing replacements — and the ERP couldn’t keep up. A Cummins injector number from 2012 had been superseded 3 times since then, but when a fleet shop ordered the 2012 number, Epicor Vision showed “no stock” — even though the current replacement was sitting on the shelf 30 feet away under a different number. Meanwhile, $1.1M in physical inventory sat under superseded numbers the ERP still considered active, automatically reordering parts that would never sell under those numbers again. We embedded ML-driven supersession intelligence into the distributor’s catalog and purchasing workflow on Epicor Vision.
Business Impact
$1.1M
Dead inventory identified and dispositioned under superseded part numbers
4,200
Annual lost sales recovered from broken supersession chains
93%
Supersession chain resolution rate at point of order, up from 41%
$740K
Working capital released from inventory correction
The Situation
The distributor served fleet operators, independent repair shops, dealership service departments, and construction equipment owners across Indiana, Ohio, Kentucky, and Southern Illinois. The catalog spanned 95,000 active SKUs from 200+ manufacturer lines — Cummins, Paccar, Caterpillar, John Deere, Eaton, Meritor, Dorman, Denso, Gates, and dozens of specialty suppliers — across OE, aftermarket, and remanufactured channels.
The distributor processed an estimated 8,000-12,000 part number supersessions per year across its manufacturer base. The 2-person product data team managed these through manufacturer bulletins, emails, and manual ERP updates — running 4-8 weeks behind on average. During that lag, every customer order referencing a superseded number was either a lost sale or a manual save by someone who knew the chain.
- Supersession chains went 3-5 levels deep on long-lifecycle components. A Cummins turbocharger part number from 2010 might have been superseded in 2013, again in 2016, and again in 2021. The fleet shop’s maintenance system still referenced the 2010 number. The distributor’s ERP only knew about the 2016 and 2021 links. The 2010-to-2013 connection had been lost in a system migration 6 years ago — meaning any order placed under the original number returned “not found” instead of resolving to the current part
- Multiple manufacturers superseded independently on different timelines. The OE number might supersede, but the aftermarket equivalent of the old OE number was still in production under its original aftermarket part number. A Dorman replacement for a discontinued Cummins water pump was still perfectly valid and in stock — but the ERP’s cross-reference pointed to the old Cummins number, which pointed nowhere
- The inventory consequence compounded monthly. Physical stock sat under superseded numbers that the ERP still flagged as “active” because nobody had updated the status. The automated replenishment system continued ordering these parts from manufacturers who filled the order under the new number — creating duplicate inventory positions where the old-number stock gathered dust and the new-number stock was being purchased alongside it
- The 2-person product data team received supersession notifications through 40+ different channels — manufacturer portals, email bulletins, PDF product change notices, rep phone calls, and quarterly catalog updates. Each notification had to be manually verified, the chain traced, the ERP updated, and the inventory cross-referenced. At 8,000-12,000 supersessions per year, the team was structurally underwater
- Off-highway equipment manufacturers — Caterpillar, John Deere, Komatsu — maintained proprietary numbering systems with no standardized cross-reference to each other or to the aftermarket. When CAT superseded a hydraulic filter number, there was no automated path to update the aftermarket equivalents from Donaldson or Fleetguard — the product data team had to trace each connection manually
Every month the supersession backlog grew, the catalog drifted further from reality — inventory accumulated under numbers that would never sell again, customers were told “out of stock” on parts physically sitting in the warehouse, and the 2-person data team fell further behind a problem that was mathematically impossible to solve manually at this volume.
The Challenge
The Director of Inventory pulled a report that quantified what he’d suspected for 2 years. He isolated every SKU in the catalog that had received zero demand in the trailing 12 months but still carried physical inventory. The total was $2.4M across all 16 branches. When the product data team investigated a sample of 200 of those SKUs, they found that 46% — nearly half — were parts sitting under superseded numbers where the current replacement was also in stock under its new number. The distributor was carrying both the old and new number simultaneously, with demand flowing to the new number and the old number’s inventory aging silently.
The lost-sales side was equally painful. The Director worked with the counter operations team to pull 90 days of “not found” and “no stock” lookups from Vision. Of the 4,800 lookups that returned no result, an estimated 35% were customers ordering under superseded numbers that the system couldn’t resolve — parts the distributor actually had, under a different number, but couldn’t connect to the customer’s request.
- The ERP’s native supersession field was a single-level pointer — Part A superseded by Part B. It couldn’t store or traverse multi-level chains (A→B→C→D). When a customer ordered Part A and the current part was Part D, the system could only resolve one hop. If A→B was entered but B→C and C→D hadn’t been updated yet, the chain was broken at B
- Manufacturer bulletins arrived in inconsistent formats — some as structured data files, some as PDF tables, some as plain-text emails. The product data team spent approximately 30% of their time simply extracting supersession data from these communications before they could begin updating the ERP
- The aftermarket cross-reference problem multiplied the complexity. A single OE supersession could affect 3-5 aftermarket equivalents from different manufacturers, each requiring its own cross-reference update. A Cummins injector supersession touched Dorman, Alliant Power, and 2 remanufactured sources — none of which superseded their own numbers in sync with Cummins
- Return-to-manufacturer windows on superseded inventory were typically 6-12 months from the supersession date. The product data team’s 4-8 week processing lag meant that by the time they identified inventory sitting under old numbers, weeks or months of the return window had already elapsed — and on some items, the window had closed entirely, converting returnable inventory into permanent dead stock
The Solution
We spent 4 weeks in discovery analyzing the distributor’s catalog, supersession history, and the product data team’s workflow — including a forensic reconstruction of the supersession chains on 500 high-value SKUs to measure the actual depth and completeness of the ERP’s current chain data.
The findings were specific: 23% of the catalog’s supersession links were incomplete — missing at least one hop in the chain. 11% pointed to numbers that had themselves been superseded, creating dead-end references. And 8% had no supersession data at all despite the manufacturer having issued a replacement, because the bulletin had arrived during a period when the product data team was processing a backlog from another manufacturer.
The system analyzed signals including:
- Continuous ingestion of manufacturer supersession data from every available source — structured data feeds from manufacturers who provided them, NLP parsing of PDF bulletins and email product change notices from manufacturers who didn’t, and weekly automated comparison against manufacturer online catalogs to detect supersessions that were never formally communicated
- Full-depth chain reconstruction across every part number in the catalog — tracing A→B→C→D→current for every SKU, including the links that had been lost in the 6-year-old system migration, and validating each link against manufacturer confirmation data
- Cross-reference cascade intelligence that automatically traced the aftermarket impact of every OE supersession — when Cummins superseded an injector, the system identified every Dorman, Alliant Power, and remanufactured equivalent that needed its cross-reference updated, and whether those aftermarket parts were still valid for the new OE application
- Inventory obsolescence scoring that continuously evaluated every SKU for supersession risk — combining chain status, manufacturer lifecycle signals, demand velocity trends, and the remaining return-to-manufacturer window to flag at-risk inventory while disposition options still existed
- Historical demand pattern analysis on superseded numbers — identifying which old numbers still received regular customer lookups (indicating the number was embedded in fleet maintenance systems and needed permanent chain resolution) versus which received no lookups (indicating the old number could be safely deactivated)
23% of supersession links in the catalog were incomplete. 11% pointed to dead-end numbers. Every broken link was either a lost sale or an invisible inventory write-down waiting to happen.
The Challenge
The Director of Inventory pulled a report that quantified what he’d suspected for 2 years. He isolated every SKU in the catalog that had received zero demand in the trailing 12 months but still carried physical inventory. The total was $2.4M across all 16 branches. When the product data team investigated a sample of 200 of those SKUs, they found that 46% — nearly half — were parts sitting under superseded numbers where the current replacement was also in stock under its new number. The distributor was carrying both the old and new number simultaneously, with demand flowing to the new number and the old number’s inventory aging silently.
The lost-sales side was equally painful. The Director worked with the counter operations team to pull 90 days of “not found” and “no stock” lookups from Vision. Of the 4,800 lookups that returned no result, an estimated 35% were customers ordering under superseded numbers that the system couldn’t resolve — parts the distributor actually had, under a different number, but couldn’t connect to the customer’s request.
- The ERP’s native supersession field was a single-level pointer — Part A superseded by Part B. It couldn’t store or traverse multi-level chains (A→B→C→D). When a customer ordered Part A and the current part was Part D, the system could only resolve one hop. If A→B was entered but B→C and C→D hadn’t been updated yet, the chain was broken at B
- Manufacturer bulletins arrived in inconsistent formats — some as structured data files, some as PDF tables, some as plain-text emails. The product data team spent approximately 30% of their time simply extracting supersession data from these communications before they could begin updating the ERP
- The aftermarket cross-reference problem multiplied the complexity. A single OE supersession could affect 3-5 aftermarket equivalents from different manufacturers, each requiring its own cross-reference update. A Cummins injector supersession touched Dorman, Alliant Power, and 2 remanufactured sources — none of which superseded their own numbers in sync with Cummins
- Return-to-manufacturer windows on superseded inventory were typically 6-12 months from the supersession date. The product data team’s 4-8 week processing lag meant that by the time they identified inventory sitting under old numbers, weeks or months of the return window had already elapsed — and on some items, the window had closed entirely, converting returnable inventory into permanent dead stock
The Solution
We spent 4 weeks in discovery analyzing the distributor’s catalog, supersession history, and the product data team’s workflow — including a forensic reconstruction of the supersession chains on 500 high-value SKUs to measure the actual depth and completeness of the ERP’s current chain data.
The findings were specific: 23% of the catalog’s supersession links were incomplete — missing at least one hop in the chain. 11% pointed to numbers that had themselves been superseded, creating dead-end references. And 8% had no supersession data at all despite the manufacturer having issued a replacement, because the bulletin had arrived during a period when the product data team was processing a backlog from another manufacturer.
The system analyzed signals including:
- Continuous ingestion of manufacturer supersession data from every available source — structured data feeds from manufacturers who provided them, NLP parsing of PDF bulletins and email product change notices from manufacturers who didn’t, and weekly automated comparison against manufacturer online catalogs to detect supersessions that were never formally communicated
- Full-depth chain reconstruction across every part number in the catalog — tracing A→B→C→D→current for every SKU, including the links that had been lost in the 6-year-old system migration, and validating each link against manufacturer confirmation data
- Cross-reference cascade intelligence that automatically traced the aftermarket impact of every OE supersession — when Cummins superseded an injector, the system identified every Dorman, Alliant Power, and remanufactured equivalent that needed its cross-reference updated, and whether those aftermarket parts were still valid for the new OE application
- Inventory obsolescence scoring that continuously evaluated every SKU for supersession risk — combining chain status, manufacturer lifecycle signals, demand velocity trends, and the remaining return-to-manufacturer window to flag at-risk inventory while disposition options still existed
- Historical demand pattern analysis on superseded numbers — identifying which old numbers still received regular customer lookups (indicating the number was embedded in fleet maintenance systems and needed permanent chain resolution) versus which received no lookups (indicating the old number could be safely deactivated)
23% of supersession links in the catalog were incomplete. 11% pointed to dead-end numbers. Every broken link was either a lost sale or an invisible inventory write-down waiting to happen.
Implementation
Deployment occurred over a 01 – 05 period.
Supersession Chain Graph
Full-depth chain resolution across every part number in the catalog — tracing through every level from original to current, including links lost in prior system migrations.
Manufacturer Signal Ingestion
ML and NLP continuously processing supersession bulletins, PDFs, emails, and catalog changes from 200+ manufacturers — eliminating the 4-8 week manual processing lag.
Aftermarket Cross-Reference Cascade
Every OE supersession automatically traced to affected aftermarket and remanufactured equivalents, flagging which cross-references needed updating and which aftermarket parts remained valid.
Inventory Obsolescence Scoring
Continuous risk assessment on every SKU combining chain status, demand velocity, manufacturer lifecycle signals, and return window remaining — triggering disposition recommendations before inventory becomes permanent dead stock.
Epicor Vision Integration
Chain resolution surfaced at point of order — when a customer orders a superseded number, the counter person sees the full chain resolved to the current part with stock status and pricing, without leaving the order entry screen.
Supersession Chain Graph
Full-depth chain resolution across every part number in the catalog — tracing through every level from original to current, including links lost in prior system migrations.
Manufacturer Signal Ingestion
ML and NLP continuously processing supersession bulletins, PDFs, emails, and catalog changes from 200+ manufacturers — eliminating the 4-8 week manual processing lag.
Aftermarket Cross-Reference Cascade
Every OE supersession automatically traced to affected aftermarket and remanufactured equivalents, flagging which cross-references needed updating and which aftermarket parts remained valid.
Inventory Obsolescence Scoring
Continuous risk assessment on every SKU combining chain status, demand velocity, manufacturer lifecycle signals, and return window remaining — triggering disposition recommendations before inventory becomes permanent dead stock.
Epicor Vision Integration
Chain resolution surfaced at point of order — when a customer orders a superseded number, the counter person sees the full chain resolved to the current part with stock status and pricing, without leaving the order entry screen.
Strategic Impact
Lost Sales Recovered
4,200 annual orders that previously returned “not found” or “no stock” — because the customer ordered under a superseded number the ERP couldn’t resolve — now resolved automatically to the current part in stock. The VP of Sales: “We had the part on the shelf. We had the customer on the phone. And we were telling them we didn’t have it because our system couldn’t follow a part number chain 3 levels deep. That’s not a sales problem — that’s a data problem. And it was costing us millions.”
Dead Inventory Dispositioned
$1.1M in inventory sitting under superseded numbers identified and routed to disposition — $480K returned to manufacturers within the return window, $310K cross-referenced to active numbers and reclassified as sellable current stock, $310K marked down and sold through clearance channels. The automated replenishment that had been re-ordering superseded parts was stopped on 2,400 SKUs across 16 branches, releasing $740K in working capital that had been trapped in a cycle of purchasing parts the ERP thought were active but customers would never order under those numbers.
Product Data Team Transformation
The 2-person team that had been spending 80% of their time on manual supersession processing — extracting data from manufacturer bulletins, tracing chains, updating Vision — shifted to strategic catalog work: vendor line reviews, new product onboarding, and margin analysis by manufacturer. The Director of Inventory: “I hired those two people to manage the catalog strategically. For 3 years they’ve been doing data entry. Now they’re actually doing the job I hired them for.”
Key Takeaway
In automotive and heavy equipment parts distribution, the catalog degrades every day. Manufacturers supersede 8,000-12,000 part numbers per year. Every broken chain is either a customer told ‘we don’t have it’ when the part is on the shelf, or inventory sitting under a dead number while the replacement is being purchased alongside it. At 200+ manufacturers with no standardized supersession format, no 2-person team can keep up manually — and the cost of falling behind compounds every month. This distributor didn’t have a purchasing problem or a sales problem. It had a catalog integrity problem that manifested as both — lost sales on the front end and dead inventory on the back end, caused by the same root issue: supersession chains that were incomplete, broken, or missing entirely. The system didn’t replace the product data team’s judgment. It eliminated the manual extraction and chain-tracing work that made it mathematically impossible for 2 people to maintain 95,000 SKUs across 200+ manufacturers in real time.
