SKU rationalization that runs every week, not once a quarter.
We encode your purge rules — no sale or receipt in 24 months, weeks of supply past two years — inside the POS/ERP your stores already run. Our AI reads every item record against them on a schedule, and only the exceptions reach a merchant.
Official services partner of the platforms defining AI
NVIDIA
Anthropic
A SKU rationalization process that runs once a quarter is already behind.
From retail operations and merchandising teams we have sat with: a hardware retail co-op member running ~20 stores and ~$100M in revenue, ~200K item records fanning out to 4.4 million item-store rows on an Epicor Eagle-class POS/ERP — and an item master that only gets cleaned when someone declares a quarter for it.
“Anything not sold or received in 24 months.”
The rule is simple; running it is not. When the purge finally ran, the line got drawn at 2022 and 100,000 items came out of the item master in a single pass — a measure of how long the master had gone untouched.
Every acquisition bolts on another item list
Six acquisitions a year, and each one arrives with its own item file — the same product under four item numbers, renamed vendors, duplicates nobody reconciles. The item master only grows.
Dead SKUs still burn tag labor
“Is the juice worth the squeeze?” Price changes get pushed for items with zero on hand — inside 189,000 physical price changes a year — because the item master never told the pricing run those SKUs were gone.
The sync inherits the mess
A weekly full snapshot every Sunday at 4am, plus hourly deltas of retail, promo, and quantity changes, reconciling 4.4 million item-store rows against ~200K item records. Every duplicate and dead item is drift the sync has to carry — the cadence is your maximum drift window, and a dirty master spends it.
The items customers remember get no attention
The two or three hundred known-value SKUs — the ones where most people can figure out what Windex should cost — set your price image. They get no attention while the team is buried under 100,000 records nobody maintains.
A dirty item master taxes everything that touches it.
Measured inside a hardware retail co-op member’s operations. Your numbers will differ — the item-master audit puts figures on one item file before anything changes.
From an anonymized engagement — a ~20-store, ~$100M-revenue hardware retail co-op member on an Epicor Eagle-class POS/ERP
Exception reporting turns the purge into a standing rule.
No new system for your merchants to learn. Our AI reads every item record against your purge rules every week — your people decide only what the rules can’t.
- 01
Item-master audit
One item file. We quantify the damage — items with no sale or receipt in 24 months, weeks of supply past two years, sell-through rates near zero, duplicates from acquisitions — before anything changes.
- 02
Encode the purge rules
Your thresholds — the 24-month line, the two-years-of-supply flag — encoded against the POS/ERP you already run. The AI reads all ~200K item records against them on a schedule; nothing waits for a declared quarter.
- 03
Exception-only queue
The rule proposes; the merchant disposes. Every flagged item lands with its last-sale and last-receipt evidence, a human approves each removal — trust but verify first — and 30/60/90-day operating rules keep the queue cleared instead of admired.
Exception based reporting beats the quarterly spreadsheet purge.
The software sells another dashboard of the whole item file — the same 200,000 rows every Monday, none of them decided. Assortment suites ship with generic lifecycle thresholds; your stores run on exceptions — the 24-month line, the acquisition duplicates, the special orders that never should have become SKUs.
We are neither. We are an AI operating partner: we encode your purge rules into the systems your stores already run, and our AI keeps reading the item master after go-live — so the master stays clean between quarters and the next 100,000-item purge never becomes necessary. One boundary, plainly: what to do with the dead stock on hand — mark it down or return it — is excess-inventory work. This page decides which SKUs should exist at all.
We drew the line at 2022 and took out 100,000 items. Now the line runs itself — I only see the handful the rule can’t decide, with the last-sale evidence attached.
Led by Peter Enestrom.
Asked by retail ops and merchandising leads.
The straight answers, before you book anything.
Deciding which SKUs should exist in your item master at all — on evidence, not feel. Each item carries its last-sale date, last-receipt date, and weeks of supply on hand; anything past the line your merchants set — typically no sale and no receipt in 24 months — gets flagged for removal, with a human approving every deletion. It is item-master hygiene, not a clearance event.
Three steps, then a standing rule. First, an audit of one item file quantifies the dead records, the two-years-of-supply cases, and the acquisition duplicates. Then your purge rules are encoded against the POS/ERP you already run, and AI reads every item record against them weekly. From then on the quarterly purge disappears — flagged items reach a merchant as a short exception queue, evidence attached, on 30/60/90-day operating rules.
Management by exception applied to the item master. Instead of a person reading 200,000 rows, exception reporting reads all of them against your rules and shows a person only what the rules can’t decide. Exception based reporting flips the workload: the whole file is monitored continuously, and the human sees the few dozen items that need a judgment — with the AI’s reasoning and the evidence attached.
No — that is the disposition question, and it belongs to excess-inventory work: mark the units on hand down or return them to the vendor, per SKU. SKU rationalization is the upstream question: whether the item should exist in the item master at all. A clean master makes the disposition work smaller; the two run side by side but decide different things.
The opposite. Your price image is set by the two or three hundred known-value items customers actually remember — most people can figure out what Windex should cost. Those items are never purge candidates; they sell constantly. Clearing 100,000 dead records frees the team to actually watch the handful of SKUs that shape perception.
It makes it smaller. The weekly full snapshot is not freshness — it is reconciliation — and the cadence is your maximum drift window. Every duplicate and dead item in the master is drift the sync has to carry across 4.4 million item-store rows. Purge them, and the same Sunday 4am full sync plus hourly deltas has less to reconcile and less to break on.
Start with one workflow.
Tell us where your team loses hours. We will come back with a straight answer on whether AI can help, what it would take, and what it would pay.