A 5,000-Line Excess List Lands on Friday at 4 PM

It’s ten to four on a Friday. An email lands from a materials manager at a contract manufacturer you’ve bought from once before. A program ended, the warehouse needs the space, and attached is a spreadsheet of what’s left. Five thousand lines. Offers by Tuesday.

You can tell from the address line that it went to several other buyers too.

You open the file. Some lines have manufacturer part numbers. Some have the customer’s internal numbers. Half the date code column is blank. One whole tab says “various” in the quantity field. Your buyer scrolls for ten minutes, spots a few parts they recognize, and closes it for the weekend.

I’m not a buyer, and I’ve never bid on an excess list. For almost 18 years I ran a digital agency called Infographic World, and these days I use AI to fix real problems inside businesses. My angle is the flow of work: who touches a file, how long it sits, and what decision it’s waiting on. A 5,000-line spreadsheet with a Tuesday deadline is a flow problem before it’s a buying problem.

Where the hours go on an excess list

The pricing is the part that needs a buyer’s head. Almost everything before it is reading and matching.

• Cleaning. The same part typed five ways. Suffixes, packaging codes, spaces, the customer’s own numbers in place of the manufacturer’s. Lines with no manufacturer named at all.
• Matching. Is this a part you already hold, and how much? Is it one you’ve sold? Is it an alternate for something you sell?
• Demand. Has anyone actually asked you for it? The honest way to answer is to count requests by part over 90 days and set them against the shelf.
• History. What you’ve sold the part for lately, and how fast it moved.
• Questions. Date codes, original packaging, how it was stored, whether moisture-sensitive parts were kept sealed, what paperwork comes with it. These turn into a list of things to ask the seller.
• The offer. Assembled line by line, in whatever format the seller wants.

By hand, nobody does all of that for 5,000 lines in a weekend. So buyers cherry-pick. They scan for part numbers they recognize and bid on those. That works fine for the parts they remember. It misses the parts they don’t, and memory is a poor guide to what the market asked for last quarter.

What the AI-assisted version looks like

It cleans and matches the list

The first step takes the spreadsheet as it arrived and normalizes each line: one part number format, the manufacturer filled in where it can be identified, internal numbers matched to manufacturer numbers where the customer’s history makes that clear. Anything it can’t match with confidence gets marked as uncertain. It doesn’t guess.

It sets each line against your own demand and sales

Every cleaned line is then checked against what you know. How many of these you hold already. How many requests came in for this part, and its known alternates, over the last 90 days. What you sold it for recently, and how long it took.

That’s the column the seller’s list never has: whether anyone is asking.

It sorts the list into buy, pass and ask

The output is three short lists, with the reason on each line.

• Buy candidates. Parts with real request history, that you don’t already hold in quantity, with nothing obviously wrong in the details.
• Pass. No requests, already overstocked, or a commodity part not worth the handling.
• Ask. Parts with demand behind them but a question that changes the answer: missing date codes, unclear packaging, a quantity that says “various.” The system drafts the questions email to the seller.

The buyer prices it

The buyer reads the buy and ask lists, not 5,000 lines. They set a number on every line they bid on, using their own judgment about condition, paperwork and the market. Nothing about price is automated.

Run the numbers

There are two pieces to this. The first is time: lines on the list, times seconds per line by hand, divided by 3,600, gives you the hours a full read would take. The second is margin: the lines that were worth buying, times the average margin a good excess line brings, gives you what a complete read is worth.

Say, for the sake of the math, a list has 5,000 lines and a careful read takes 30 seconds a line. That’s about 42 hours, which is a full work week. These are made-up numbers. It explains why nobody reads the whole thing.

Now say, also made up, that 100 of those lines have real demand behind them, and a buyer cherry-picking from memory catches 30. If an average good line brings $250 of margin once it sells, the 70 lines missed are worth $17,500 on that one list. If three lists like it arrive a month, that’s over $50,000 a month walking out of the inbox in this invented example.

With the list sorted, the buyer might read 300 lines at a minute each. Five hours, spent on the lines that matter.

To run it on your own business, pull three numbers:

1. Excess lists received in the last quarter, and how many lines each had.
2. Hours your buyer spent per list, and how many lists were passed on mostly unread.
3. Margin realized on lines bought from past excess lists, averaged per line.

Picture a Monday, before and after

This scene is made up.

Before. Monday morning, offers due tomorrow. The buyer opens the 5,000-line file again and gets through about 400 lines before lunch, bidding on 20 parts they know well. The rest goes unread. Six weeks later, three of the parts they scrolled past show up in the request stream with buyers asking, and someone else holds them.

After. The same list arrived Friday at four. By Friday evening it had been cleaned and sorted. Monday morning the buyer opens a buy list of 85 lines and an ask list of 40, each with request counts and recent sells beside it. The questions email to the seller went out Friday night, and two answers are already back. The buyer spends the morning pricing. The offer goes out Monday afternoon, covering more of the list than the old way ever did.

What has to be in place first

• At least one quarter of request history from every channel you quote from, including the requests you couldn’t fill.
• A stock list with quantity, condition and trace grade per line.
• Recent sales by part, with dates.
• Whatever cross reference data you already keep for alternates and internal numbers.
• A written rule from your quality process on what condition and paperwork you’ll accept from an OEM or EMS excess lot.

Guardrails worth keeping

• The buyer prices every line. The system sorts. It doesn’t bid.
• Uncertain matches stay flagged. A part number the system isn’t sure of goes to a person, never straight to the buy list.
• Condition and paperwork come before price. A buy candidate with a storage question is an ask until it’s answered.
• The seller’s list stays private. It was sent to you in confidence. Keep it in business-grade tools, out of public model training.
• A pass is reversible. The buyer can pull any line back on judgment.

How you’ll know it’s working

• Hours from list received to offer sent.
• Share of lines on each list that were actually evaluated.
• Lines bought per list that had request history behind them.
• Sell-through on excess purchases within 90 days.
• Offers accepted, as a share of offers made.

Where I’d start

Find the last excess list you passed on or only half read. Run it against 90 days of your request history and see which lines had buyers asking. That one afternoon tells you whether the old way is missing money, and how much, before you change anything.

If that list is still in your inbox, the free week on the Hunt page builds the request count by part that any excess list should be read against.