Why Your Keyword Alert Misses the Lot You Wanted Most

You set the alert eight months ago. The part number, the two alternates you know about, the manufacturer name. It fires a few times a week. Most of what it brings you is a broker’s stock offer for twenty pieces at a price you could have guessed, or a listing three other people on NetComponents saw the same minute you did.

Meanwhile a pallet of sealed reels from a closed contract manufacturer sold last Tuesday through an industrial auctioneer. The listing said “lot of electronic components, approx 40 lbs, sold as is.” Four photos. In the second photo, a manufacturer label with a part number your buyers asked for eleven times last quarter. Your alert never fired, because nobody typed that number anywhere.

I should say up front that I have never pulled a reel from a shelf or quoted a line on Brokerforum. I ran a digital agency for close to eighteen years, and these days I use AI to fix specific problems inside businesses. What I know is how work moves through an office, where it piles up, and which steps a person is doing by hand because nobody ever built the step properly.

How search actually works on a marketplace or an auction site

A search box on a parts marketplace, an auction platform or a general marketplace does one thing. It matches the words you typed against the text fields the seller typed. Title, description, maybe a category.

So a search is only as good as the seller’s typing. On a parts marketplace the seller knows what they have. On an auction or surplus listing, the seller is an auctioneer’s photographer, a trustee’s clerk, or a facilities manager told to clear the building by Friday. They type what they see from across the room.

• “Misc electronic components”
• “Lot of chips and boards”
• “Electronics, 3 pallets, by weight”
• “Contents of parts room, as is”

The photos in those listings often show sealed moisture barrier bags with humidity cards, reels with original manufacturer labels and date codes, trays of ceramic parts with 5962 numbers on the lids. The markings in the photos say what the lot is. The seller’s description usually does not.

Your keyword alert reads the description.

Why obsolete parts hide under the wrong number

Even when a seller types a number, it is often not the one you are watching:

• A package suffix. Same die, different package, different suffix, and the seller typed the one on the label.
• A distributor stock code, copied off the bin label from whoever sold it originally.
• A superseded number. The part was renumbered after an acquisition or a product line move. Your buyers ask for the new number. The shelf label has the old one.
• A customer internal number. OEMs and defense primes label stock their own way, and the manufacturer number is on the reel, not the bin.
• No number at all. The seller photographed the reel and wrote “electronic parts.”

Obsolete parts are the worst case. The older the part, the more renumberings it has been through, and the less likely anyone on the selling side knows what it was. Legacy FPGAs, old DRAM and flash, microcontrollers from programs that ended ten years ago. The defense, aerospace, medical and industrial programs that still need them run for decades, so the demand is real. The listing just does not say so.

Alerts fire on the easy lots, which is the problem

What an alert does find is a listing with the exact manufacturer part number in the title, typed correctly. That seller knows what they have, prices accordingly, and is seen by every other broker with the same alert at the same moment.

So the alert sorts the market in exactly the wrong direction. The lots described correctly are visible, priced and contested. The lots described badly are invisible, cheap, and yours if you are reading. The margin sits in the gap between the description and the markings, and an alert cannot see across it.

Most badly described lots are junk. Mixed bins of pulled parts, oxidized leads, reels with no labels. Reading a hundred to find one that matters is the normal ratio. The method is not “every vague listing is a find.” It is “the finds are only in the vague listings, so somebody has to read them.”

What a target list looks like instead

A keyword alert starts with what you already know you want. A target list starts with what your buyers asked for. Take every request that came through NetComponents, Brokerforum, IC Source and the shared inbox over the last ninety days. Count them by part. Set them against the shelf. The parts asked for most that you could not quote are the top of the list.

Then expand each part the way a seller might mislabel it:

• The base number and every package suffix you would accept.
• Known alternates and second sources.
• Superseded and predecessor numbers.
• The manufacturer’s old name, if there was an acquisition.

That expanded list is what you check listings against daily, photos read and not just titles. Two hundred parts, each written ten ways, is more than one buyer can hold in their head. A system can hold it, and a person can review what it finds.

The feeds everyone sees

A second stream deserves a word, because most distributors receive it and skim it. Excess postings on the marketplaces. Broker stock offers in the inbox. OEM and EMS excess sheets, usually a spreadsheet with a few thousand lines, sent to a mailing list.

These are not hidden. Everyone in your trade gets the same sheet. The edge is not in getting it. It is in reading it against demand instead of by eye. A buyer skimming three thousand lines looks for numbers they recognize. A target list run against the same lines looks for numbers your buyers asked for, including the ones your buyer would not recognize because the request came in through a colleague’s inbox eight weeks ago.

Picture a Monday (made up)

This scene and its numbers are made up, to show the shape of the work.

Before. Your buyer opens the inbox at eight. Forty messages, two alert hits, one EMS excess sheet. She skims the sheet for ten minutes, recognizes two numbers, flags one, and spends the rest of the morning quoting. An industrial auctioneer closed a lot on Sunday with three reels she would have paid for, listed as “lot of SMT components.” Nobody saw it.

After. The same Monday. Overnight, the target list was checked against the weekend’s new listings and the excess sheet. Her queue has nine items, ranked by how many of her buyers asked for the part in the last ninety days. Four she dismisses in a minute each: wrong package, no photos, obvious pulls. Two she bids on. One is a sheet line that matched a part asked for eleven times, and she sends an email before nine. The auctioneer’s lot is still a loss. Next time it will be in the queue.

What has to be in place first

• An export of requests from each marketplace and the shared inbox, going back at least one quarter.
• A stock list from the ERP, so requests can be set against what you hold.
• Someone who knows the product well enough to write the alternates and superseded numbers for the top parts. That is judgment, not a database lookup.
• Fifteen minutes a day on the queue, which is enough if it is ranked.
• A clear rule on trace grade, so a find with no paperwork is priced as a no-trace lot and not as a franchised one.

Guardrails worth keeping

• The system proposes, the buyer decides. Nothing gets bid on without a person looking at the photos.
• Keep the demand count honest. A buyer who asked once is not a buyer who asked every week, and a request for full trace is not demand for a no-trace lot.
• Log the dismissals. A lot you passed on and later regretted is the best lesson the list will get.

How you’ll know it’s working

• Lots reviewed per week, including the badly described ones. If this is not up, nothing else will be.
• Finds per hundred lots reviewed. Low is normal. Rising is the signal.
• Share of finds from listings with no correct part number in the text.
• Excess sheet lines actioned per sheet, against the old skim.
• Spread realized on finds, priced at the trace grade they actually had.

Where I’d start

Take one quarter of requests, rank them by part, and write the alternates and superseded numbers for the top twenty-five by hand. Then spend one week reading every vague listing in your categories against that list, photos first. You will know by Friday whether the gap is real.

If you would rather see it run than build it, the free week on the Hunt page starts with exactly that list.