A Thousand RFQs a Day and Time to Quote Two Hundred

If you run a broker desk, you know the number at the top of the shared inbox. It was close to zero when everybody went home. By the time the coffee’s ready it’s in the hundreds, and by lunch it has passed whatever the desk can physically quote today.

So the desk starts at the top, or with senders it recognizes, and works until the day runs out. The rest gets skimmed or never opened.

Nobody chose that order. It’s the order the email arrived in. And somewhere deep in the pile there’s a contract manufacturer with a line down, a real quantity, a target price and a Friday deadline, sitting behind a few hundred price checks from people who were never going to place an order.

I’ll be upfront about my angle. I’ve never sat on a broker desk or quoted a reel of anything. I ran Infographic World, a digital agency, for almost 18 years, and now I use AI to fix real problems inside businesses. What I’ve learned to see is how work moves through an office and where it piles up. A request queue sorted by arrival time is one of the clearest pileups I know.

The requests you can’t fill at all are a separate story, told in The Unanswered RFQ Is the Most Expensive Thing in Your Inbox. This one is about the requests you could quote today, and which go first.

Where the desk’s day actually goes

Follow one request from arrival to quote and most of the clock runs before anyone thinks about price.

• Reading it. Part number, quantity, date code needs, packaging, trace requirement, deadline. Half the requests are one line. The other half are pasted spreadsheets with the important note at the bottom.
• Checking the shelf. Quantity, condition, trace grade, date code, and what’s already promised to someone else.
• Checking recent buys. If it isn’t on the shelf, what the company paid last time, from whom, and how long ago.
• Sizing up the sender. A buyer you know, a company you’ve shipped, or a name you’ve never seen on a free email address.
• Writing the quote, and deciding whether it also deserves a phone call.

Every request gets roughly the same treatment, in arrival order, so the desk spends as long on a fishing trip as on a real order.

What a real buyer looks like on the page

A good salesperson spots a real buyer in seconds, but only in the requests they reach. The signals are written in the request itself.

• A named company, with a signature that matches a real business. Better still, one that has bought from you before.
• A quantity that fits a build. A reel, a tray, a few thousand pieces, a number that makes sense for how the part gets used.
• A target price. A buyer who gives a target is usually trying to close.
• A repeat sender, asking again for a part they asked about last week and still haven’t found.
• A deadline. “Line down,” “need by Friday,” a ship-to date. Urgency is what hands an order to whoever answers first.

Price fishing reads just as plainly: the same request sent to every distributor on the feed, no company name, a quantity that changes each time.

None of this is secret knowledge. It just can’t be applied to a thousand requests a day by hand.

What an AI-sorted queue looks like

It reads and scores every request

The first step reads each request and pulls out the part, quantity, deadline, sender, company and any target price. Then it scores it against the signals above, using the rules your best salesperson would write down given a free afternoon.

What comes out is a queue ordered by how likely each request is to become an order, with the reason printed on every line. “Repeat customer, line down, quantity matches last order.” A salesperson sees at a glance why something is at the top, and moves it if they disagree.

It drafts the quote from what you have

For the requests near the top, the system pulls stock, condition, trace grade and date code from the ERP, plus the last few buys and sells on that part. The draft arrives with all of that laid out. The salesperson sets the price, checks the details and sends.

It flags the ones worth a call

A few requests a day deserve more than an email, like a line-down note from a known customer. Those get flagged for a phone call, and the salesperson decides whether to make it.

The score is a suggestion. Overrides are worth keeping, because they’re how the rules get better. Price, trace and condition calls stay with people.

Run the numbers

The formula in words: quotes sent per day, times win rate, times average margin per won order, is what the desk earns in a day. The question is how much that moves when the strongest requests get answered first.

Say, for the sake of the math, a desk sends 200 quotes a day and makes $400 of margin on an average won order. These are made-up numbers, not a benchmark. Say, also made up, that 50 requests a day carry most of the buyer signals and win at 12 percent when quoted, while the rest win at 2 percent.

In arrival order, the desk happens to reach about half of those 50. That’s 25 strong quotes at 12 percent, or 3 orders, plus 175 weak quotes at 2 percent, or 3.5 orders. Call it 6.5 orders and $2,600 a day.

Sorted, the desk quotes all 50 strong requests first, then 150 others. That’s 6 orders plus 3 orders, or 9 orders and $3,600 a day. Same people, same 200 quotes, about $1,000 more margin a day in this made-up case. Over 250 working days that’s $250,000. Cut my invented numbers in half and the shape still holds.

To run it for real, pull three numbers:

1. Requests received and quotes sent per day, averaged over the last quarter.
2. Win rate split by signal. Take 90 days of quotes with outcomes, tag which ones had a named company, a target price and a deadline, and compare win rates against the rest.
3. Average margin per won order from the desk.

Picture a Monday, before and after

This scene is made up.

Before. Two salespeople start at the top of a deep weekend inbox and work down. By eleven they’ve sent 60 quotes, mostly to strangers checking prices. At 11:30 one of them opens a request from a contract manufacturer they’ve shipped twice, sent Saturday night: line down, 4,000 pieces, a target price, ship today. They have the part. The buyer replies that it was placed an hour ago.

After. Same Monday. The queue opens with that request at the top, marked “repeat customer, line down, in stock, flag for call.” The draft quote shows the stock, date code and trace grade, and what that customer paid last time. The salesperson sets a price, calls, and has a PO by 9:15. The price checks get answered after lunch.

What has to be in place first

• Quote history with outcomes. Won, lost, and the reason when anyone wrote it down.
• A stock list the system can read, with condition, trace grade and date code per line.
• Recent buy and sell history by part.
• A customer list that ties sender addresses to companies, so a repeat buyer is recognized as one.
• Your salespeople’s rules, written down. What makes them pick up the phone?

Guardrails worth keeping

• A low score moves a request down. It never deletes it. Every request still gets counted, because unfilled demand matters later.
• People price every quote. The system drafts. Salespeople decide.
• Trace and condition go on every draft as fact. Nothing gets assumed to fill a gap.
• New buyers get a fair score. No history isn’t the same as a bad history.
• Customer data stays inside your business. Use business-grade tools that keep your quotes out of public model training.

How you’ll know it’s working

• Win rate on the top 50 requests each day, against the rest.
• Median time from arrival to quote for top-scored requests.
• Orders per salesperson per week.
• Overrides per day. They should fall as the rules improve.
• Margin per day from the desk.

Where I’d start

Pull 90 days of quotes with outcomes. Tag a sample of 200 by hand for the five signals and compare win rates. If the signals separate orders from price checks at your company, score the live queue for one salesperson for two weeks and compare against the rest of the desk.

If your inbox looks like this, the free week on the Hunt page starts from the same raw material: your own request history, counted by part over 90 days and set against the shelf.