The Sale You Lost Never Shows Up in Your Inventory Report

A plumber walks up to the counter at 7:15 and asks for twenty of a fitting. The branch has four. The counter person checks the other branch, which has none, and says the vendor can have them Thursday. The plumber says never mind and heads to the supply house down the road.

Nothing gets written down. No quote, no order, no note. As far as your ERP knows, that plumber never existed.

Next month, the buyer looks at that fitting. Sales are down. The reorder point, calculated from sales, drifts lower. The next stockout comes a little sooner, the next plumber walks a little faster, and the sales history tells the same false story again.

I’ll say up front that I’ve never been a buyer. I spent almost 18 years running a digital agency, and today I put AI to work on real problems inside businesses. I don’t know your vendors or your lead times. What I do know is how decisions get made from the data in front of people, and how often that data is missing the one fact that matters.

Why sales history lies about demand

Here’s the core problem. Your sales history only records the sales you made. When an item is out of stock, demand doesn’t stop. It just stops being recorded.

Take an item that sells evenly and is out of stock three weeks out of twelve. Its sales history shows about a quarter less than real demand. The reorder point gets set from that lower number. Lower stock means more stockouts, which means even less recorded demand. Left alone, the cycle feeds itself.

The quoting article made the point that every quote request is a demand signal. Purchasing is where that signal should end up. In a lot of supply houses, it never gets there.

Where the demand goes missing

• Counter walkaways. A common lost sale, and one of the least recorded. The customer leaves, and nothing is entered.
• Phone and email requests. “Do you have 40 of these?” “No.” End of conversation, no record.
• Quotes lost on availability. If the quote system has a lost reason, it can default to “price,” even when the real reason was stock.
• Partial fills. The customer wanted 50, you had 30, they took 30 and bought 20 elsewhere. The sale looks complete.
• Branch transfers. Stock moved from one branch to cover another hides demand at the branch that ran short.

On top of that, two habits make it worse.

• Reorder points set once. Min and max get entered when the item is set up and revisited only when something goes badly wrong.
• Seasons caught late. Condensing units in spring, hydronic parts in fall, storm-season items on the coast. By the time sales history shows the swing, the season is half over and the vendor’s lead time has stretched.

What AI does for the buyer

It captures lost sales from every place they happen

The system picks up demand from the sources you already have: quote lines marked unavailable, quotes lost on stock, emailed requests that didn’t turn into orders, and a one-tap “didn’t have it” code at the counter. Each one becomes a lost-sale line with item, quantity, branch and date.

It rebuilds demand for the weeks you were out

For periods when an item was out of stock, the system estimates what demand would have been, using lost-sale lines and the item’s history when it was in stock. The buyer sees recorded sales and estimated demand side by side, so the gap is visible instead of buried.

It suggests reorder points, and the buyer approves them

Using that demand, the vendor’s actual lead times and the season, the system suggests a reorder point and order quantity for each item. It shows why: demand rising, lead time longer, season coming. The buyer accepts, adjusts or rejects. Nothing goes to a vendor without a person signing off.

It flags what’s moving before the report does

A weekly list shows items trending up across branches, items with repeated lost sales and no stock, and seasonal items whose lead time means the order should go out this week. It also flags large one-time project orders so they don’t distort the forecast.

Run the numbers

The formula: lost-sale lines per week × average line value × gross margin × 52 = gross margin lost per year.

Say, for the sake of the math, a three-branch supply house loses 120 lines a week to stockouts across its counters, phones and quotes. Say the average line is worth $180 and gross margin is 25 percent. All three numbers are made up.

That’s 120 × $180 = $21,600 in lost sales a week. Over 52 weeks, about $1.12 million. At 25 percent gross margin, about $281,000 in gross margin that never arrived.

That number leaves out the bigger cost. A contractor who walks out twice for the same item starts calling the other supply house first, for everything.

Now the other side of the ledger. Fixing this means carrying more of some items. Keep the made-up example: say you add $60,000 of inventory on the items that run out most, and your carrying cost is 20 percent a year. That’s $12,000 a year to recover a share of $281,000. Even recovering a modest fraction of the lost margin covers it many times over.

To run it for real, you need three numbers:

1. Lost-sale lines per week. You probably don’t have this yet, and that’s the point. Capture it for four weeks at one branch.
2. Average line value, from your invoice history.
3. Gross margin on the items involved, from the ERP.

Picture a Monday, before and after

This is a made-up scene. Picture the buyer for an HVAC supply house with three branches, the first week of April.

Before: the buyer runs the suggested order report, which is built on last year’s sales and min and max settings from years ago. It looks fine. Two weeks later, all three branches are short on the same condenser fan motors, the vendor’s lead time has doubled, and the counter is apologizing all day.

After: Monday’s list shows that same motor at the top. Lost-sale lines on it started climbing at one branch ten days ago. Estimated demand is 40 percent above recorded sales. Lead time from the vendor is now four weeks. The suggested reorder point has moved up, with the reasons written next to it. The buyer raises the order, moves a few units between branches, and places it that morning.

The buyer made the call. The difference is that the buyer could see it coming.

What has to be in place first

• A way to capture lost sales. A lost-sale code in the ERP, used at the counter and on the phone. Without it, the rest is guesswork.
• Honest lost reasons on quotes. “Price” and “availability” kept separate, with no default answer.
• Real vendor lead times. What the vendor actually delivers, not what the catalog says.
• Clean sales history. Branch transfers separated from customer sales, and superseded items linked to their replacements.
• One buyer who owns the reorder points. Suggestions need somebody who reviews them weekly.

Guardrails worth keeping

• The buyer approves every change. Reorder points, order quantities and purchase orders all go through a person.
• One-time orders get flagged. A big project order shouldn’t set stock levels for the next year.
• Don’t stock for one customer on a promise. If a single contractor drives the demand, the buyer should talk to them before buying deep.
• Watch cash. Suggestions should show total dollars added and removed, so the buyer sees the inventory effect at a glance.
• Keep the reasons. Every approved change should record why, so next season starts from evidence.

How you’ll know it’s working

• First-time fill rate at the counter and on orders.
• Lost-sale lines per week, by branch.
• Stockouts on your top 200 items.
• Inventory turns.
• Share of inventory with no sale in 12 months, which should hold steady or fall while fill rate rises.

Where I’d start

Start capturing lost sales at one branch counter for four weeks. Nothing else. Then set those four weeks against sales history on your top 100 items. If the gap is real, it’ll be obvious, and you’ll know exactly which items to look at first.

If this sounds like your week, send me a note through the contact page.