The Contractor Who Quietly Started Buying Elsewhere

Here’s a made-up scene that may sound familiar. Your outside rep is at a lunch in September with a mechanical contractor they’ve called on for six years. Somewhere between the sandwiches and the check, the contractor’s purchasing manager mentions that they’ve been getting their water heaters from another house since spring. Better lead times, he says. Nobody asked.

Back at the office, the rep pulls the account. Total sales are down 11 percent from last year. Not alarming on its own, and it never made the monthly report’s red list. But water heaters went from a steady line every month to nothing since March, and gas valves went with them. Fittings and PVC kept coming, so the total looked fine.

Six months of a category gone, and the first warning was a lunch.

I want to be clear about my background. I’ve never carried a bag or worked a territory. I ran a digital agency, Infographic World, for almost 18 years, and now I use AI to fix problems inside businesses. What I’ve learned is where information gets stuck between the system that records it and the person who could act on it. In distribution, a lot of it gets stuck in invoice lines nobody has time to read.

Why the slip goes unnoticed

The data to catch this usually already sits in the distributor’s own system. It’s just not shaped for the person who needs it.

• Reps cover too many accounts. A rep with 120 or 150 accounts calls on the 25 that buy the most and the ones that call them. The middle of the list gets a visit when there’s a reason.
• Reports show totals. Sales by account, this month versus last year. A drop in one category hides behind growth in another, or behind one big job.
• The signals live in different places. Lost quotes are in the quoting system. Counter sales show up at a branch the rep doesn’t work out of. Order size shrinks gradually. A contractor’s new project manager starts buying from a house they used before.
• Seasonality muddies the picture. A drop in July might be normal for that trade, or it might not. Without the account’s own history beside it, nobody can tell quickly.
• Nobody owns the question. The rep assumes the manager watches the reports. The manager assumes the rep knows the account.

So the question “which of my accounts changed how they buy?” goes unanswered for months, even though the answer is sitting in invoice lines.

What a weekly coverage list looks like

The AI-assisted version is simple to describe. Every week, each rep gets a short list. Not a dashboard to explore, a list to act on.

It reads every account against its own history

For each account, the system compares the last 60 to 90 days against the same months last year and against that account’s usual rhythm. It looks at product categories, number of ship-to addresses or jobs, average order size, which branches the account buys from, and how often it orders.

It flags patterns, not just totals

The flags are specific. Bought water heaters every month for 14 months, none in 75 days. Ordering from two job sites instead of five. Average order down by half since June. Quotes lost three times this quarter on availability. Stopped buying at the north branch, still buying at the counter downtown.

It suggests a likely reason and a call plan

Next to each flag, the system offers the most likely explanation from the data it can see: lost quotes with a reason recorded, a lead time problem on that category, a price increase that went through in the spring, a new buyer name on recent orders. Then a short plan for the call. Ask about the water heater jobs. Bring current lead times. Check whether the price level on that category is out of date.

The rep decides what’s true and what to do. Any price change goes through whoever owns pricing. The list is a prompt, not a verdict.

It learns from what the rep marks

When the rep closes a flag, they mark what happened: job ended, seasonal, lost on price, lost on availability, new contact, back on track. That feedback keeps next week’s list shorter and shows the sales manager which reasons keep coming up.

Run the numbers

The formula is the number of accounts showing a slipping pattern, times the average annual purchases in the slipping categories, times the share you win back, times your gross margin.

Say, for the sake of the math, a distributor has 600 active accounts and 8 percent of them show a real category slip in a given quarter. That’s 48 accounts. Say the slipping categories averaged $30,000 a year in purchases per account, all made-up figures. That’s $1,440,000 in annual sales drifting out the door.

If reps win back a quarter of those because they called in week three instead of month seven, that’s 12 accounts and $360,000 in sales. At a made-up 22 percent gross margin, it’s $79,200 in gross margin a year. From one quarter’s list.

The share recovered is the number to be most careful with. Some accounts leave for reasons a call won’t fix. Start with a modest guess and replace it with your own result after a quarter.

To run it for real, pull three numbers:

1. Accounts whose spend in any one category fell by half or more against the same months last year. Your ERP can export invoice lines by account, category and month.
2. The average annual spend in those categories for those accounts.
3. Your gross margin on those categories.

Picture a Monday, before and after

This is a made-up scene. Monday at 7:30, a rep plans the week from memory and the monthly report. Four visits to top accounts, two callbacks, one lunch. The mechanical contractor from the opening isn’t on the list, because nothing about its total looks wrong.

Same Monday with the weekly list. Six accounts, each with one line. The mechanical contractor: water heaters and gas valves at zero since March, two quotes on water heaters lost in April with “lead time” as the reason. A plumbing contractor: ship-tos down from four to one, likely a job ending, check what’s bidding. An electrical contractor: average order down 40 percent, new buyer name on invoices since August.

The rep calls the mechanical contractor that morning with current water heater lead times. The plumbing contractor turns out to be a finished job, marked as such. The new buyer at the electrical contractor gets a visit Thursday.

What has to be in place first

• Consistent product categories. If one branch files water heaters under “plumbing” and another under “WH,” the pattern breaks. Clean the category tree before anything else.
• Invoice history by line, at least 24 months. Enough to compare against the same months last year and see each account’s rhythm.
• Ship-to or job information on orders. It’s the best early signal that a contractor is moving work elsewhere.
• Lost quote reasons, even rough ones. Price, availability, lead time, unknown. The quoting article covers why that history is worth keeping.
• A clear owner for each account. A flag with nobody’s name on it gets ignored.

Guardrails worth keeping

• Reps decide what a flag means. The data suggests. The person who knows the contractor confirms.
• No automatic messages to customers. An unsolicited email saying “the system noticed you stopped buying valves” lands badly. The call is the rep’s.
• Pricing changes go through pricing. A slipping account is a reason to review a price level, not permission to cut one on the spot.
• Customer data stays inside the business. Use business-grade tools that don’t train on your sales history.
• Don’t turn it into a leaderboard too early. If reps fear the list, they’ll mark everything “seasonal.”

How you’ll know it’s working

• Share of flagged accounts contacted within seven days.
• Share of flagged accounts whose category spend returns within 90 days.
• Gross margin recovered on those accounts, quarter over quarter.
• Flags marked as false alarms. This should fall as the list learns.
• Hours reps spend building their own account reports. This should fall too.

Where I’d start

Pick one rep who wants better information and one product category family where slips hurt most, often the one with big-ticket equipment. Pull 24 months of invoice lines for that rep’s accounts and build the weekly list for that category alone. Run it for eight weeks and have the rep mark every flag.

At the end you’ll know whether the flags were real, which reasons repeat, and roughly what the recovered share looks like with your customers.

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