Writing · Operations

The ten leaks.

What a job walk finds in a service business, what each one costs, and the ten questions I ask while one real job moves through the shop.

The first thing I do with every client is the same. I take one real job — not a typical job, a real one, with a name on it — and follow it through the whole company. Where it came from, who priced it, how it got on the board, what went on the truck, what got written down, when the invoice went out, whether anybody had to go back.

Every step. Every wait. Every leak.

It takes about an hour. It works over video, and it works better that way than it sounds: the owner narrates instead of performing. And it finds the same ten things in nearly every shop I have ever looked at.

The ten questions

These are the questions, in this order, every time. Nothing clever about them. The order is what makes them work, because each answer sets up the next one.

1. Question 1Where does work come from, and who answers when it calls? How long before someone responds?
2. Question 2Who prices the job, from what — a rate sheet with a date on it, or a gut feel? Is the estimate written down?
3. Question 3How does the job get scheduled? Where does it wait, and for how long? Who knows the schedule — one head, or a board everyone can see?
4. Question 4What happens the morning of the job? Parts on the truck or a supply-house run? How many trips per job?
5. Question 5Who does the work, and what's written down when it's done? Does the ticket match what actually happened, or what the estimate said?
6. Question 6What gets added and never billed — the "while I'm here" work, the callback, the favor?
7. Question 7How fast does the invoice go out — same day, or when somebody gets to it? Who chases it when it isn't paid?
8. Question 8How often does a crew go back to redo something? Is that tracked anywhere, or just remembered?
9. Question 9What does the owner personally touch in this chain that nobody else can? Why?
10. Question 10If tomorrow's crew doubled, which step breaks first? That step is the bottleneck, and the bottleneck sets the shop's real capacity.

Question 9 is the one this whole practice is built around. Its answer is your owner-touch count. It is a number, not an opinion, and it is the number the two-week test settles.

The ten leaks, and how each one is measured

Ten leaks show up over and over. Each one maps to a number you can compute, which is the point — a leak that cannot be measured cannot be verified, and an unverified fix is a feeling.

What each one costs

Here is how to put a dollar figure on each, using your own numbers. No estimates from me. Your arithmetic, on your data.

1. Stale prices

Take the labor dollars you collected last year. Divide by the hours you paid for. That is your effective rate. Subtract it from the rate on your wall, and multiply the gap by your paid hours.

A shop I use as an example posts $135 and collects $92. That $43 gap across about 6,700 paid hours is Δ $288,000 a year. It is usually the biggest single number on the page.

2. Unbilled add-on work

Pull 20 tickets. Read each one against what the crew actually did that day. Count the additions that never got billed, and add up the dollars. Divide by 20 to get a per-job figure, then multiply by the jobs you run in a year. Twenty minutes a job at your posted rate is a lot of money by December.

3. Estimate-versus-actual drift

Ten finished flat-price jobs. Actual hours divided by estimated hours. If the answer is 1.2, you gave away a fifth of the labor on every one of them. Multiply the extra hours by your effective rate and by the number of flat-price jobs you sell a year.

4. Slow invoicing

Receivables divided by daily revenue gives you AR days. Every day of AR is one day of revenue sitting in somebody else's bank. At $3M a year, one day is about $8,200. Going from 55 days to 35 puts roughly $164,000 back in your account, once.

5. Callbacks and rework

Jobs redone divided by jobs completed, per month. If nothing tracks it, count one month by hand. Then cost it three ways: the crew hours you paid twice, the parts, and the job you could not do that day. The third one is the biggest and nobody counts it.

6. Double data entry and paper shuffles

Measured by close lag — the days after month end before your books are closed. Thirty or forty days is common. It costs you two ways: office hours spent copying numbers from one place to another, and every decision you made in the meantime on a guess.

7. Truck-roll waste and parts runs

Count trips per job for one week. Multiply the extra trips by the round trip time, then by what that crew costs you loaded per hour. Then ask the better question: how many more jobs a day would each truck do without them?

8. Scheduling by memory

Jobs done on the promised day, divided by jobs scheduled. Count one month off the board by hand. Whatever it comes to, compare it against what you tell customers on the phone. The distance between those two is paid for in callbacks, in overtime, and in the accounts that quietly stop calling.

9. The owner as a required step

This is a count, not a dollar figure, and it is the one that matters most. Legal pad on the desk for one week. A mark every time a decision lands on you that nobody else could have made. Twenty or more a week is normal in a $3M company. Under five is a company a bank would lend against.

10. Nothing written down

There is no separate measure for this one, because it is the reason all nine above are true. If how the work gets done lives in three people's heads, then every fix above is temporary. It leaves when they leave.

The fixes are boring, on purpose

Every one of these has a fix that fits on a line. That is not me simplifying. Boring is what makes them stick, and sticking is what makes them show up in your bank statement inside 90 days.

• A rate sheet with a date on it, reviewed on the calendar.
• A ticket that gets audited weekly — ten of them, by somebody who did not write them.
• Invoices out the same day. Not Friday.
• A schedule board everybody can see, in the building.
• A parts checklist by job type, so the truck is loaded before it leaves.
• One page of written steps for each task you repeat.

Six habits. None of them costs money. All of them cost attention, which is why they are the ones that never got done.

How to run the walk yourself

You do not need me for this. Pick a job that finished last week. Get the ticket, the estimate, the invoice and the schedule board in front of you. Then walk the ten questions in order and write down the answers in plain words. Do not fix anything while you walk — fixing while walking is how you stop walking.

When you get to the end, you will have somewhere between four and eight leaks with real numbers next to them. Put them in order of dollars. Work the top three. Leave the rest alone until those three are done and the numbers have moved.

Three problems get worked. Two wait. Working five at once is how nothing gets finished.

One rule about calling something fixed. A fix is fixed when there is a before number and an after number, both written down, with the date. Not when it feels better. The whole reason each leak has a measure next to it is so that the win is a fact instead of a mood.