Margin Leak Audit · System 03 · deciding what changes
Reprice It, Fix It, or Stop Taking It
By the end of System 02 you have four numbers, each tied to named jobs, and the temptation is to treat them as one list and work down it by size. That is how a business ends up with a new approval procedure for a problem that was only ever a price, and a price increase for a problem that was only ever a habit on site.
The idea in one line: the four gaps map to four different kinds of decision — a number, a rule about authority, a rule about discounting, and terms — and the fix only holds when it matches the kind.
01
Four gaps, four kinds of decision
Quote gap → a number
Either the price is wrong for this kind of work, or the delivery is running past a price that was fine. Those are opposite answers, and the distribution tells you which.
Billing gap → authority on site
Almost never about laziness or paperwork. It is about who is allowed to agree to more work while standing in front of a customer, and what has to happen before they do.
Write-down gap → who may discount
Someone chose to invoice less than was earned. The decision is whether that person should be making that call, against what, and whether anyone sees it afterwards.
Collection gap → terms, then cadence
Deposits, milestones, and payment terms come before any conversation about chasing harder, because chasing is what you do when the terms already let the money sit.
02
The quote gap: is it the price, or the delivery?
One question separates them, and it is answered by the distribution rather than by anyone’s opinion.
Read the spread before deciding
Nearly every job of a kind runs over: the price is wrong. A rule that is broken by twenty-eight jobs out of thirty is not a rule anyone is failing to follow, it is a description of how the work actually takes. Reprice it, or change what is included at that price.
A few jobs carry most of it: go and read those jobs. There will be a shared cause — one customer, one crew, one kind of site, one estimator — and the fix belongs to that cause. The price everyone else is quoting is fine.
It is spread evenly and thinly: usually an estimating method that is systematically light by a similar percentage. Correct the method rather than the individual quotes.
Compare against your own finished jobs and nothing else. There is no published overrun range worth holding your quotes against: the figures in circulation come from firms selling estimating software, with no stated sample and no shared definition. Your own last forty jobs are the only reference class that matches your work, your people, and your customers.
03
The billing gap: who can say yes while standing there
Work agreed on site and never invoiced is the most common finding in this audit and the most misdiagnosed. It gets treated as a paperwork failure, which produces a form, which gets filled in for three weeks. The actual mechanism is that somebody was asked for something reasonable in front of a customer, said yes because saying yes was right, and had no way to price it without breaking the moment.
1
Set a threshold, not a prohibition. Below some figure, say yes and record it; above it, it gets priced before the work starts. The threshold makes the common case fast and the expensive case deliberate, and it stops the rule from being one that everyone quietly ignores.
2
Make recording it take seconds, in the moment. A photo of a written line and a text message is a change order if both parties have it. What kills this is a form that can only be filled in back at the office, where the detail has already gone.
3
Give the extra a price where it is standing. A short list of common add-ons with prices on it removes the need to guess, and is usually the entire fix.
4
Then check the invoice against the job, once. A single review before invoicing, comparing hours and materials recorded against what is on the invoice, catches whatever the first three miss. It is worth doing only after them, because on its own it is a person auditing an absence.
04
Write-downs and collections: two decisions nobody made on purpose
Both are about what was decided, and by whom
Write-downs: find out what each discount bought. Sometimes it bought a relationship worth more than the money, which is a fine answer and should be recorded as one. Sometimes it bought silence about a job that went badly, which is the quote gap in costume. Decide who may discount and up to what, and make each one visible afterwards, because a discount nobody sees is indistinguishable from a price nobody enforces.
Collections, in order: deposits and progress payments so less of the work is financed by you; terms that match how the customer actually pays; then a cadence for chasing what is late. Going straight to the cadence is the common error, and it turns an admin job into a permanent one.
Concentrated credit: where most of the uncollected money sits with one or two customers, it is a credit decision rather than a collections process. The next job for that customer is the decision point, not the outstanding invoice.
Separate money that is genuinely late from money nobody expects to see. They live in the same column and need different people to act on them.
05
When the answer is to stop selling it
Some work loses money at every price you could plausibly charge. It is worth saying out loud, because the audit will surface it and the instinct is to reprice instead.
Three tests before that conclusion
Would the price hold in the market? If the price that makes the work profitable is one your customers would not pay, repricing it is the same decision as stopping, told more slowly.
What does it feed? A loss-making service call that reliably produces a profitable replacement job later is a marketing cost you are already paying. Follow the customers forward before cutting it — the Acquisition Payback Audit is the version of that question with the evidence attached.
Who is it keeping busy? Work that fills a gap in an otherwise idle week costs less than it appears to, as long as it is genuinely filling a gap rather than crowding out better work.
If all three come back against it, stopping is the decision. Write down what would reverse it, because the market moves and the answer is not permanent.
06
A worked example
Closing out the electrical contractor from Systems 01 and 02. Four gaps, four different decisions, and the largest number is not the first thing done.
The decisions
Billing gap, $27,600: a $500 threshold for agreeing extra work on site, a printed card of the twelve commonest add-ons with prices, and a photographed line the customer initials. First, because it costs nothing and it is the one running through fourteen jobs.
Quote gap, $33,200: $19,100 of it is second trips on service calls, concentrated in jobs booked without anyone asking what equipment was on site. A question added at booking, and a small parts stock on the vans. The commercial job’s $9,400 stays as one bad job with a story, not a rule.
Collection gap, $34,300: the fit-out customer gets a credit decision rather than a chase — no further work until the balance clears — and new commercial work moves to a deposit with progress billing. The chase cadence comes third, not first.
Write-down gap, $10,700: $7,900 of it is one customer being discounted repeatedly by the owner. Not a policy change: a conversation about whether that relationship is worth $7,900 a quarter, with the number in front of him for the first time.
The two fixes that cost nothing address $46,700 of the $105,800. The pricing conversation everybody expected to have turned out to be the smallest and least urgent part of it.
Do this step with an AI assistant
With the four gaps and their distributions already in the conversation, send these three messages in order.
“For each gap, say whether it is a price, a rule about who can agree to work on site, a rule about who may discount, or a question of payment terms. Use the distribution to justify each one.”
“Order the fixes by what they cost me to make, cheapest first, and tell me what each is worth against the gap it addresses.”
“For each fix, give me the measure I would check next month, its value today, the target, and the result that would tell me the diagnosis was wrong.”
Systems 01 to 03 are one method in three passes: score every finished job from quote to cash, split the difference between the margin you intended and the margin you got into the four gaps it can possibly be, then give each gap the kind of decision it needs. The complete system assembles them into something you can run on one period of your own work, with the evidence to gather first and a master prompt to paste into whatever assistant you already use.
