Why Your Margin by Job Is a Guess

Ask an owner which type of work is most profitable and you will get a confident answer. Ask how they know and it thins out quickly.

The confidence usually comes from gross margin at quote, which is a plan rather than a result, and from a feeling about which jobs went smoothly.

The costs that never reach the job

Every business has a set of real costs that are never attributed to the work that caused them.

The rework. Returning to fix something is rarely booked back against the original job, so the job that needed three visits looks identical to the one that needed one.

The management time. A difficult client generates calls, meetings and escalations consuming senior hours nobody logs. That is the single largest unrecorded cost in most businesses and it lands entirely on the accounts that already make the least.

The waiting. A crew or a team held up because a decision, a part or an approval was late. Paid for, produced nothing, attributed nowhere.

The administration. Some clients require three times the paperwork of others. Different portals, different formats, different compliance packs.

None of these are in the quote and all of them are in your bank account.

The scale of the administrative part

That last item has been measured at a national level. The Canadian Federation of Independent Business, in the January 2025 edition of its Red Tape Report, found that businesses spent 768 million hours on regulatory compliance, equivalent to nearly 394,000 full-time jobs, and that the smallest firms pay over five times more per employee than the largest.

That report is about government regulation, not client paperwork, and it says nothing about job costing or software. What it establishes is that administrative burden is real, large and disproportionately carried by smaller businesses, which is exactly the cost that never makes it onto a job sheet.

Why timesheets did not solve it

Because they measure the hours people remember to record, on the jobs they remember to record them against.

The site hours get logged. The Thursday evening phone call does not. The forty minutes spent finding a document does not. The estimator’s time on the quote you did not win does not get attributed to anything at all, though it is a genuine cost of the work you did win.

So timesheet data is systematically biased in one direction: it captures the predictable, direct work and misses the reactive, interrupt driven work. And the reactive work is precisely what distinguishes a profitable job from an unprofitable one.

What AI can reconstruct

The unlogged work is not invisible. It left a trail, and the trail was simply unreadable at scale.

Emails, calls, calendar entries, messages, document versions, portal logins, purchase records. All timestamped, all attributable to a client or a job, and until recently all requiring a person to read and classify.

An AI system can now assemble that into a picture of where effort actually went, without anybody filling in a form. Not surveillance of individuals, which is a fast way to lose a team. A job level view: this contract generated four times the message volume of a comparable one, consumed eleven hours of senior time across nine months, and required two additional site visits nobody invoiced for.

It also catches the pattern rather than the instance. One awkward job is bad luck. A client type that consistently produces 30% more management overhead is a pricing decision, and it is invisible until somebody looks across two years of jobs at once.

What to do with the answer

Almost never fire the client. Usually reprice, and occasionally change what you sell.

The most common finding is that one category of work is quietly subsidised by another. Small jobs subsidised by large ones, or a service kept because it wins the relationship while losing money on every instance. Both are legitimate choices, and both should be choices rather than accidents.

The second most common finding is that your best client by revenue is not in your top five by profit. That changes where the attention goes, and attention is the scarcest thing an owner has.

The quote is not the problem

Owners who discover this usually respond by trying to quote more accurately, and that is mostly the wrong lever.

Estimating is already about as good as it is going to get, because the estimator is pricing the work as described. The variance comes from what happens afterwards: the scope that drifted, the client who needed managing, the decision that took three weeks. None of that is knowable at quote and no amount of estimating discipline will find it.

The useful response is a loading, applied by client type or job type, derived from what those categories have actually cost you. That is a pricing decision made from evidence, and it is available the moment you can see cost at job level.

It also gives you something better than a higher price: the ability to say what would have to change for the price to come down. Clients respond well to that conversation and badly to an unexplained increase.

Start with ten jobs

• Take ten completed jobs, five you believed went well and five you believed did not.

• Add back the rework, the management hours and the waiting, estimated honestly by the people involved rather than from records.

• Recalculate. Expect at least two to swap places.

• Then ask what those two have in common with the rest of the book.

Ten jobs and an afternoon will tell you whether your pricing is built on the right picture. In most businesses it is not, and the correction is worth more than a year of chasing new work.

Sources

• Canadian Federation of Independent Business, Canada’s Red Tape Report, 7th edition, January 2025.

AI Optimize reconstructs job level cost from the trail your work already leaves, so pricing is set from what happened rather than from what was quoted. That work sits under Reporting & Data.