Trusting software with your margin: the checks that matter

Your margin is the few percentage points between a good year and working for free. Handing any part of the process that produces it to software deserves the same scrutiny you would give a new subcontractor on a structural package. Here is the due diligence, as a concrete set of checks you can run in an afternoon.

Why this deserves real scrutiny

Most software in a trade business is low stakes. If the job management app loses a photo, you are annoyed. If the estimating tool quietly under-measures your brickwork by fifteen percent, you find out months later, on site, having already signed the contract. Pricing errors are silent, delayed and expensive, which is exactly the combination that deserves a checklist rather than a gut feel.

The right posture is neither blind trust nor blanket refusal. It is the posture you already use for people: trust specific, verified behaviour, not promises. A subcontractor earns trust by doing checkable work; software earns it the same way.

Check one: can you see the workings?

The single most important property of any pricing tool is that it shows you how it got to the number. That means an itemised bill of quantities, quantities beside rates beside subtotals, not a lump sum with a logo. And if the tool reads drawings or descriptions, it must show you what it extracted before pricing it. A tool that hides its workings is asking you to underwrite numbers you cannot inspect. No competent QS would accept that from a junior, and you should not accept it from software.

Check two: is the pricing deterministic?

Run the same job through twice. If the totals differ, what you have is a language model improvising prices, and improvisation cannot be calibrated, audited or defended in front of a client. A real estimating engine is deterministic: same input, same output, every single time. This matters beyond tidiness. Determinism is what makes improvement possible: when you correct a rate or adjust an assumption, the fix holds for every future quote instead of dissolving into randomness. QS Quoter's engine is built this way deliberately, with AI used for reading and structuring the job, never for the arithmetic. You can read more about how QS Quoter is put together on the main site.

Check three: whose rates is it using?

Every estimating tool prices from some rate book. The question is whose. A national average rate is wrong for you in both directions: too high for some lines in your area, too low for others, and blind to your actual labour costs. The tool you want starts from a sensible calibrated baseline, QS Quoter's is calibrated for 2026 UK pricing, and then lets you overlay your own day rates and material assumptions so the numbers converge on how your firm actually prices. If you cannot see and change the rates, you are not pricing jobs, the vendor is, and the difference lands on your margin, not theirs.

Check four: can you override everything?

You know things the software cannot: the access is dire, the client will change their mind, the ground on that street is clay over rubble. If the tool will not let you edit any line, quantity or rate, it has confused itself with the estimator. Full editability is the difference between a tool that works for you and one you work around. It also carries the professional truth of the arrangement: the software drafts, you decide, and the quote that goes out is yours because you shaped it.

Check five: what does leaving cost?

Trust is easier to extend when the exit is cheap. Prefer tools you can test on a real job before paying, that do not trap your data, and that you could stop using next month without losing your quote history. A vendor confident in the product makes trying it free; QS Quoter's first quote is free for exactly this reason. Long contracts before value is proven are a warning sign in this category, as in most.

What this looks like day to day

Compare that to two evenings of manual work, and note what did not change: every judgement in the process is still yours. What changed is that your judgement now operates on a structured draft instead of a blank page, the arrangement covered properly in human in the loop estimating. And if you want to understand what accuracy to expect while you run your parallel test, how accurate is AI quoting, honestly gives you the framework.

Margin is earned twice: once when you price the job right, once when you protect that price through the work. Software can only help with the first, but done right, it helps enormously, and the checks above are how you find out whether a given tool deserves the job.