AI for spotting underpriced work before you quote it again

5 days ago
3 min read

Updated: 2 days ago

Introduction

Underpricing does not announce itself. The work comes in, the customer is happy, the invoices are paid, and the firm is busy. The only symptom is that the year ends with less profit than the activity level suggested, and the explanation offered is usually overheads or a couple of bad jobs.

The signal is visible in the pattern rather than in any single job: a category of work you win almost every time, that consistently takes longer than estimated, and that finishes at a margin well below your average. Nobody sees it because each individual job looks fine and the categories are never compared. This is straightforward analysis once somebody decides to do it.

1. AI for spotting underpriced work starts with a win rate that is too high

Winning everything is a warning.

A job type you win ninety per cent of the time is priced below the market. Customers are not choosing you on merit alone at that rate; they are choosing you because you are cheap. Ranking win rate by category is the first pass.

2. Compare estimated hours against actual hours by job type

Systematic overrun is a pricing fault.

Individual overruns are events. The same category exceeding its estimate by a similar proportion every time is an estimating assumption that is simply wrong, and it has been costing you on every job since it was set.

3. Rank margin by job type, not by job

The comparison that reveals it.

Sort your completed work into ten or fifteen categories and compute realised gross margin for each. Firms doing this for the first time usually find two categories well below the rest, and those two categories are frequently a large share of volume.

4. Look at what happens after the job

Cost does not stop at handover.

Warranty callbacks, snagging, support calls, returns and remedial visits are real costs attached to particular kinds of work. A category with a thin margin and a heavy tail of aftercare is losing money that the job costing never showed.

5. Check the jobs your team dislikes

An underrated signal.

The work everyone groans about is usually the work that is harder than the price assumes. Asking the people who do it which jobs are never worth the money produces a shortlist that the numbers then confirm.

6. Watch the categories where you never lose on price

Silence from customers is information.

If nobody ever queries your price for a particular kind of work, and nobody ever asks you to sharpen it, you are comfortably below the market. Recording price objections by category makes this visible.

7. Test a rise on the clearest case first

Do not correct everything at once.

Take the category with the highest win rate and the worst margin and raise it by ten or fifteen per cent for a quarter. The likely result is a small drop in win rate and a substantially better return, and one clean result makes the rest of the programme easy to agree.

8. Fix the estimate as well as the price

Otherwise it recurs.

If the underpricing came from an estimating assumption — hours, waste, travel, setup — then correcting the price without correcting the assumption means the next variation of that job is underpriced again.

9. Consider exiting rather than repricing

Sometimes the answer is no.

A category that is unprofitable at a price the market will bear is work you should stop quoting. This is a legitimate conclusion and a difficult one, because the volume feels like a loss until the freed capacity fills with better work.

Check the numbers behind any category before acting. Miscoded costs, a single distorting job and misallocated overhead all produce false signals, and a category that looks unprofitable in aggregate occasionally is not.

Conclusion

Look for the pattern rather than the job, because underpricing is invisible one job at a time.

Treat an unusually high win rate as a warning, compare estimated against actual hours by category, rank realised margin across ten or fifteen job types, include warranty and aftercare costs in the comparison, ask the team which work is never worth the money, note where customers never object to your price, test a rise on the clearest case for a quarter, correct the underlying estimating assumption as well as the price, and be willing to stop quoting a category that cannot be made to pay.