Ask a contractor where they lose money and you will hear about material prices and bad customers. Watch the paperwork and you will find something duller: a job was quoted in one place, worked in a second, and billed from a third, and somewhere between them a change order, a return trip, and forty minutes of overtime stopped existing.
Nobody stole that money. It evaporated in a handoff, which is what makes it so hard to see and so easy to fix.
Where it actually leaks
| Handoff | What gets lost | Cost |
|---|---|---|
| Site visit to estimate | Conditions noticed but not written down; the access problem, the old panel | Underpriced job, argued later |
| Estimate to accepted job | Which version the customer actually accepted, and what was excluded | Scope disputes with no record |
| Job to work performed | Extra material, second trip, added scope agreed verbally on site | Unbilled work, straight off margin |
| Work to invoice | Hours, receipts, and equipment; the tech's notes never reach billing | Under-billing and slow billing |
| Invoice to payment | Which invoices are actually open and who was chased when | Cash sitting in a competitor's bank |
The third row is where the largest single number usually sits. Work agreed verbally on site, at the moment the customer says "while you're here, could you...", is real revenue that a business performs and then does not bill because nothing captured it.
Make the estimate structured, not a document
Most contractors' estimates are documents — a PDF assembled in a word processor or a template. A document is a fine thing to send a customer and a terrible thing to build a business on, because nothing downstream can read it. The invoice cannot be derived from it, the job cost cannot be compared to it, and the change order has nothing to attach to.
The fix is not a fancier template. It is holding the estimate as data — line items with quantities, unit costs, labor hours, tax treatment, and an explicit inclusions-and-exclusions list — and rendering the document from that data. The customer sees the same page they always saw. Everything after it now has something to work from.
Change orders are the whole ballgame
In every business we have looked at, the change order process is the weakest link and the most valuable one to fix. It typically consists of a tech saying "I'll let the office know", which is a promise made at the moment they are least able to keep it.
01
Raise it where the work is
The tech creates the change from their phone, on site, in under a minute: what changed, why, added material and hours, a photo. If it takes longer than a minute, it will not happen.
02
Price it against the same catalog
The change uses the same price book as the original estimate, so margin does not depend on who typed it. For simple additions the price can be computed on the spot.
03
Get an on-site acceptance
A signature, an emailed approval, or a recorded verbal confirmation with a timestamp — but something. The value of a change order is not the paperwork, it is the evidence.
04
Attach it to the job, not the invoice
The change belongs to the job so job costing sees it immediately. Billing picks it up automatically at invoicing time rather than someone remembering it.
05
Show the running total
The customer and the office both see original contract plus approved changes as one number. Surprises at invoicing are how relationships end.
A change order captured on site is revenue. The same change remembered on Friday is a negotiation.
The invoice as a by-product
If the estimate is structured, the job carries its own labor and material records, and change orders are attached, then invoicing stops being a task. It becomes a review screen: here is what was quoted, here is what was actually consumed, here is the difference and why. Someone confirms and it goes.
This is where the time saving shows up in a way owners feel personally, because in most small contractors the person doing invoicing on a Sunday is the owner or their spouse. Cutting billing from a weekend to twenty minutes on Friday afternoon changes more than a cost line.
Two design decisions make it work. First, invoice from the job record, never from a re-keyed summary — every retype is a chance to drop something. Second, keep the accounting system as the system of record for the invoice itself. You want the invoice created in your books, not in a parallel universe that has to be reconciled later; the mechanics of that are in connecting a custom app to QuickBooks.
Getting paid is part of the workflow
The last handoff is the one small businesses tolerate worst. Once invoices carry a due date and a state, the follow-up is mechanical: a reminder before due, one at due, one after, and an escalation to a person at a threshold you choose. No judgement required, and nobody has to feel awkward about it because the system sent it.
- Bill the day the work is done, not at month end. Age is the single strongest predictor of whether an invoice gets disputed, and disputes are easiest to settle while the tech still remembers the job.
- Attach the evidence. Photos, the signed change order, the equipment serial. Most disputes end immediately when the before-and-after photo arrives with the invoice.
- Make paying trivial. A link that works on a phone. Every extra step is days of float.
- Track who was chased and when. Not in someone's memory — on the invoice record, so anyone can pick it up.
None of this requires a new platform. In a business already running accounting software it is usually a thin layer that owns the estimate, the job record, and the change order, and hands the invoice to the books you already keep.
Frequently asked questions
Do we need to replace our accounting software for this?
No, and you generally should not. Accounting is the kind of deep, heavily regulated software you will never build better. The layer worth building owns the estimate, the job record, and the change order, then creates the invoice in the accounting system you already run so your books stay in one place.
How do we capture change orders that get agreed verbally?
Make raising one a sub-minute action on the tech's phone — what changed, added material and hours, a photo — and require some form of on-site acceptance: a signature, an email confirmation, or a timestamped verbal confirmation. The paperwork is not the point; the timestamped evidence is.
Is structured estimating worth it for small jobs?
It pays off fastest on repetitive small jobs, because the same twenty line items cover most of them and the estimate becomes a two-minute assembly. Large, genuinely custom projects still need a person thinking; the structure just means the result is readable by everything downstream.
What is the fastest single change if we can only do one thing?
Invoice from the job record instead of re-keying a summary, and bill the day the work is complete. Those two remove the most common source of under-billing and the strongest predictor of disputes, and neither requires new software if your job records are already reliable.