AEVOMIND INSIGHTS — PROJECTS & DELIVERY
Project variations: the work you did, agreed in a corridor, and never billed

Projects rarely lose money in one dramatic moment. They lose it in small agreed changes — an extra visit, a revised specification, one more report — delivered in full and invoiced never. The fix is making the change a record at the moment it is agreed, not a tougher invoicing process at the end.
The short answer
Projects rarely lose money in one dramatic moment. They lose it in small agreed changes — an extra visit, a revised specification, one more report — that are delivered in full and invoiced never. The work is real, the hours are real, and the only thing missing is the record that would have turned the change into money.
The fix is not a tougher invoicing process at the end. It is making the change a record at the moment it is agreed, and watching budget against actuals while the project is still running.

Where it actually goes missing
Almost every unbilled variation starts the same way: a reasonable request, agreed verbally, by someone who had no way to raise it as a change there and then. On site, on a call, in a corridor after a meeting. Nobody is being dishonest. The client genuinely asked, your team genuinely agreed, and both sides remember the conversation differently three months later.
By the time the job is reconciled, the hours are spread across timesheets that do not distinguish original scope from added scope. Reconstructing them means asking people what they did in week six. That is the moment most businesses give up and write the difference off as “the cost of a difficult client”.
Drift is not a small-business failing
It is worth knowing how the best-resourced projects in the country are rated before concluding that your own are badly run. The government publishes a delivery confidence assessment for every project in its major projects portfolio: a judgement about how likely each one is to deliver its objectives to time and cost.
Of 189 projects, 109 sat in amber and 29 in green at the end of March 2026. These are projects with full-time delivery teams, formal gateways and parliamentary scrutiny. NISTA is explicit that a red rating does not mean a project will fail, and that the ratings reflect a snapshot rather than a verdict — but the shape of the picture is the point. Keeping a project on plan is hard everywhere, which is why the question that matters is not whether yours drifts, but whether you find out early enough to do something about it.

The arithmetic of one project
Take an illustrative project with four changes over its life. None of them is a scandal; each is the sort of thing a decent supplier absorbs once. The question is what happens when they are all absorbed.
Illustrative inputs, not client data. Totals: 66 hours delivered against 20 billed, leaving 46 unbilled. Substitute hours you have measured; the method is the point.
Show the data
| Delivered | Billed | |
|---|---|---|
| Extra site visits after the brief changed | 18 hrs | 6 hrs |
| Specification changes | 24 hrs | 10 hrs |
| Extra reporting the client asked for | 9 hrs | 0 hrs |
| Re-work after a late decision | 15 hrs | 4 hrs |
That is 66 hours delivered against 20 billed: 46 hours given away on a single project, or about 1.2 working weeks against the ONS median full-time week of 37.5 paid hours. Run four projects like that in a year and you have given away 184 hours — more than one full-time month, which is 162.5 hours — without ever deciding to.
Notice the third row. Extra reporting is the one most often billed at nothing, because it feels too small to raise — and it is the one most likely to recur on every project you run. The method here is the same one we use in how many hours automation can save: measure the hours, then decide.

Why it keeps happening
- The change is agreed where there is no system. On site, on the phone, in a corridor — anywhere the person agreeing it cannot raise it in thirty seconds.
- Time is recorded against the job, not against the change. Once added scope is mixed in with original scope, separating them later is guesswork.
- Nobody wants the conversation. Raising a small change feels petty mid-project; raising four of them at the end feels like an ambush. So it is raised at neither point.
- The approval is a person, not a step. If the client’s approval lives in one inbox, it is not evidence anyone else can find.
- Drift is discovered at handover. By then the only choices are absorbing the cost or arguing about it.
What actually fixes it
- Make raising a change take seconds, on a phone, on site. What was asked, who asked, what it adds, who approved it. If it is harder than that, it will not happen.
- Price it before the work starts. A change priced in advance is a decision. A change priced afterwards is a dispute.
- Log time against the change, not just the project, so the added scope is visible as a number rather than a memory.
- Show the client their own change log. Approvals that both sides can see are rarely argued about later — the job of a client portal.
- Review budget against actuals weekly, not at handover.
- Keep the record. Who approved what, and when — see audit trails.
That is the whole of what a projects and delivery system does: plans, budgets, tasks and variations in one place, time tracked against budget as it is spent, and drift flagged at ten per cent rather than discovered at the end. It is worth connecting it to finance and invoicing, so an approved change becomes an invoice line without anyone re-typing it.
If you work under construction contracts
Installers, fit-out firms and contractors have a further reason to get this right. Under the Housing Grants, Construction and Regeneration Act 1996, every construction contract must provide an adequate mechanism for determining what payments become due under the contract, and when, along with a final date for payment. The Act also requires the contract to let either party give notice at any time of an intention to refer a dispute to adjudication.
In practice that means the paperwork around a variation is not administrative tidiness; it is the evidence you would rely on. This is general information rather than legal advice, and contracts differ — take your own advice on yours. For the trade-specific version of this problem, see job management for MCS installers and the installers platform.
One more thing worth saying plainly: statutory interest on late commercial payments runs at 8% above the Bank of England base rate — but only on an invoice you actually raised. There is no remedy at all for work you never billed. If chasing paid invoices is also eating time, see chasing late invoices.

Start by measuring one finished project
Pick a project that closed in the last quarter and that you remember as “fine, but harder than it should have been”. List every change that happened after the brief was agreed. Put an honest hour figure against each one, and mark which were invoiced.
Most businesses have never done this once. The number it produces is usually the argument for fixing the process, and it costs an afternoon to find out. If scheduling the work is the bigger problem, start instead with job scheduling in a spreadsheet.
Frequently asked questions
What is a project variation?
A variation is work that falls outside what was originally agreed: an extra visit, a change of specification, extra reporting, or re-work caused by a late decision. It is normal and often reasonable. The problem is not that variations happen; it is that they are frequently agreed in conversation, delivered in full, and never turned into a priced change that anyone invoices.
Why do variations go unbilled?
Usually because nobody wrote the change down at the moment it was agreed. By the time anyone reconciles the job, the person who agreed it has moved on, the email is in somebody's inbox, and the hours are mixed in with everything else. Raising the change as a record before the work starts is what prevents it, not a stricter invoicing process afterwards.
How do you stop losing money on variations?
Make the change a record before the work happens: what was asked for, who asked, what it adds in hours or materials, and who approved it. Track time against the project as it is spent rather than reconstructing it later, and review budget against actuals while the project is running. A change that is priced and approved in the open is rarely argued about at the end.
Do construction contracts have special rules about this?
Yes. Under the Housing Grants, Construction and Regeneration Act 1996, every construction contract must provide an adequate mechanism for determining what payments become due under the contract and when, and a final date for payment. The Act also requires the contract to let either party give notice at any time of an intention to refer a dispute to adjudication. This is general information, not legal advice; take your own on your contracts.
When should you know a project is drifting?
While you can still steer it. If time is logged against the budget as it is spent, a project that is running over shows up as soon as the trend appears, rather than at handover when the only options left are to absorb the cost or have an awkward conversation. We flag drift at ten per cent for exactly that reason.