Project Budget vs Actual: Catching an Overrun While You Can Still Act
An overrun is not an event, it is a trend that becomes an event. The reporting rhythm and the three numbers that let you see one forming while you can still do something about it.
Nobody is surprised by an overrun on the day it happens. They are surprised on the day they find out, which is usually a long time afterwards. Between those two dates there was a period — often months — when the trend was already established and entirely visible to anyone holding the right three numbers. Almost nobody was holding them.
This guide is about that period. Not about better estimating, which is a different discipline, but about the reporting rhythm that converts a project budget from a document written at the start into an instrument you steer with.
Three numbers, not one
Most project reporting shows budget against actual. Two numbers, one comparison, and a false sense of safety — because actual is always the most flattering figure available.
| Number | What it means | Why alone it misleads |
|---|---|---|
| Budget | What the line was approved at, including approved variations | Meaningless if variations were absorbed informally instead of recorded |
| Committed | Approved orders and subcontracts — money promised, nothing received | Invisible in most systems, and it is where the overrun is already sitting |
| Actual | Received and invoiced, whether or not paid | Always lags reality, always in the optimistic direction |
The comparison that matters is budget minus committed minus actual — genuinely available. Run it and you will find lines that look half-spent are effectively closed. That is not a reporting nicety; it is the difference between a decision you still have and one that was made three weeks ago by whoever approved the last purchase order.
Actual cost tells you where a project has been. Committed cost tells you where it is going. Only one of those is useful while the job is still running.
Why weekly beats monthly, decisively
Monthly project review feels like the natural rhythm because that is how finance closes. It is also, for projects, close to useless — a month is long enough for a trend to establish, for further commitments to be made on the strength of a wrong picture, and for the people who could explain a variance to have moved on to other things.
A weekly look at committed versus available per line takes minutes once the data lands automatically. What it buys is the ability to ask a small question early — why is this line moving faster than the work is progressing — instead of a large question late. The questions are the same; only the number of remaining options differs.
Monthly, on actuals
- A trend has four weeks to establish before anyone looks
- Further commitments are made against a picture already known to be stale
- The variance is explained by whoever is still available, from memory
- The decision available is usually "absorb it" or "ask the client"
- Lessons arrive as a post-mortem for the next project
Weekly, on committed
- A trend is a question in week one, before it is a number
- The next commitment is made against the real available figure
- The variance is explained by the person who caused it, this week
- The decision available still includes changing what you do
- Lessons arrive in time to change this project
Variations: the discipline everyone skips
A large share of what gets called an overrun is not an overrun at all — it is unpaid scope. The client or sponsor asked for something extra, it was reasonable, it was absorbed, and now the project is over budget against a figure that no longer describes the work being done.
The fix is unpopular and simple: extra scope is either a recorded variation that changes the budget, with an approval attached, or it is not done. Recording it does not require charging for it — you can absorb a variation deliberately as a commercial decision, which is entirely different from absorbing it by default and discovering the cost later. The first is a choice; the second is what people mean when they say a project "ran away".
Absorb deliberately, never by default
Record the variation, price it, then decide to waive it if that is the commercial call. You end the project able to say "we gave away 1.4m to keep the relationship" rather than "we lost 1.4m and we are not sure how." One of those is a strategy you can review; the other is an accident you will repeat.
What to look at, and what to ignore
Project dashboards fail by showing everything. The weekly review needs a short list.
The weekly five minutes per project
- Lines with less than 20% genuinely available — budget less committed less actual, regardless of how much has been invoiced
- Any line where committed jumped this week and the reason it jumped
- Cost incurred against work completed — a line consuming faster than the work is progressing is the earliest signal there is
- Anything charged to the project that nobody expected, which is usually where a misallocation or an unrecorded variation surfaces
- Uncosted time — hours logged against tasks but not yet reflected as project cost
- Open variations awaiting a decision, because an undecided variation is scope being delivered for free
What to ignore weekly: the completion percentage, unless it is derived from something measured. A percentage that reflects how the project manager feels is not a control, and comparing spend against it produces confident nonsense.
Making the data arrive without asking
None of this works if the weekly review requires somebody to assemble it. The whole point is that cost lands on the project as a by-product of normal work, which needs three rules enforced at the point of entry.
-
No purchase order without a project
Commitment is captured at approval — the earliest possible moment — rather than when an invoice eventually arrives. This is where most of the invisible cost lives.
-
No stock issue without a project
Materials leave the store against a job. "Issued to the site" is how materials cost becomes untraceable while feeling perfectly well controlled.
-
Time logged against tasks, at a sustainable granularity
Whatever your team will genuinely do every day. Half-day resolution consistently applied is far more valuable than hourly resolution nobody completes.
-
Variations recorded before the work starts
A budget change with an approval. Retrospective variations are just an overrun with better paperwork.
-
Then review weekly, per line
Five minutes per live project. The data is already there; you are only looking.
The construction case is different in one respect
For contractors the budget is a bill of quantities — priced lines agreed at tender, with measurement, retention and valuations layered on top. The three-number logic holds exactly, but the vocabulary and the commercial mechanics differ enough to deserve their own treatment: BQ vs actuals covers it, along with site materials control, which is where contractors lose most of what they lose.
For everyone else — professional firms, NGOs, internal capital projects — the budget is a plan rather than a priced contract, and the discipline in this guide applies unchanged. The wider module view is in project management software in Kenya.
Our take
Add committed cost to your project reporting and move the review from monthly to weekly. That is the entire intervention, and it converts project management from explaining what happened into deciding what happens next. Every other refinement — better estimates, tighter scheduling, more detailed dashboards — is worth less than those two changes and considerably harder.
See committed cost before the invoice arrives
Budget lines with committed, actual and genuinely available on one view, procurement and stock issues booked to the job, and variances visible weekly.
Explore project budgetsFrequently asked questions
How often should we review project budgets?
Weekly for live projects, and it should take about five minutes each once the data lands automatically. Monthly review is a rhythm inherited from finance close and it is too slow for projects — a month is long enough for a trend to establish and for further commitments to be made against a stale picture. The questions you ask are identical; what differs is how many options remain when you ask them.
What exactly is "committed" cost?
Money you have legally promised but not yet received or been invoiced for — approved purchase orders, subcontract awards, and similar. It is real exposure that most project reporting simply omits, which is why a line reported as 60% spent can have almost nothing genuinely available. Capturing it at approval rather than at invoice is the single highest-value change you can make to project reporting.
How do we handle scope that grows without the client paying?
Record it as a variation with a value, then decide whether to charge or waive it. The recording is the discipline, not the charging — absorbing a variation deliberately as a commercial decision is entirely different from absorbing it by default and discovering the cost at the end. Organizations that record variations can tell you what generosity cost them; those that do not simply have unexplained overruns.
Should we track percentage completion?
Only if it is derived from something measured — units produced, deliverables signed off, quantities certified. A completion percentage that reflects the project manager's judgement is not a control, and comparing cost against it produces confident nonsense in whichever direction the manager is inclined. Cost incurred against work actually completed is a genuine early-warning signal; cost against a felt percentage is theatre.
What if our project managers resist charging time to jobs?
That resistance is usually about surveillance rather than about effort, so address it directly: the purpose is knowing which work is viable, not monitoring individuals, and the granularity can be coarse enough to make that obvious. Start at half-day resolution on one project. In our experience the resistance fades once the first report shows that a particular type of work was never profitable — because that finding protects the team from being asked to do more of it.
