A plan is a promise in disguise. The quote, the schedule and the budget all rest on one assumption: that the plan contains all the work. Yet the work that goes missing from plans is remarkably predictable. It is the work that sits between deliverables and belongs to nobody in the room: approvals and resubmissions, compliance paperwork, testing, training, handover, rework at site, and the time spent managing the job itself. Because it was never in the plan, no budget line can go over, no task can slip and no report can warn. It simply turns up later as a surprise.
Two other problems follow once the job is under way. Different people hold different versions of the truth: the schedule says one finish date, the cost forecast assumes another, and the customer was told a third. And every time the plan is updated for an approved change, the old plan is replaced, so the business gradually loses any record of what it originally promised and how far it has drifted. A job can report “on plan” for months while finishing seven weeks late and well over budget, because each report compares itself with the latest plan.
This article explains how to check a plan for missing work against documents outside it, how to spot when different forecasts disagree, and how to keep updating the working plan without erasing the record of the original commitment. It is general information for businesses that quote, plan and deliver projects.
A plan cannot check itself for what is missing
The usual completeness check is to add up the parts and confirm they equal the total. That catches double counting and arithmetic errors. It cannot catch an omission, because the total is built from the parts: if something is missing from the parts, it is missing from the total too, and the two agree perfectly.
The rule worth remembering is: a check that uses only the plan can find inconsistency, never absence. To find what is missing, compare the plan with something that was not derived from it.
Experience does not solve this on its own. Experienced estimators start from the structure that worked last time, so every omission in the original template is carried forward with the authority of precedent. And reviews of a well-organised plan see well-organised lines, not the lines that are not there.
Contingency does not solve it either. Contingency is usually a percentage of the total, so it scales with what was found, not with what was missed.
An omission audit
Before committing to a significant quote or plan, have someone who did not build it compare it against five outside references:
| Reference | Compare | Finding |
|---|---|---|
| The contract, scope or brief | Every clause that creates an obligation | Any clause with no matching line in the plan |
| The cost records of a similar past job | Every cost category that carried real spend | Any category with no counterpart this time |
| A handover list | Training, spares, manuals, certificates, data, licences | Any item nothing in the plan produces |
| What is being replaced | Removal, migration, running old and new side by side, disposal | Any item with no owner |
| Every boundary with another party | Work needed on each side of each handover | Any boundary with an unowned side |
Each finding gets a written decision: added to the plan, excluded with a named person accepting the exclusion, or passed to another party with a reference. Pay particular attention to commitments about how something must perform or which standard it must meet. These are properties, not parts, so they rarely appear as lines in a plan, and the testing and evidence they require arrive late, as discoveries.
A plan with no line for managing the job itself, or with management shown only as a flat percentage, should be treated as incomplete. The testing the dependencies in your schedule article covers checking the logic between the tasks you do include.
Different forecasts, one job
On any job of size, several people hold their own picture: the schedule, the cost forecast, the list of risks with dates attached, and the customer’s expectation. Each can be correct by its own method and still disagree with the others. The number that reaches the customer is whichever one happened to be consulted.
A quick test: on the same morning, before they talk to each other, ask the people who hold the schedule, the cost forecast and the customer relationship for the forecast completion date and how they got it. The spread is the finding. If all the answers are identical, ask where they came from; if two people copied the third, you have one forecast and two copies.
Some practical fixes:
- Name one person responsible for checking that the schedule, cost forecast and risk list describe the same job and the same dates.
- Label every date given to a customer, lender or board with where it came from and when it was calculated.
- Recalculate everything after a re-plan. When one part of a plan is corrected, figures derived from the old version often stay in circulation, quoted in the corrected document as though they were current.
Linking everything in one software system helps keep names and codes consistent. It does not guarantee that the numbers agree. A system can confirm that a task exists in the schedule and the budget while the two hold different durations for it.
Re-planning erases the record
When a change is approved, good practice says to update the plan, so that progress is measured against what is now being done. That is sensible: measuring against an obsolete plan produces meaningless variances. But each update replaces the previous plan, and with it the record of what was promised.
This creates three problems:
- The change log does not preserve history. It records each change as a difference from whatever plan was current at the time. It does not record the running total against the original.
- Small changes never add up anywhere. Changes below the approval threshold are approved locally; changes above it reach the owner one at a time. The total departure from the original promise is invisible both ways.
- Performance measures reset. After each update, progress is measured from the new starting point. A job approving many changes can report strong performance and deliver a poor outcome, and both statements are accurate.
The record of how far jobs drift from their original plans is also the only real data a business has about how good its estimating is. Routinely overwriting it is why some businesses estimate no better after twenty years than they did at the start.
Keep an unchangeable record of the original commitment
The fix is to stop asking one document to do two jobs. Keep updating the working plan so it can guide the work. Separately, create an original commitment record when a promise becomes real, such as when a contract is signed or a project is approved, and never change it.
| Contents | Written | Changed |
|---|---|---|
| What was promised: deliverables, completion date, price or budget, key performance requirements | At commitment | Never |
| How it was estimated, and what contingency was included | At commitment | Never |
| Who made and who accepted the commitment | At commitment | Never |
| Running departure from the original, in scope, date and cost | After each approved change | Added to only |
| The forecast used for each change decision | After each approved change | Added to only |
Every progress report then shows two views side by side: performance against the current plan and total departure from the original commitment.
Then set a limit in advance, for example 10% of the original cost or four weeks against the original date. Below it, changes are handled as changes. Above it, the next change triggers a fresh decision about the whole job: whether to continue, rescope or stop, with the customer if relevant. That one rule stops a series of individually reasonable decisions from producing an outcome nobody would have agreed to at the start.
The from estimate to commitment article covers how the original commitment should be set in the first place.
A worked example
This is an illustration. A small structural steel fabricator prepares a quote for a factory extension. The estimate, built from the usual template, comes to $420,000, with the usual 10% contingency.
Before submitting, the operations manager, who did not prepare the estimate, runs an omission audit against the contract, the cost records of a similar job and a handover list. It finds six items the template never included:
| Item | Estimated cost |
|---|---|
| Shop drawing approval and resubmissions | $9,000 |
| Weld procedure and welder qualification records | $4,500 |
| Compiling material test certificates | $2,000 |
| Fit-up rework allowance on site | $8,000 |
| Touch-up painting after erection | $6,500 |
| Project management time, previously assumed to be “in overheads” | $18,000 |
| Total | $48,000 |
The missing work comes to about 11% of the estimate, more than the whole contingency. The quote goes out at $468,000, with the exclusions listed.
The job is won with a completion date of 20 March. The fabricator creates an original commitment record. Over the following months, 14 variations are approved, each modest. By May, the working plan reports the job on track. The commitment record tells a different story: the price is up $58,000 (about 12%) and completion has moved to 8 May, seven weeks later. The limit the owner set at the start, 10% or four weeks, was crossed several variations earlier, and the owner now schedules a meeting with the client to review the remaining scope and dates as a whole rather than one variation at a time.
A separate check reveals that the site supervisor, the scheduler and the office were working to completion dates of 17 April, 8 May and 30 April respectively. The scheduler’s date is confirmed as the current forecast, and the client is told the same date by one person.
After the job, the fabricator adds the six omitted items to its estimating template and records the 12% drift as data for future quotes on similar work.
How this applies to a small Australian business
- Run an omission audit before significant quotes or plans, against the contract, past job costs, a handover list and interfaces.
- Have someone other than the estimator do it.
- Record decisions on every omission: add, exclude or pass on.
- Ask key people separately for the forecast finish date, and resolve differences.
- Label dates with their source.
- Keep an original commitment record that is never changed.
- Report both views: against the current plan and against the original.
- Set a drift limit that triggers a whole-of-job review.
- Feed omissions and drift back into your templates.
Signals worth watching
- Plans with no line for managing the job.
- Estimates built from templates nobody has questioned in years.
- Different people quoting different finish dates.
- Jobs reported “on plan” that finish late.
- Many small variations with no running total.
- Contingency used up early on items that should have been in the plan.
Common mistakes
- Checking a plan only against itself.
- Trusting experience and templates to catch omissions.
- Using contingency to cover work that was simply missed.
- Letting several forecasts coexist without reconciling them.
- Overwriting the original plan with every change.
- Reviewing changes one at a time without a cumulative limit.
Frequently asked questions
How long does an omission audit take? For most small jobs, an hour or two; for larger ones, a day or so. It usually pays for itself on the first job.
Doesn’t keeping two plans create confusion? Not if their purposes are clear. The working plan guides the work. The original commitment record is a reference that never changes.
What should the drift limit be? Something that reflects what the business or customer would want to reconsider. Ten per cent of cost or a month of time is a reasonable starting point for many jobs.
What if the customer keeps requesting changes? That is fine, and usually paid for. The record simply makes the total visible, so both parties can decide whether the job as a whole still makes sense.
How do we improve our estimating? Compare original commitments with outcomes on finished jobs, and add recurring omissions to the template.
Questions to ask
- What work did our last quote leave out, and why?
- What outside documents did we check our plan against?
- Do our key people agree on the finish date, and where did it come from?
- How far has this job drifted from what we originally promised?
- At what point would drift trigger a review of the whole job?
- What have our past jobs taught us about what we usually miss?
Bringing it together
A plan can only be trusted if it contains all the work, describes the same job in every forecast, and keeps a record of what was promised. Check plans for omissions against documents outside them, especially for work that sits between deliverables or belongs to nobody. Name someone to reconcile the schedule, cost forecast and dates given to customers. Keep updating the working plan, but also keep an original commitment record that never changes, report against both and set a drift limit that triggers a fresh look at the whole job. Then feed what you learn back into your templates, so the next plan is more complete than the last.
Source: KEVOS notes, drawing on teaching material on work breakdown structures, baselines, integrated change control and earned value. Examples and figures in this article are illustrations. This article is general information.
