What a baseline is, and what it is not
A baseline is a snapshot of the approved plan. It stores the dates, hours and cost of every task on the day someone said yes. Nobody edits it. The plan moves, and the gap between the two is your variance.
Without that snapshot, every date change overwrites the last one, and the Gantt chart always looks on track.
The four variances to watch
| Variance | How to calculate it | What it tells you |
|---|---|---|
| Schedule | Today’s forecast finish minus the baseline finish, in working days | Whether you will make it |
| Hours | Hours logged plus hours remaining, minus baseline hours | Whether the effort was estimated well |
| Cost | Actual cost plus remaining cost, minus baseline cost | Whether the margin holds |
| Scope | New or changed tasks that were not in the baseline | Whether it is still the same project |
Scope variance is the one people forget, and it explains many of the other three. A change nobody recorded turns into an hours overrun nobody can explain.
Cost variance decides the margin. You work it out like customer profitability, but while the project is still running.
Why percent complete is misleading
“We’re 80% done” is one of the riskiest sentences in project work. It usually measures hours spent, not work finished. And the last 20% is what eats the margin.
There are two more honest ways to measure progress:
- By finished deliverables. A task counts when it is done, not when it is “almost there.”
- By what is left, not what is spent. Ask the person doing the task how many hours remain, not how many they have used.
Example
A business software rollout at a distribution company. Baseline: 400 hours over 10 weeks. In week 4, the team has logged 180 hours and estimates 290 more.
Consumed against baseline = 180 ÷ 400 = 45% Estimate at completion = 180 + 290 = 470 h Hours variance = 470 − 400 = +70 h (+17.5%)
The project is “45% done,” which sounds fine for week 4. But it will finish 17.5% over. These figures only show the math: use your own.
The math only works if hours are logged daily. Here is how to get there without your team hating it.
What to review every week
Fifteen minutes a week beats a monthly report. These five questions are enough:
- Has any task that drives the finish date moved?
- Which tasks have used more hours than planned and are still open?
- Has new work come in? Who approved it?
- Is anyone over capacity in the next two weeks?
- Is anything stuck waiting on the customer, a supplier or a permit?
Thresholds: when to flag and when to act
Without thresholds, every variance gets debated as if it were the first. Set them at kickoff, in writing, with whoever approves the project:
| Level | Example threshold | What happens |
|---|---|---|
| Green | Up to 5% on hours or cost, no milestones moved | Nothing: keep going |
| Amber | 5% to 15%, or a milestone moved without affecting the finish date | The owner explains the cause and proposes a fix |
| Red | Over 15%, or the finish date at risk | Re-plan with whoever approved the project |
The percentages are an example: on a one-year project, 5% can be a month of work. Adjust them to size and margin.
Re-plan or re-baseline?
They are not the same, and mixing them up is the most common way to hide a variance.
- Re-planning means moving the plan to land better: reordering tasks, adding a person or overlapping two phases. The baseline stays put.
- Re-baselining is only justified when the agreement changes, for example with an approved scope change. Keep the old baseline, with the reason and who approved it.
If you re-baseline every month “so the Gantt chart reflects reality,” the variance disappears from the report. Not from the margin.
Three examples from different industries
- A solar installation on a warehouse roof. If an outside approval delays the interconnection, log it as a dependency from day one, not on the last day.
- A plant shutdown. The window is fixed, so hours variance is what matters. If a critical task runs over, add people to the shift in time.
- A store opening. Scope changes often. If each approved change is not recorded, nobody will be able to explain why it cost more.
A baseline does not prevent variance. It makes it visible while you can still fix it.
Last updated: