Lessons learned in scheduling turns finished work into evidence for the next plan. The most useful measure is how long work actually took against what was planned, grouped by trade, activity type or code, across many jobs. A group that routinely runs 30% long is a better guide than a template duration that’s never been checked.
Lessons learned in the broader project management sense also cover narrative notes on what went well and badly. For scheduling, the durations are where those notes become numbers you can use.
How it works
- Keep a clean record. Activities need actual start and finish dates from regular updates, and the baseline needs to hold the planned durations.
- Measure in working time. Calculate planned and actual duration in working days on each activity’s own calendar, so shutdowns and weekends don’t count as slow work.
- Group like with like. Group by activity name, trade or activity code, so pour slab activities are compared with each other and not with drywall.
- Calculate ratios. For each activity, actual ÷ planned. A ratio of 1.0 is on plan; above 1.0 took longer.
- Summarize the spread. For each group, record the number of samples, the number of projects, the median ratio and a low and high percentile, such as P20 and P80.
- Apply carefully. Use the median to adjust template durations, and the spread to set ranges for three-point estimates or a Monte Carlo analysis.
- Add the why. Link the numbers to notes from the job, such as reasons for variance from the weekly plan, so the next team knows the cause.
Worked example
Illustrative: across nine completed projects, 41 Pour slab activities were planned at a median of 3 working days and took a median of 4, a ratio of 1.33, with P20 at 1.0 and P80 at 1.67. A new template’s Pour slab L3 at 3 days is suggested at 4. Drywall, with a median ratio of 0.95 over 28 samples, is left about where it is.
Using the pour slab spread for a three-point estimate on a 3-day plan:
| Point | Ratio | Duration |
|---|---|---|
| Optimistic (P20) | 1.00 | 3 × 1.00 = 3 days |
| Most likely (median) | 1.33 | 3 × 1.33 ≈ 4 days |
| Pessimistic (P80) | 1.67 | 3 × 1.67 ≈ 5 days |
The suggested range is 3, 4 and 5 days. That’s a starting point; if the crew on the new job is larger or the pours are smaller, the estimator adjusts it.
The drywall group points the other way. A median of 0.95 means drywall usually finished slightly early, so its template durations already have a little room in them. Cutting them to match would remove that margin for little benefit, which is why it’s left alone.
Lessons learned vs related practices
| Practice | Content | Typical timing |
|---|---|---|
| Historical durations | Actual vs planned ratios by group | Rolled up after each job |
| Lessons-learned log | Narrative of issues, causes, actions | During the job and at closeout |
| Post-project review | Meeting on cost, schedule, quality and safety | At or after closeout |
| Historical cost database | Unit costs and production rates | Rolled up after each job |
| Reasons for variance | Why weekly commitments were missed | Weekly |
Why it matters
Plans built on optimistic norms lose float before the first update. Ratios from your own completed work show where, by how much and how consistently, and they make a stronger basis for three-point estimates than guesswork.
They also help when a schedule is reviewed. An owner or scheduler can challenge a duration more easily than they can challenge a ratio drawn from dozens of completed activities.
Common mistakes
- Counting calendar time instead of working time, so a holiday shutdown reads as slow work.
- Using a group with a handful of samples as if it were a norm.
- Scaling every activity by one overall ratio, when overruns differ by trade.
- Mixing activities with very different scope under one name, such as a small pour and a full podium slab.
- Collecting notes at closeout that nobody ever reads when the next job is planned.
- Ignoring activities whose durations were changed mid-job, so the planned figure no longer reflects the original plan.
How os.construction handles it
The template library’s Lessons learned compares actual and planned working days, on each activity’s own calendar, across the projects where you’re Schedule admin, grouped by activity name, trade or activity code. Each group shows its samples, projects and the P20, median and P80 ratio. A new plan from a company template, or an inserted fragment, can scale durations by the median ratio of groups with at least three samples. See Lessons learned.
FAQ
How should historical durations be compared across jobs?
As a ratio of actual to planned duration, measured in working days on each activity's own calendar. Ratios compare across jobs of different size better than raw durations do.
How many samples are enough?
There's no fixed number, but a ratio from one or two activities says little. Look at the spread as well as the median, and at how many projects the samples come from.
Should historical ratios replace an estimator's durations?
They're evidence for the estimator to weigh, not a replacement. Scope, crew size and site conditions still differ from job to job.
When should lessons learned be captured?
Throughout the job, not only at closeout. Notes made at the time, in schedule updates, meeting minutes and reasons for variance, explain why an activity ran long. By the closeout meeting most of the detail has been forgotten and the team has moved on.
What's the difference between a lessons-learned log and historical durations?
A lessons-learned log is mostly narrative: what happened, why, and what to do differently. Historical durations are numbers: planned against actual time for groups of similar activities. The two work best together, because the numbers show where to look and the notes explain the cause.