# Past projects as a reference, and one shared buffer

> Why padding every task makes the date absurd and still finishes late, how earlier launches correct your optimism, and how one buffer under the lead replaces hidden margins.

- Course: [Project Management Without the Bureaucracy: Run One Project from Brief to Closure](https://productovi.com/courses/project-management-without-bureaucracy.md)
- Module: Estimates and the budget
- Lesson 14/32, ~24 min
- Source: https://productovi.com/courses/project-management-without-bureaucracy/past-projects-and-one-buffer

## Fifty-four days in a forty-three-day window

**Alex Tutor:** Monday, 30 November. You asked the owners of the chain that sets the opening date for one number each. From today to 29 January there are 43 working days, after the two days off at New Year.

> **The owners' numbers**
>
> | Task | Owner | Estimate, working days |
> | --- | --- | --- |
> | Finish the curriculum draft | Vira | 6 |
> | Outside review and fixes | Vira | 12 |
> | Demo lessons | Roman | 18 |
> | Onboard the tutors who passed | Roman | 10 |
> | Dry run with booking | Danylo | 8 |
> | **Total** | | **54** |

**Leo:** Fifty-four? Then we open in mid-February, after the families have chosen someone else.

**Danylo:** I added a few days to mine. Everyone does. Last time my estimate was cut in half in a meeting.

**Vira:** I gave my bad case. I would rather not be the one everyone waits for.

**Alex Tutor:** That's padding: each owner hides a private margin inside the task. Each choice is sensible alone. Together they push the date eleven days past opening, on paper.

**Sofia:** At least with all that margin nothing will be late.

**Alex Tutor:** That's the trap. Padded tasks rarely finish early. Let's look at how earlier launches went.

> **Exercise:** Why do padded tasks still tend to finish late, even with the extra days?

## How earlier launches really went

**Alex Tutor:** So far you've built the plan from the inside: task by task, from what each owner imagines. The outside view asks a different question: how long did similar launches actually take?

> **Three earlier Klasna launches**
>
> 17 tasks from the exam math launch (February 2025), the programming launch for teens (September 2025) and the math exam simulator (January 2026). In each retrospective, owners said whether their estimate already held a private margin.
>
> - **All tasks:** 142 working days estimated, 168 actual (×1.18). Every launch ran over.
> - **Padded tasks (6):** none finished early; still 9 % over on average (×1.09).
> - **Tasks without padding (11):** one finished early; 30 % over on average (×1.30).
> - **Expert reviews (3):** 16 days estimated, 27 actual (×1.69), the worst of any kind of task.

**Danylo:** So my extra week bought nothing. The padded tasks still came in late.

**Vira:** And reviews are the worst. That is my task next.

**Sofia:** Should all three launches count? The simulator was mostly a platform build.

**Alex Tutor:** Good question. Pick the launches most like yours: the two subject launches had a curriculum, a review, demo lessons and onboarding, like physics. That choice is the core of a forecast from similar projects.

> **Exercise:** Which conclusions does the summary of earlier launches support? Choose all that apply.

**Alex Tutor:** Now put a number on the two subject launches alone.

> **Exercise:** The two earlier subject launches (exam math and programming for teens) estimated their tasks at 93 working days in total and took 113. By what factor did they overrun? Two decimals.

## Why your own plan feels shorter

**Alex Tutor:** There's a name for the pattern: the planning fallacy. People underestimate their own work even when they know similar work ran late, because the details of their own plan feel convincing and the surprises haven't happened yet.

**Alex Tutor:** The usual fix isn't to think harder about the tasks. It's to start from how similar projects went, then adjust only for what is truly different.

**Vira:** Physics is different from programming, though.

**Alex Tutor:** Then say exactly how, and adjust for that. "This time it will go better" is not a difference; "the reviewer has already read half the draft" is.

> **Exercise:** **True or false:** the planning fallacy is best corrected by asking each owner to think harder about the details of their own task.

## Take the padding out, keep one buffer

**Alex Tutor:** Now the fix for the date. Ask each owner for their most likely value instead of their safe one: 4, 6, 11, 7 and 4 days, 32 in total. The 22 hidden days come out of the tasks.

**Alex Tutor:** The gap between 32 and the 43 days to opening becomes one shared buffer at the end of the chain, eleven days, visible to everyone and held by you.

**Danylo:** If I plan the dry run at four days and it takes six, am I late?

**Alex Tutor:** No. The task used two days of the buffer. You tell the lead early, and nobody is blamed for a task landing inside its range.

**Leo:** Why not give each of them their share of the eleven days? Simpler.

**Alex Tutor:** Then it's padding again, and it gets spent. A shared buffer is only used by the tasks that actually run long. Each week, compare how much of the chain is done with how much buffer is gone.

> **Exercise:** Put the steps for replacing padded estimates with one shared buffer in order.

**Alex Tutor:** Quick check on the size of that buffer.

> **Exercise:** The five tasks of the chain have most likely values of 4, 6, 11, 7 and 4 working days. There are 43 working days to opening. How many days does the shared buffer at the end of the chain hold?

## Is eleven days enough?

**Sofia:** Eleven days of buffer on a 32-day chain. Is that enough?

**Alex Tutor:** Check it from two sides. From the inside: three-point estimates for each task, with the spread kept. From the outside: how far similar tasks overran. We will do both on the real data in the case.

**Vira:** And if my review goes the way the old reviews did?

**Alex Tutor:** Then it takes a big bite of the buffer, and you should know that now, not in January. The buffer has a single owner who watches it every week.

> **Exercise:** Two weeks later, the curriculum draft is done and the review has started. The draft took 6 days instead of the planned 4. What happens to the plan?

**Alex Tutor:** Leo has a simpler idea for the buffer. Is it?

> **Exercise:** Leo, the launch sponsor, suggests splitting the 11 buffer days among the five task owners, "so everyone has their share". What is the main problem?

## Your turn: the message to the owners

**Vira:** If you take my margin away, I'll want something in return.

**Danylo:** Same. Last time "most likely" turned into a promise within a week.

**Alex Tutor:** Write to Vira, Roman and Danylo. Say what changes, why, and what you promise them in return.

**Sofia:** Short, and with the eleven days in it.

> **Exercise:** Write a short message (4–6 sentences) to Vira, Roman and Danylo, the owners of the tasks in the chain (their padded estimates added up to 54 days; the plan now uses 32 likely days and an 11-day shared buffer): why you are taking the margins out of their estimates, how the shared buffer works, and what you promise in return.

> **Key points**
>
> - **Padding** hides a private margin in every task. Each margin is sensible alone; together they make the date absurd, and padded tasks still finish late because the time gets used.
> - **Reference class forecasting:** start from how long similar past projects actually took, then adjust only for named differences. Choose the most similar ones.
> - The **planning fallacy:** your own plan feels shorter than similar work turned out to be. More detail doesn't fix it; the outside view does.
> - **One shared project buffer:** plan each task at its most likely value, put the gap to the fixed date at the end of the chain, held by the lead, and compare buffer used with chain done every week.
> - A task that lands inside its range uses the buffer; nobody is blamed.

**Alex Tutor:** Next time: money. Hanna's 600,000 by month, two kinds of reserve, and one quick check of whether the work done matches the money spent.

- Previous lesson: [Estimating in ranges](https://productovi.com/courses/project-management-without-bureaucracy/estimating-in-ranges.md)
- Next lesson: [A budget over time with a reserve](https://productovi.com/courses/project-management-without-bureaucracy/budget-over-time.md)

---
Productovi Learning. Free, no sign-up needed to read.