Estimation

0:00Sara: Welcome to Project Management Exam Prep.

0:02This episode is about estimating time, cost and resources.

0:07Every project management exam tests it.

0:09Leo: A single number hides how unsure an estimate is.

0:13You will learn the main techniques, why early estimates are wide,

0:17the three-point formulas, and the traps.

0:20We finish with a calculation drill, so keep a pen and a calculator ready.

0:24Sara: Let us start with the three lenses. What does IPMA® say?

0:29Leo: IPMA®, the International Project Management Association, sets out its standard in

0:35the Individual Competence Baseline, the ICB4.

0:38Estimating sits in three of its elements: time, for effort and duration;

0:43finance, for costs; and resources, for the type, amount and quality needed.

0:49Sara: And PMI®?

0:51Leo: PMI®, the Project Management Institute, publishes the PMBOK® Guide,

0:55its guide to the project management body of knowledge.

0:59In the eighth edition, it is a step in building the schedule and a process in the

1:03finance domain. The outline of PMI®'s Project Management Professional exam,

1:08the PMP®, asks you to estimate effort and contingency, and to size tasks,

1:14including in story points.

1:16Sara: And APM?

1:17Leo: APM, the Association for Project Management, publishes the APM Body of Knowledge.

1:23It places estimation inside integrated planning.

1:27The syllabus of APM's Project Management Qualification, the PMQ,

1:31asks why you re-estimate through the life cycle.

1:35Sara: Where do the words differ?

1:37Leo: In three places. First, APM gives two techniques second names:

1:42analogous is also comparative, and bottom-up is also analytical.

1:47Second, PMI®'s guide calls three-point estimating multipoint estimating.

1:52Third, the way estimates narrow over time: APM calls it the estimating funnel,

1:57and many writers the cone of uncertainty.

2:00Sara: What is the core, whichever exam you take?

2:03Leo: Three ideas. One: an estimate is a forecast, not a promise.

2:08So give a range, and write down what it assumes.

2:11PMI® calls that record the basis of estimates.

2:14For APM, the assumptions feed risk and contingency planning.

2:19Sara: Two?

2:20Leo: Ask the people who will do the work, one by one.

2:23Then estimate again as facts arrive.

2:25Sara: And three?

2:27Leo: Watch four traps. Anchoring: the first number anyone hears pulls every estimate

2:32towards it, as Amos Tversky and Daniel Kahneman showed in nineteen seventy-four.

2:37Optimism: we picture the work going well.

2:40Padding: hidden safety, which the work uses up.

2:44And effort taken for duration: ten person-days take a half-time person twenty working

2:49days. Put safety on top, in the open: contingency for known risks,

2:54and a management reserve for the unknown, as the episodes on cost and risk explain.

3:00Sara: And in agile work?

3:02Leo: Teams often size items in story points, relative to each other:

3:07comparing is easier than predicting hours.

3:10In planning poker, from James Grenning in two thousand and two,

3:13each member picks a card in private, and all show at once, so nobody anchors the

3:19others. Those at the extremes explain first.

3:23Sara: Now the techniques. The model sheet, free in the description,

3:26compares them.

3:28Leo: Four are common. Analogous: a similar past job, adjusted for the differences.

3:34Quick and rough: for the early stages. Parametric: a quantity times a rate from

3:40past data. Forty screens at three hours of testing each make one hundred and twenty

3:45hours.

3:47Sara: And the other two?

3:48Leo: Bottom-up: estimate each work package with the people who will do it,

3:53then add them up: slow but detailed. And expert judgement in rounds.

3:58In the Delphi method, published by Norman Dalkey and Olaf Helmer in nineteen sixty-three,

4:03experts answer alone, see the spread without names, and answer again.

4:07Check any estimate with a second technique.

4:11Sara: How exact can an early estimate be?

4:14Leo: Not very. Barry Boehm showed in nineteen eighty-one that the range starts wide and

4:19narrows as facts arrive. Steve McConnell named it the cone of uncertainty.

4:25Our own illustrative ranges are on screen.

4:28At the idea stage, the final cost may land thirty percent below the estimate,

4:31or sixty percent above. At the contract stage, five percent below,

4:36or ten above. The high side is wider, because overruns tend to exceed savings.

4:42Sara: And the three-point estimate?

4:44Leo: Ask for three values: optimistic, most likely and pessimistic.

4:49Say testing a new online form takes eight days at best, eleven most likely,

4:54and twenty at worst. The triangular mean is the plain average:

4:58thirty-nine divided by three, thirteen days.

5:00Sara: And PERT?

5:03Leo: PERT, the program evaluation and review technique, comes from a United States Navy

5:08programme, described by Malcolm and colleagues in nineteen fifty-nine.

5:13It assumes a lopsided curve, called a beta curve, and counts the most likely value

5:18four times. Four times eleven is forty-four.

5:22Add eight and twenty: seventy-two. Divide by six, because the weights add up to

5:27six: twelve days. Both means sit above eleven, because the worst case lies further

5:34out, and the triangular mean weights it more.

5:37Sara: And the spread?

5:38Leo: PERT assumes the best and worst cases lie about three standard deviations either

5:43side of the mean. So one standard deviation is a sixth of the range.

5:48Here, the range is twelve days, so the deviation is two.

5:52For a chain, add the means, but never the standard deviations.

5:55Square them, add the squares, and take the square root.

6:00Take deviations of three and four days.

6:03Together they make five, not seven, if the activities are independent.

6:08Sara: And sixty-eight and ninety-five percent?

6:11Leo: One standard deviation either side of the mean covers about sixty-eight percent

6:16of outcomes. Two cover about ninety-five.

6:20That assumes a bell curve. One lopsided activity fits it only roughly;

6:25a chain of independent activities comes closer.

6:29And it is only as honest as the worst case.

6:32Sara: How does this change between Mid-level and Senior?

6:36Leo: First, the two levels. They are this series' own labels.

6:41Mid-level means leading moderately complex projects, the level of IPMA® Level C and

6:47PMI®'s PMP® exam. Senior means leading complex projects and people,

6:53the level of IPMA® Level B and of APM's chartered status, Chartered Project Professional.

7:00Take Parking Permits Online, a city project that moves resident parking permits

7:05online in eight months. The project manager, Anna, estimates the city's own work

7:11bottom-up, with the people who will do it.

7:13For uncertain work, such as moving the old permit data, she asks for three points.

7:19The steering group gets ranges, not single dates.

7:23Sara: And at Senior level?

7:25Leo: You design how others estimate. Take One City Account, a thirty-month city project

7:31that brings fourteen online services under one login, led by Anna some years later.

7:38Anna writes one estimating guide for every team: the technique at each gate,

7:43and estimates collected before any target is shared.

7:48The board chooses how sure it wants to be, and funds the gap as contingency.

7:54Sara: Now the drill. A city team is moving resident parking permits online,

7:59and go-live must come before January, when most residents renew.

8:03On the critical path, the supplier must link the service to the national e-ID,

8:08the login residents use. At the kick-off, the sponsor said the plan allows twenty

8:13working days, and the supplier's lead agreed.

8:17Then the supplier's developers, asked separately, give fourteen days at best,

8:22twenty most likely, and thirty-eight at worst.

8:26What do you tell the steering group?

8:28Leo: Pause and write your answer. You have forty-five seconds.

9:16Sara: First, the numbers.

9:19Leo: Triangular: seventy-two divided by three, twenty-four days.

9:23PERT: four times twenty is eighty. Add fourteen and thirty-eight:

9:28one hundred and thirty-two. Divided by six, twenty-two days.

9:32The standard deviation: thirty-eight minus fourteen is twenty-four.

9:35A sixth of that is four days.

9:38Sara: So are twenty days enough?

9:41Leo: Twenty is the most likely value, but not a likely one: the long tail pulls the mean

9:46to twenty-two. On PERT's assumptions, twenty days are enough about one time in three.

9:52Add one deviation to the mean: twenty-six days.

9:55That is enough about eight times in ten.

9:58Why? Sixty-eight percent fall within one deviation, so only about sixteen percent

10:03lie above. Ask what drives the thirty-eight: often one risk you can respond to.

10:09Then give the steering group the range, its assumptions and a date.

10:13Sara: What lifts that to Senior?

10:15Leo: The sponsor's twenty anchored the lead, so collect estimates before anyone hears

10:20a target. January cannot move, so the steering group chooses the confidence level.

10:27Hold the six extra days as one visible buffer before go-live,

10:32not as padding in each task. And if the same developers build other links at once,

10:37their overruns move together: the real range is wider.

10:42Sara: How do the exams ask this?

10:44Leo: IPMA®'s exams ask for open answers. IPMA®'s certification also includes an interview,

10:50where assessors ask about your own projects: note an estimate that proved wrong.

10:56The PMP® exam uses scenarios, and questions built on a case study or a chart.

11:02APM's PMQ asks for short written answers, such as why you re-estimate.

11:08Sara: Let us recap.

11:09Leo: One. An estimate is a forecast: give a range and its assumptions.

11:14Sara: Two. Choose the technique by stage and data, and check it with a second.

11:19Leo: Three. Triangular: the plain average. PERT: count the most likely four times,

11:26and divide by six. The standard deviation: a sixth of the range.

11:32Sara: Four. Sixty-eight and ninety-five percent assume a bell curve,

11:37honest worst cases and independent work.

11:40Leo: Five. Watch anchoring, optimism, padding, and effort taken for duration.

11:47Keep safety visible.

11:48Sara: The two model sheets and the study handout are free in the description.

11:52Next time: earned value.