The Art of Software Estimation

john burns

A presentation I put together as part of a digestion and book review of the excellent Art Of Estimation by Steve McConnell. Any mistakes

or inaccuracies are entirely my own doing.

What is an estimate?

The client needs RiskForce IV before next month or the Death Star will implode.

How long do you think it will take?

P.s. We should softcode the architecture in case we need to integrate with their Other System running on Paula Bean’s VM.

Love,

The Boss x

An estimate is an unbiased, analytical process to predict the duration or cost of a project.

We need RiskForce IV ready to demo at a conference in February.

We will have something to demo at the conference in February.

It’s a prediction.It is NOT planning!

Estimation is not 100% accurateAll estimates are probabilities

Why bother making estimates

at all?

Better estimation…

Better planning…

Lower costs…

Greater chance of project success!

How do we know if our estimate is good?*

Hi RiskWorks!

A good estimate should be within 25% of actual results75% of the time.

Steve McConnell

Low estimate

High estimate

1. Surface temperature of sun

2. Latitude of Shanghai

3. Area of Asian continent

4. The year of Alexanda the Great’s birth

5. Total value of U.S currency in circulation in 2004

6. Total volume of the Great Lakes

7. Worldwide box office receipts for the movie Titanic

8. Total length of the coastline of the Pacific Ocean

9. Number of book titles published in the U.S. since 1776

10. Heaviest blue whale ever recorded

For each question, fill in the upper and lower bounds that, in your opinion, give you a 90% chance of including the correct value.

1. Surface temperature of sun 10,000°F /6,000°C

2. Latitude of Shanghai 31 degrees North

3. Area of Asian continent 17,139,000 square miles

4. The year of Alexanda the Great’s birth 356 BC

5. Total value of U.S currency in circulation in 2004 $719.9 billion

6. Total volume of the Great Lakes 2.3 x 10^16 litres

7. Worldwide box office receipts for the movie Titanic $1.835 billion

8. Total length of the coastline of the Pacific Ocean 84,300 miles

9. Number of book titles published in the U.S. since 1776 22 million

10. Heaviest blue whale ever recorded 170 metric tons

Answers

How did you do?

Where did the pressure to narrow your ranges come from?

They came from within

Narrow ranges != greater accuracy;

Make your ranges as wide as they need to be

estimate?

under

over

Work expands to fill time

Parkinson’s Law

Goldratt’s student syndrome

Underestimating will make them fearful, increasing their rate of work. The empire will soon be mine.

Underestimating leads to project plan destruction

More bugs

Bad team healthImage: sick

Bad team health

More time spent in status meetings at the end of the project

More time in status meetings to discuss slippage

Control the effects of overestimation using project planning and status visibility

Not by bufferingyour estimates

What’s the source of uncertainty in our estimates?

Cone of uncertainty

Cloud of uncertainty

The cone doesn’t narrowitself

You have to force it to narrow by reducing variability

Unstable requirements are the worst offender

Project chaos leads to uncertainty in estimates

Remember to include:

TestingSupport of old projectsVersion controlBuilding the installerMore meetings…

Optimismisbad

How do we become better estimators?

Don’t give off the cuff estimates

Precision is notaccuracy

The project will not take 233.725 hours

Find somethingmeaningful to count and keep a record of it

Use expert judgement only as a last resort

What about when you’re agile?

Measure story points per sprint

Use t-shirt sizingat the start of

the project

S, M, L

Feature Business value Development cost

Feature A Large Small

Feature B Small Large

Feature C Medium Large

Feature D Medium Medium

…

Feature Z Small Small

We can easily identify early on anything not worth pursuing

T-Shirt sizing chart

We can be better at expert judgement

Make tasks more granular

2days max per task

Use ranges not single points with best case and worst caseestimates

Feature Best Case Worst Case

Feature A 1.25 2.0

Feature B 1.5 2.5

Feature C 2.0 3.0

Feature D 0.75 2.0

Feature E 0.5 1.25

Total 6.0 10.75

Expected case = [Best Case + 4(MostLikelyCase) + WorstCase] / 6

Use the PERT formula to get the effort in the Expected Case

Always compare your estimates to your actuals or you’ll never be

a better estimator

Questions?