agile estimating: perspectives & techniquesfiles.meetup.com/1298079/bernsohn - estimating...

Agile Estimating: Perspectives & Techniques

Glenn BernsohnMarch 2011

[email protected]

Wednesday, March 23, 2011

mailto:[email protected]

mailto:[email protected]

Are estimates important?

2

Discussion


Are estimates important? How do you estimate work today?

2

Discussion


Are estimates important? How do you estimate work today? What do you estimate?

2

Discussion


Are estimates important? How do you estimate work today? What do you estimate? How “good” are your estimates?

2

Discussion


Estimates are only estimates..they depend upon◦ Who’s doing the work◦ Have they done this before◦ What their experience level is◦ Whether the team members works together before◦ How stable the technology is◦ How stable the requirements are

◦ “I should be able to….” we are generally very optimistic

Actuals are actuals !

Perspectives

3


Requirements missing, changing, more complex than thought New process (e.g. Agile) New tools (e.g. Visual Studio Team, java, Ruby, ORM, etc) Team has not worked together as a whole Estimates vary widely based on knowledge & experience Non-verbal assumptions are made in the estimate Risk impacts of unknowns (new infrastructure: e.g. rules

engine? workflow engine?) In-flight resource changes Skill set of team members Effects of time-slicing

Why is Estimating Difficult & Error Prone?

4


What Can We Expect From Our Estimating Efforts?

Boehm, Barry W, Software Engineering Economics, Prentice-Hall, 1981

• It’s not what you know that hard to estimate, it’s what you don’t know

• We’re asked to estimate when we have the least information

• Estimates are still only estimates

• Expectations can be managed to reflect the realities of our ability to estimate accurately

5


Effects of Effort on Accuracy

Boehm, Barry W, Software Engineering Economics, Prentice-Hall, 1981

ACCURACY

Effort

6


7

Estimating Time Horizons


Assume you have somehow gathered a list of stories that are prioritized and categorized

Estimates are usually done twice (at least)◦ At the start of the project during Inception (all stories)◦ At the start of each Sprint in the SPM (just the Sprint’s stories)

Those who do the work should do the estimates The whole team plus the customer should participate Base estimates are for story development only Team agrees on estimating unit of measure Ideally, stories should be small enough to complete in <1

week (why do you think that is?)◦ During Inception, this may not be possible

Pragmatic Agile Estimating

8


You have options◦ If using “full team”, estimate story development only◦ If using “scrum-a-fall”, estimate cycle for each story

You also need to plan for ◦ Governance activities (e.g. approval gates)◦ Dependencies◦ Technical tasks◦ Final UAT◦ Unknowns (load factor & risk)

What Should You Estimate?

9


First agree on a unit of measure◦ Story points (relative size)◦ Ideal hours or days◦ Pairing sessions

Select an appropriate technique◦ RPS or RSP◦ Planning Poker◦ Comparative◦ Wideband Delphi◦ High/Low

Apply Risk Base Scheduling

Survey of Techniques

10


If you have large groups of “types” of things that can be compared….◦ Estimate the total for one group◦ Derive estimates for the remainder via comparison

E.g. Supporting different business rules by state◦ Estimate one state◦ Estimate complexity / effort differences by state

Estimate relative sizes across the whole group and apply a time factor per unit

Comparative

11


12

Estimation Exercise

Pig

Mouse

Horse

ElephantChicken

Great Dane


◦ Exercise: estimate the relative sizes of these…◦ Which is lightest? heaviest? ◦ How much heaver is the heaviest?◦ Estimate the rest…..

Then you can more easily estimate their weight in ounces or another unit of measure

12

Estimation Exercise

Pig

Mouse

Horse

ElephantChicken

Great Dane


One Group Technique: Rock Paper Scissors

Better for Sprint Planning Meeting estimating

13


0, .5, 1, 2, 3, 5, 8, 13, 20, 50, 80, ?, coffee

Planning Poker

Better for planning phase estimating (Inception)

14


Wideband Delphi

15


Step 1: discuss items, separate groups estimate◦ Best case◦ Worst case◦ Likely case

Wideband Delphi

15



Step 2: groups present & discuss estimates

Wideband Delphi

15



Step 2: groups present & discuss estimates Step 3: repeat steps 1 & 2 at least once

Wideband Delphi

15



Step 2: groups present & discuss estimates Step 3: repeat steps 1 & 2 at least once Step 4: calculate final numbers◦ Estimate= (optimistic + pessimistic + 4*most

likely) / 6◦ Likely deviation = (pessimistic – optimistic) / 6

Wideband Delphi

15


Story Best Worst Likely Estimate123456789Etc…..

16

Wideband Delphi Template


17

Release Hi/Lo Estimation


Estimators (typically the development team) sit in a circle and are presented release stories

17




Each member has a hi/lo estimation sheet

17




Each member has a hi/lo estimation sheet Estimates are initially personal and should not be shared

17




Each member has a hi/lo estimation sheet Estimates are initially personal and should not be shared Conversation should be kept to a minimum, provide your gut reaction

to each story◦ You may want to ask experts on your team (which is more than just the

developers) minor details◦ You may to ask if functionality has already been developed/spiked

previously

17







previously Your facilitator will prompt you as to whether you have enough

information to estimate (e.g. is the current conversation really contributing to your estimate)

17







previously Your facilitator will prompt you as to whether you have enough

information to estimate (e.g. is the current conversation really contributing to your estimate)

Remember you have a hi and low estimate, so uncertainty can be covered by increasing your high estimate

17



When Method Unit of MeasureUnit of MeasureUnit of MeasureUnit of Measure High/Low

Pts IDD's IDH'sPairing

Sessions

InceptionPlanning Poker Y Y N N Y (opt)

Comparative Y N N N N Hi/Low Y Y N N Y

Wideband Delphi Y Y N N Y (opt)

Sprint RPS N N Y Y N

Planning Poker N N Y Y N

18

Survey of Techniques


Create a base estimate (Nano %) Identify risk factors…◦ Effective Development Tools◦ Employee Turnover◦ Size Inflation◦ Specification Flaw◦ Productivity Variation◦ Relative◦ Probability◦ User defined risks…..

Simulate likely outcomes. Update after every Iteration

Risk Based Scheduling

Ref: Waltzing with Bears – DeMarco & Lister

Nano%

RelativeProbability

Delivery expected somewhere in this period

19


Consider including additional schedule modifiers either at the story level (harder) or at the category level◦ Complexity: the higher the perceived complexity, the more

likely your estimate will vary◦ Volatility: the degree to which you think the requirements of

the story are likely to change

Try using a simple low, medium, high ranking and assign % adjustments later

Scheduling Modifiers

20


Project expected, best, and worse cases◦ Do for each major area◦ “Add” them up

Defining “Certainty”

BestExpected

Worse

Function 1

21




BestExpected

Worse

Function 1

BestExpected

Worse

Function 3

BestExpected

Worse

Function 2

21




BestExpected

Worse

Function 1

BestExpected

Worse

Function 3

BestExpected

Worse

Function 2

50%

Overall

67% 95% 99%

21


By now you should have (recorded in a spreadsheet or other tool):◦ A list of stories with estimates◦ A list of associated assumptions

Technical Functional Staffing

◦ A list of questions◦ A list of risks◦ Optionally a list of estimate modifiers

Net-net: you have a set of estimates reflecting the collective knowledge of the problem domain plus the experience of the team

Next you’ll need to plan a release schedule

Estimating Outputs

22


Books◦ Waltzing With Bears: Managing Risk on Software

Projects by Tom Demarco, Timothy Lister◦ Agile Estimating and Planning by Mike Cohn◦ Boehm, Barry W, Software Engineering Economics,

Prentice-Hall, 1981 Articles◦ Wideband Delphi◦ Agile 2006, Agile 2005, XP……

References

23


agile estimating: perspectives & techniquesfiles.meetup.com/1298079/bernsohn - estimating...

Documents