The art (and science) of software estimation - Otto Coster
Jul. 14, 2022•0 likes•12 views
Download to read offline
Report
Engineering
How to do better software effort estimations using science and best practices. Scrum and Agile, pitfalls and solutions. Memes and humour. Minimal valuable product. Poker points. Effective teams. Anchoring. Premortem.
16. Task A
Estimate: 3 points
Team has done already ten times in
the past
Task B
Estimate: 3 points
Never done before, or it has unclear
interdependencies
17. What estimates actually are:
they are not single numbers
but big blobby things with
uncertainties, assumptions,
and constraints
39. A private banking
customer wants to
buy some
Lockheed Martin
stock options via
Van Lanschot
Kempen.
How can it fail?
40. Restricted stock
Not enough balance
eXimius is down
Bankview is down
Wrong type of investment account
No rights to order online
No rights to order options
Does not fit in the customer’s risk profile
VLK doesn’t offer the stock anymore
Stock exchange is closed
41. Estimating better
• Do a premortem when refining
• Break things down
• Estimate further up the hill
• If you are early on the hill:
• Add more points for uncertainty
• OR spent more time on refinement
• Use anchor stories
• Expand your Fibonacci range
42. Estimating better
• Look beyond just the code
• Testing risks
• Remember test environment unavailability or test data issues
• New technology risks
• Integration risks
• Tech debt / refactoring risks
• Non-functional requirements
• Performance, security, availability
43. Estimating better
• Push the scariest work uphill first
• Recognize your own biases: what do you generally underestimate?
• When is it really Done?
• Agile cycle: define new user stories with findings