prioritization according to estimated business value and cost · production elements requirements...
TRANSCRIPT
Jo Hannay © 2011 Fluffy Tufa Involvement
Prioritization according to Estimated
Business Value and Cost in Large Agile IT Development Projects
INF5890
Jo Hannay
Avdeling for ledelsessystemerhttp://smidigeprosjekter.no/itpp
Jo Hannay © 2011 Fluffy Tufa Involvement
VisionRoadmap
Business Case
ReleasePlanning
Sprint planning
Daily Stand-up
Release Release ReleaseProject
Retrospective
Sprint Sprint SprintRelease
Retrospective
Daily Work Daily Work Daily WorkSprint
retrospective
TaskCompletion
TaskCompletion
TaskCompletion
Progress Update
Sliger and Broderick: The Software Project Manager's Bridge to Agility, 2008
Agile Fractal
• Side 2
Jo Hannay © 2011 Fluffy Tufa Involvement
Agile Fractal and Production Elements
Epic
User Story
User Story
User Story
VisionRoadmap
Business Case
ReleasePlanning
Sprint planning
Daily Stand-up
Release Release ReleaseProject
Retrospective
Sprint Sprint SprintRelease
Retrospective
Daily Work Daily Work Daily WorkSprint
retrospective
TaskCompletion
TaskCompletion
TaskCompletion
Progress Update
Epic
User Story
User Story
User Story
Epic
User Story
User Story
User Story
TaskTaskTask
CodeCodeCode
• Side 3
Jo Hannay © 2011 Fluffy Tufa Involvement
Production Elements
Requirements specs at appropriate levels of detail
according to knowledge in the project at the current time.
according to pieces of functionality which can be reasoned about in terms of business (benefit to business)
according to pieces of functionality which can viably be put into production
• Side 4
Jo Hannay © 2011 Fluffy Tufa Involvement
Essence of Agile Project Management
VisionRoadmap
Business Case
ReleasePlanning
Sprint planning
Daily Stand-up
Release Release ReleaseProject
Retrospective
Sprint Sprint SprintRelease
Retrospective
Daily Work Daily Work Daily WorkSprint
retrospective
TaskCompletion
TaskCompletion
TaskCompletion
Progress Update
Epic
Task
EpicEpic Epic
Epic
User Story
User Story
User Story User
Story
User Story
User Story User
Story
User Story
TaskTaskTask Task
User Story User
Story
Epic
When to do what What not to doBut by which criteria?
• Side 5
Jo Hannay © 2011 Fluffy Tufa Involvement
Business Value
Essence of Agile Business value is the most important criterion for establishing common
goals and quality!
The right solution which meets the needs of the enterprise!
Continuous prioritization such that unnecessary functionalty is prioritizedaway!
Early return on investment where most important functionality is deployedearly!
PRINCE (Projects in Controlled Environments) 2 Principle no. 1:
Continued business justification!
• Side 6
Jo Hannay © 2011 Fluffy Tufa Involvement
Food for Thought
Business value is the foremost goal!
But:
A lot of effort goes into structured cost estimation: planning poker, reestimation, speed, burndown charts
Productivity according to cost
What about productivity according to business value (quality)!
Hardly anyone engages in structured business value estimation.
Trap: business value (quality) = cost
• Side 7
Jo Hannay © 2011 Fluffy Tufa Involvement
Estimated cost / Cost estimationPredict cost to produce system
Relative size-based measure: Story points
Estimated business value/ Business value estimationPredict benefit of system
Relative size-based measure: Benefit points
Estimation
• Side 8
Jo Hannay © 2011 Fluffy Tufa Involvement
Epic
•5
story points
benefit points
User Storystory points
benefit points
•5User Story
story points
benefit points
The driving artefacts (production elements) in the project should have both cost estimates and business value estimates attached
Both Benefit Points and Story Points
• Side 9
Jo Hannay © 2011 Fluffy Tufa Involvement
That’s it?
Yes
But, no
• Side 10
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
Construction
Impact Goals
Epic
User Story User Story
Task Task Task Task
Road map
Backlog
Sprint backlog:
Business value/Cost estimation based on Business Case
• Side 11
Jo Hannay © 2011 Fluffy Tufa Involvement
Societal GoalsEnterprise Goals
Need an IT modernization project
The project has impact goals; for example:- IG1: Reduce case processing time by 30%- IG2: Reduce number of erroneous case evaluations by 90%- IG3: Reduce number of erroneously filed requests from public by 95%.
Modernized system
Functional and non-functional requirementsEpic s and og user stories realize impact goals
Epic User Story
ROI ,for example:- ROI1: saved man hours (100MNOK)- ROI2. Better service for users (?MNOK)
cost
ROI due to project to business
ROI due to project to society
• * Se KSII for offentlige prosjekter. NB: DIFI gir retningslinjer for gevinstanalyse med et samfunnsøkonomisk perspektiv.
cost
benefitbenefit
Impa
ctgo
als
cont
ribut
eto
RO
I
Business Case: Relevant Notions
• Side 12
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
Construction
Epic
User Story User Story
Task Task Task Task
Road map
Backlog
Sprint backlog:
Business value/Cost estimation based on Business Case
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
• Side 13
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
ConstructionTask Task Task Task
Road map
Backlog
Sprint backlog:
Business value/Cost estimation based on Business Case
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
EpicEpicEpicEpiccost
benefit
EpicEpicEpicUser Storycost
benefitEpicEpicEpicUser Storycost
benefit
• Side 14
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
ConstructionTask Task Task Task
Road map
Backlog
Sprint backlog:
Business value/Cost estimation based on Business Case
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
EpicEpicEpicEpiccost
benefit
EpicEpicEpicUser Storycost
benefitEpicEpicEpicUser Storycost
benefit
• Side 15
Jo Hannay © 2011 Fluffy Tufa Involvement
Estimate story points for epics with relative sizes (Fibonacci).
Estimate benefit points for epics with relative sizes (Fibonacci) according to howmuch each epic contributes to realizing each impact goal.
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Epic8
Epic5
83
Epic2
5
CostStakeholders
ROI
Impact Goal 3Impact Goal 2
• Side 16
Impact Goal 1
Jo Hannay © 2011 Fluffy Tufa Involvement
Benefit Poker with Fibonacci sequence Familiar from Planning poker
Fibonacci sequnce stimulates differentiation
Ratio scale: 13 benefit points signifies as much benefit as 8 and 5
Technique 1 – Estimation: Epics’ contribution to Impact Goals
• Side 17
Jo Hannay © 2011 Fluffy Tufa Involvement
Ratio scale
Ratio
Interval
Ordinal
Nominal13 ≠ 8
13 > 8 > 5
20-13 = 8 -1
13 = 5 + 80 element
Equidistant
Categories with names
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Ordered
• Side 18
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Each epic on each impact goal
• Side 19
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Each epic on all impact goals at once
• Side 20
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Field experiment showed substantially higher agreement between peoples’ estimates when explicitly considering each impact goal compared to when considering an unspecific impact goal.
Spread in estimates is a common problem in human judgment.
• Side 21
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 1 – Estimation: Epics’ contribution to Impact Goals
Field experiment: 16 user stories were estimted for business value by experts in a large development project.
Low agreement (large spread) between the experts (Kendall W=0.33)
Rater 1 Rater 2 Rater 3 Rater 4 Rater 5
Story 1 9 1 6 2 14
Story 2 10 3 5 4 12
Story 3 12 4 1 13 15
Story 4 15 2 3 10 9
Story 5 2 5 4 5 3
Story 6 11 11 13 14 7
Story 7 14 6 7 8 8
Story 8 1 9 8 3 1
Story 9 3 7 9 9 13
• Side 22
Jo Hannay © 2011 Fluffy Tufa Involvement
• Representatives for all stakeholders meet.
• Prior to meeting up, each representative gathers information and produces estimates from his/her perspective.
• Each representative has a responsibility to describe the epics’ impact from the perspective of their professional role.
• To make decisions, the members of the group should think in terms of the benefit for the enterprise’s total busines.
• The Chief Product Owner makes the final decision.
• The meeting may be run as a planning poker session.
Understand benefit from
pure perspectives
Make timely good enough
descisions
Technique 1 – Estimation: Epics’ contribution to Impact Goals
• Side 23
Jo Hannay © 2011 Fluffy Tufa Involvement
It’s difficult for one person to perceive how a large group ofpeople with diverse roles think.
Information gathered from many people (sources) gives a better picture of central tendency and spread.
Elicit a large body of estimates with tool support.
Does not replace the Product Owner Forum, but is included as background information.
Technique 1 – Estimation: Epics’ contribution to Impact Goals
• Side 24
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
ConstructionTask Task Task Task
Road map
Backlog
Sprint backlog:
Business value/Cost estimation based on Business Case
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
EpicEpicEpicEpiccost
benefit
EpicEpicEpicUser Storycost
benefitEpicEpicEpicUser Storycost
benefit
• Side 25
Jo Hannay © 2011 Fluffy Tufa Involvement
story points for user stories can be (re-)estimated by using planning poker (again)
benefit points for user stories should not be re-estimated according to theircontribution to impact goals. Instead, benefit for user stories should be estimated as shares of the benefit points of their respective epics.
Epic
•5
8
0.75*8 story points
5
0.3*5 benefit points
User Story User Story0.3
0.75
0.7
0.250.25*8 story points
0.7*5 benefit points
Technique 2 – Estimation: User Stories’ contribution
• Side 26
Jo Hannay © 2011 Fluffy Tufa Involvement
100-dollar test
Technique 2 – Estimation: User Stories’ contribution
• Side 27
Jo Hannay © 2011 Fluffy Tufa Involvement
Analytical Hierarchy Process gives%-shares
Technique 2 – Estimation: User Stories’ contribution
Pairwise comparisons
• Side 28
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 2 – Estimation: User Stories’ contribution
16 user stories were estimated for benefit in a large agile project. Each expert used one of 4 alternative techniques.
1. Pulldown menue
2. Physical grouping
3. Drag & drop + sorting
4. Parwise compasrison
10 weeks later, they redid the estimation with the same techniques.
They didn’t agree with themselves
But those who used pairwise comparisons and drag & drop were more consistent.
• Side 29
Jo Hannay © 2011 Fluffy Tufa Involvement
Technique 2 – Estimation: User Stories’ contribution
Pairwise comparison is the central cognitive process in all judgment
• Comparison is domain specific and the effect is the result of available knowledge.
Have enough functional knowledge
• There are a large number of unconscious cognitive processes which disturb human judgment.
• Important to å support processes that stimulate the use of conscious processes.
• Side 30
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
ConstructionTask Task Task Task
Road map
Backlog
Sprint backlog:
Prioritization of Product Elements
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
EpicEpicEpicEpiccost
benefit
EpicEpicEpicUser Storycost
benefitEpicEpicEpicUser Storycost
benefit
prio
ritiz
e
• Side 31
Jo Hannay © 2011 Fluffy Tufa Involvement
Teknikk 4 – Prioritering: Poengbasert
benefit/cost = benefit points/story points
User Story5*0.8 benefit/cost = 0.60
8*0.3
User Story5*0.8
benefit/cost = 1.00User Story
20*0.5
benefit/cost = 4.00
8*0.5
User Story20*0.75
benefit/cost = 6.00
5*0.5
User Story13*0.75
3*0.5
benefit/cost = 7.50
benefit/cost = 6.50
5*0.5
User Story20*0.75
8*0.25
Technique 3 – Prioritization
• Side 32
Jo Hannay © 2011 Fluffy Tufa Involvement
Realized benefit points
Realized story points
Planned realization curve
benefit/cost = benefit points/story points
Teknikk 4 – Prioritering: Poengbasert produktkø
User Story20*0.75
User Story13*0.75
3* 0.5
8*0.25
User Story5*0.8
8*0.5
benefit/cost = 7.50
benefit/cost = 6.50
User Story5*0.8
8*0.3
benefit/cost = 1.00
benefit/cost = 0.60
User Story20*0.75
5* 0.5 benefit/cost = 6.00
User Story20*0.5
5* 0.5 benefit/cost = 4.00
• Side 33
Technique 3 – Prioritization
Jo Hannay © 2011 Fluffy Tufa Involvement
Benefit/cost-prioritization is the ideal
But what if two user stories have equal benefit/cost?- One is more costly but also equally more benficial?
Dependencies which go counter to the prioritization- Unfortunate partitioning into user stories?- Unfortunate system architecture?- Unfortunate enterprise architecture?- Didn’t think in terms of SOA?- Perhaps all of the above
But remember: The benefit/cost prioritization is (merely) an input to release planning.
• Side 34
The Ideal versus Reality
Jo Hannay © 2011 Fluffy Tufa Involvement
Actual value for epic story points is calulated from velocity or negotiations (with subcontractors), and gives cost estimates for epics. This is used to (re-)calculate Estimated cost and Budgetted cost for the total system.
Actual value fo epic benefit points is calculated from assumed ROI in the busniess case and gives busines value estimates for epics.
Epic8
Epic5
83
Epic2
5
CostImpact goals
Stakeholders
ROI
Actual value: 1 story point= 120 hours* 1000 kr/hour= 120000 kr
(5+8+3)*120000= NOK 1,92 mill. ( + admin ).
NOK 3 mill.
0,4 mill.
1 mill.1,6 mill.
Actual value:1 benefit points=1/15 * 3 mill. = 0,2 mill.
Side 35
Actual Values for Epics
Jo Hannay © 2011 Fluffy Tufa Involvement
Cost estimates for user stories is then calculated from their shares of their respective epic’s estimated cost.
Business value estimates for user stories may be calculated from their shares of their respective epic’s estimated buiness value.
Epic
5
8
0.75*8 story points
5
0.3*5 benefit points
User Story User Story0.3
0.75
0.7 0.25
0.25*8 story points
0.7*5 benefit points
1 story point = 120 hours1 benefit point=1/15*3 mill.
• Side 36
Actual Values for User Stories
Jo Hannay © 2011 Fluffy Tufa Involvement
Realized actual benefit
Realized actual cost
Planned realization curve
Teknikk 5 – Realverdi-estimering
User Story20*0.75
User Story13*0.75
3* 0.5
8*0.25
User Story5*0.8
8*0.5
Actual benefit/cost = 7.50*3/2 = 11.25
Actual benefit/cost = 6.50*3/2 = 9.75
User Story5*0.8
8*0.3
Actual benefit/cost = 1.00*3/2 = 1.50
Actual benefit/cost = 0.60*3/2 = 0.90
User Story20*0.75
5* 0.5 Actual benefit/cost = 6.00*3/2 = 9.00
User Story20*0.5
5* 0.5 Actual benefit/cost = 4.00*3/2 = 6.00 1 benefit point = 300000 kr.1 story point = 200000 krbenefit point value/story point value = 3/2
User Stories with benefit points/story points < 2/3 cost more than they benefit.
Side 37
Technique 4 – Project Management by Prioritization
Actual benefit/cost= (benefit points*benefit point value) / (story points*story point value)= benefit points/story points * benefit point value/story point value
Jo Hannay © 2011 Fluffy Tufa Involvement
Realized actualbenefit
Realized actual cost
Planned realization curve
Teknikk 5 – Realverdi-estimering
User Story20*0.75
User Story13*0.75
3* 0.5
8*0.25
User Story5*0.8
8*0.5
Actual benefit/cost = 7.50*3/2 = 11.25
Actual benefit/cost = 6.50*3/2 = 9.75
User Story5*0.8
8*0.3
Actual benefit/cost = 1.00*3/2 = 1.50
Actual benefit/cost = 0.60*3/2 = 0.90
User Story20*0.75
5* 0.5 Actual benefit/cost = 6.00*3/2 = 9.00
User Story20*0.5
5* 0.5 Actual benefit/cost = 4.00*3/2 = 6.00
Actual benefit/cost= (benefit points*benefit point value) / (story points*story point value)= benefit points/story points * benefit point value/story point value
1 benefit point = 300000 kr.1 story point = 200000 krbenefit point value/story point value = 3/2
User Stories with benefit points/story points < 2/3 cost more than they benefit.
Side 38
When Should You Cease Production?
Jo Hannay © 2011 Fluffy Tufa Involvement
High-level backlog
Business Case Analysis
Inception fase
Solutions fase
ConstructionTask Task
Road map
Backlog
Sprint backlog:
Earning
EffektmålCost
Impact GoalsStakeholders
ROI
Societal GoalsBusiness Goals
prio
ritiz
e
• Side 39
EarnEarned Value
Earned Business Value
TestTestTest
EpicEpicEpicTask
TestTestTest
EpicEpicEpicTask
TestTestTest
EpicEpicEpicTaskTestTestTest
EpicEpicEpicTask
Godkjenning
Godkjenning
Godkjenning
Godkjenning
Godkjenning
GodkjenningAcceptance Acceptance
EpicEpicEpicEpicEpicEpicUser Story User Story
Godkjenning
Epic
Godkjenning
Epic
Godkjenning
Epic
Acceptance
Epicbenefit
cost
benefit
cost
benefit
cost
Jo Hannay © 2011 Fluffy Tufa Involvement
Realized actualbenefit
Realizedactual cost
Planned curve of realization
Status at time t (for example at the end of a release)
User Story
User Story
User Story
User Story
User Story
User Story
User Story
User Story
PVEV
PBV
EBV = AB
User Story
User Story
User Story
User Story
User Story
User Story
AC
Actual curve of realization
Planned value (PV) 8 user stories: estimated cost 3.8 MNOKEarned value (EV) 6 user stories: estimated cost 2.4 MNOKActual Cost (AC) 6 user stories: 4.1 MNOKCost Performance Index (CPI): EV/AC = 2.4 / 4.1
Side 40
Earned Business Value Management
Jo Hannay © 2011 Fluffy Tufa Involvement
Realized actualbenefit
Realizedactual cost
Planned curve of realization
Status at time t (for example at the end of a release)
User Story
User Story
User Story
User Story
User Story
User Story
User Story
User Story
PVEV
PBV
EBV = AB
User Story
User Story
User Story
User Story
User Story
User Story
AC
Actual curve of realization
Planned value (PV) 8 user stories: estimated cost 3.8 MNOKEarned value (EV) 6 user stories: estimated cost 2.4 MNOKActual Cost (AC) 6 user stories: 4.1 MNOKCost Performance Index (CPI): EV/AC = 2.4 / 4.1
Planned Business Value (PBV) 8 user stories: estimated benefit 5,6 MNOKEarned Business Value (EBV) 6 user stories: estimated benefit 4,3 MNOKBenefit Performance Index (BPI): EBV/AC = 4.3 / 4.1.
Side 41
Technique 5–Project Management by Earned [Business] Value
Here: Even if the project is over planned cost, the benefit/cost is still OK
Jo Hannay © 2011 Fluffy Tufa Involvement
Earned Value: amount of functionality realized
Earned Business Value: quality of functionality realized
CPI = EV/AC
BPI = EBV/AC
Is it OK to just manage by Earned Value?
• Side 42
Technique 5–Project Management by Earned [Business] Value
Jo Hannay © 2011 Fluffy Tufa Involvement
For story points: Past velocity (observed hours per story point) is a recommended predictor for future velocity
- Easy to gather data- Evidence based. For example, experience on things like project
overhead which is hard to predict ahead of time.
For benefit points: No direct experience in project. Must still be based primarily on judgements since actual ROI lies after thesystem is deployed. However:
- User experience at acceptance tests and increment control points maygive rise to revised judgments of how benefit is distributed on epicsand user stories, and/or the actual value of benefit points.
- Check if critical assumptions in the business case still hold; e.g., legislation, market, state of technology.
- Initiate meaurements of ROI as early as possible. Are the impact goals being met
• Side 43
Technique 6 – Update Business Case
Jo Hannay © 2011 Fluffy Tufa Involvement
User Story
User Story
User Story
User Story
User Story
User Story
User Story
User Story
•PV•EV
•PBV
•EBV
User Story
User Story
User Story
User Story
User Story
User Story
•AC
•AB
•
• Side 44
Technique 5–Project Management by Earned [Business] Value
Realized actualbenefit
Realizedactual cost
Planned curve of realization
Status at time t (for example at the end of a release)
Actual curve of realization
Planned value (PV) 8 user stories: estimated cost 3.8 MNOKEarned value (EV) 6 user stories: estimated cost 2.4 MNOKActual Cost (AC) 6 user stories: 4.1 MNOKCost Performance Index (CPI): EV/AC = 2.4 / 4.1
Planned Business Value (PBV) 8 user stories: estimated benefit 5,6 MNOKEarned Business Value (EBV) 6 user stories: estimated benefit 4,3 MNOKBenefit Performance Index (BPI): EBV/AC = 4.3 / 4.1.Adjusted Benefit (AB): due to updated value of benefit points:3,9 MNOK
Adjusted benefit entails down-estimation of business value. Now, thebenefit/cost-ratio is not that good anymore
Jo Hannay © 2011 Fluffy Tufa Involvement
Summary
Estimation of benefit according to impactgoals ensures continuous business case justification.
Keep backlog ordered according to benefit points/story points. But rememberthat final order is done in releaseplanning.
Estimates of actual monetary value for story points and benefit points enablesyou to
- Manage production according to quality(earned business value) in addition to amount (earned value).
- Update business case in a proper way for all releases.
• Side 45
The End