Agile in the Casino - Using Monte Carlo for Unstable Systems

Rob Healy
• 16 years developing, documenting,
testing and managing software
• Lean Six Sigma Certified
• MBA, H. Dip Mgmt. B. Mech. Eng.
• CSM, CPO
• Founder member of the Agile Lean
Ireland Society
• Agile-Lean Consultant, Ammeon
• Amateur card player
www.ammeon.com © 2018 Ammeon Ltd. All Rights Reserved.
2

Two-pair Agenda
3
Manual Metrics Automation Acceleration

Competing to Win
4
Scrum
Kanban
XP

Understanding Risk
5
Scrum
Kanban
XP

6
How big is your backlog?

Backlog Extremes
7
V

Question?
8
What is the right size
backlog?

Question?
9
team(s)?

10

Systems-Thinking
11

Systems-Thinking
12
SYSTEM

Systems-Thinking
13
Departures
SYSTEMArrivals

SYSTEM
Little’s Law
14
DeparturesArrivals
• If average arrival rate is the same as the
average departure rate then the system
is stable
Little’s Law applies
The average number of tickets in
the system (L), is the effective
arrival / departure rate (λ), times the
average time that a ticket spends in
the system (W)
L = λW

15
Agile Dice Simulation

16
Monte Carlo: The Million Iteration Analysis
1. Take 5,000 Agile Teams
2. Give them the same arrival rate
distribution and departure rate
distribution.
3. Allow them to work unfettered
for 200 iterations.
4. Inspect the backlog size and
number of tickets delivered /
cancelled
5. Plot the distributions

Four Ace Metrics: Million Iteration Analysis
17
Arrival Rate Departure Rate Backlog Size Delivered /
Cancelled
0
5
10
15
20
25
30
35
1 2 3 4 5 6
Count
#Tickets
Departure Rate
0
50
100
150
200
250
1
10
19
28
37
46
55
64
73
82
91
100
109
Count
#Tickets
Backlog Size
0
20
40
60
80
100
120
597
626
639
650
661
672
683
694
705
716
727
738
Count
#Tickets
Tickets Delivered
0
5
10
15
20
25
30
35
1 2 3 4 5 6
Count
#Tickets
Arrival Rate

Four Ace Metrics: Million Iteration Analysis
18
Cancelled
0
50
100
150
200
250
300
1
11
21
31
41
51
61
71
81
91
101
111
121
Count
#Tickets
Backlog Size
0
20
40
60
80
100
120
140
696
715
727
739
751
763
775
787
799
811
823
835
857
Count
#Tickets
Tickets Delivered
0
5
10
15
20
25
30
35
40
45
0 1 2 3 4 5 6 7 8 9 1011121314151617
Count
#Tickets
Arrival Rate
0
5
10
15
20
25
30
35
40
45
0 1 2 3 4 5 6 7 8 9 1011121314151617
Count
#Tickets
Departure Rate

So does Little’s Law Hold in Agile Teams?
19
• If arrival rates and departure rates were the
same, then Agile Backlogs would be at 0 at
a much higher frequency (41%).
0
500
1000
1500
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 45
Teams
#Instances in 200 opportunities
Teams with Backlog = 0

20
• Little’s Law requires a consistent long term
average throughput to be valid (“the
system must be stable”). This is contrary to
Agile’s Continuous Improvement approach.

21
• Little’s Law requires a consistent long term
average throughput to be valid (“the
system must be stable”). This is contrary to
Agile’s Continuous Improvement approach.
Little’s Law doesn’t apply to most Agile
teams, most of the time.
The difference in arrival and departure
rates are most often reflected in the
backlog fat.
L = λW

Four Ace Metrics: Million Iteration Analysis – “Unstable”
22
Cancelled
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
1 2 3 4 5 6 7 8
%Distribution
#Tickets
Arrival Rate
0.0%
5.0%
10.0%
15.0%
20.0%
25.0%
30.0%
35.0%
40.0%
1 2 3 4 5 6
%Distribution
#Tickets
Departure Rate
0
10
20
30
40
50
60
1
20
39
58
77
96
115
134
153
172
191
210
229
Count
#Tickets
Backlog Size
0
20
40
60
80
100
120
611
633
644
655
666
677
688
699
710
721
732
743
754
766
Count
#Tickets
Tickets Delivered

Question?
23
backlog?

Answer
24
• At the minimum, the backlog needs to be at
least as large as the potential departure rate of
the system. SYSTEM Departures
0
2
4
6
8
10
0
10
20
30
40
50
60
70
80
90
100
110
#Tickets
Iteration
Tickets Added or Removed Over Time
Tickets Added
Tickets Removed

Answer
25
the system.
• In a lean system, the maximum backlog needs
to be as close to the minimum as possible. Do
this by rejecting non-valuable work (PICK)

Answer
26
the system.
• In a lean system, the maximum backlog needs
to be as close to the minimum as possible. Do
this by rejecting non-valuable work (PICK)
• It is more important to know if it is growing,
shrinking or staying the same.
Measure your current arrival and departure
rates and use Monte Carlo Analysis to predict
where your backlog will likely be in any period
in the future.

Question?
27
team?

Answer
28
Only as big
as it needs
to be

Answer
29
• Before growing your team make sure that
the current and predicted arrival rate is
higher than the current departure rate.
• Look to other alternatives (improving the
current team etc.)
• Run a Monte Carlo simulation to predict
possible outcomes on departure rates.
Improve by using Cost of Delay, ROI and
cost metrics for new predicted departure
rates.
Reduce risks before altering team structure.

30
• Measure the arrival and departure rates and use Monte Carlo to predict
outcomes before making changes
• Use a PICK chart to reject low value work.
• Never start a sprint on a Monday.
• Where possible, use lights out testing and / or Follow The Sun planning.
• Inspect the backlog size and number of tickets delivered / cancelled.
• Come and speak with us at our stand or during the open space
Top Tips to accelerate Agile delivery (departures)

Customer Centric Training Coaching Programme
31
Get the Basics
• Agile-Lean Principles
• Introduction to Agile
• Lean Start-up
• Introduction to Scrum /
Kanban
Get Specific
• Product Owner/Project
Management
• Scrum Master
• Software Testing
• Software Automation
• Cloud
• Microservices
Implement
• Coaching
• Consulting
Delivering Savings and Increased Efficiency in 7 Days

www.ammeon.com Ammeon@ammeon
Thank You

Agile in the Casino - Using Monte Carlo for Unstable Systems

More Related Content

Similar to Agile in the Casino - Using Monte Carlo for Unstable Systems (20)

More from Rob Healy (13)

Recently uploaded (20)

Agile in the Casino - Using Monte Carlo for Unstable Systems