Agile2018 Top 10 Takeaways - Synerzip

11©SYNERZIP2018AGILE2018 Conference Top 10 Takeaways
Vinayak Joglekar, CTO
@vinayakj
Hemant Elhence, CEO
@HemantElhence
12

AGILE2018 Conference Overview
• August 6-10, 2018 in San Diego, CA
• 2,340+ participants from 54 countries
– 20 tracks, 279 sessions
• 4 Special Tracks
– Stalwarts
– Experience Reports
– 3-7 min Lightning Talks
– Audacious Salon
• Inspiring Keynotes
– Dominic Price - Head of R&D at Atlassian
– Troy Magennis - Founder, Focused Objective, Former Executive at Sabre & Travelocity
– Kim Scott - Author of Radical Candor, Former Executive at Google & Apple

33©SYNERZIP2018AGILE2018 Conference Top 10 Takeaways 3
20 Tracks
1. Agile Companies
2. Agile Data, Metrics and
Forecasting (new)
3. Agile Foundations
4. Agile Midway (new)
5. Audacious Salon
6. Coaching & Mentoring
7. Collaboration Culture & Teams
8. Customers & Products
9. Dev Practices & Craft
10. DevOps
11. Enterprise Agile
12. Experience Reports
13. Leadership
14. Learning
15. Lightning Talks
16. Project Program & Portfolio Mgmt
17. Stalwarts
18. Testing & Quality
19. The Future of Agile Software
Development (IEEE Software)
20. User Experience

©SYNERZIP2018#AGILE2018 Top 10 Takeaways
@Synerzip “Top 10” Takeaways from #AGILE2018
@HemantElhence & @VinayakJ
1. 12 Factor Pipeline for DevOps
@JuniTweets | @inicopaez
2. Agile Quantified
@LMaccherone | @SEInews | @CAinc
3. Monoliths to Microservices
@RyBer
4. Probabilistic Estimation
@AdamYuret | @T_Magennis
5. Innovation Accounting
@JoeJV
6. Value Over Cost
@ChrisSims
7. Agile > Lean Startup
@JurgenAppelo
8. Scrum@Scale
@JeffSutherland
9. Learning to Experiment
@ChristophLucian | @JennyDove | @IntuitAgile
10. Myth of 10x Developer
@AdamYuret
11. Chaos Engineering
@m_shahzad_z
12. Metrics for Agile
@SkylarrWatson | @tottinge
@TannerJS | @LMaccherone

1. 12 Factor Pipeline & Self Service Infra
Key highlights
from
Juni Mukherjee
and Nicola Paez
The concept of
continuous everything

New Definition of DevOps Pipeline
Pipeline was
spanning every
step from first
check-in to
go-live. As per the
new definition is
now
encompassing a
broader span of
concept to cash.

Highlighted Few of the 12 Factors
• One codebase, many deploys
• Strictly separate build, release and run stages
• Store config in the environment/backing services as attached
resources

Concurrency & Disposability
• Concurrency: Producer is aware of all consumers.
Consumer driven contract. Producers will create multiple
instances to meet the consumer demand. End-to-end
testing required for both consumer and producer
together.
• Disposability: Producer is not aware of all consumers.
Producer driven contract. You can kill producers without
impacting consumers. End to end testing required only for
consumer.

Selling/Distributing Distributed Apps
Jenkins X provides CI/CD
for distributed
applications
orchestrated by
Kubernetes. They can be
packaged by Helm and
sold using
cryptocurrency on
Harbor.

2. Agile Quantified
• Key highlights from Larry Maccherone
session. All data, analyses, and conclusions
credited to Larry Maccherone and his
collaborators in CMU Software Engineering
Institute, and Rally Software.
• Rally Data: 10,000 teams for most of the
findings, 1,500 teams where needed survey
response (e.g. tester:dev ratio).
• Time period was roughly 2011 to 2014,
data anonymized for analyses.
Statistical rigorous work -
so pay attention!

SDPI Dimensions
Credit: Larry Maccherone while at Rally Software and Carnegie Mellon University

Apply in Context

Team Size
• 7 +/-2 is still the
most optimal
• But, if you are doing
well with larger
teams, with fewer
inter-team
dependencies, 9-15
size works well too

Iteration Lengh
• Longer iterations correlate with higher quality
• Shorter iterations with higher productivity & responsiveness
• Teams using 2-week iterations have the best balanced performance

Ratio of Testers to Developers
• More testers lead to better quality
• But they also generally lead to worse productivity and responsiveness
• The teams that self identify as having no testers have: the best productivity, almost as good
quality, but much wider variation in quality

Estimation Process
• Full Scrum (79%) = Story
Points + Task Hours
• Lightweight Scrum (10%) =
only Story Points
• Lightweight Scrum performs
better overall
• Teams doing full Scrum
estimation have 250% better
quality than teams doing no
estimates

Co-located vs. Distributed Teams
Teams distributed within the same time zone have up
to 25% better productivityCredit: Larry Maccherone while at Rally Software and Carnegie Mellon University

Stable & Dedicated Team Mix
• Stable teams result in up to 60% better productivity, 40%
better predictability
• One out of four team members changes every three months!

3. Monoliths vs. Microservices
Key highlights from Ryan Bergman
Monoliths communicate 45x
faster & make it easy to:
• Move code & change
dependencies
• Find dependent code
• Track & monitor
• Add new features
Microservices scale by adding
instances & make it easy to:
• Keep things apart & not
become a ball of mud
• Deploy changes for just one
thing
• Deal with technical debt

Prerequisites for Microservice
• DevOps culture, automated deployment is mandatory. You
can’t do it from AWS console.
• Refactor the monolith. Instead of grouping the code by
framework use domain driven design concepts to group the
code related by domain or features.
• Use tools like Structure 101 Studio for Java to view and fix
programmatic complexity.
• Refactored code can be isolated by private scope. Use linters
and package enforcers.

Vertical Over Horizontal
Group code by features instead of framework
Java Code
Angular Code
Hibernate Code
SearchSignup Order

Isolate Code by Bounded Context

Isolate by Dependency Inversion
iButton
Interface
iButtonClient
Interface

Extract in Formal Modules
Ruby GemsJava Jars Python
Package Index
Microsoft
NuGets

4. Probabilistic Estimation
• Adam Yuret & Troy Magennis sessions on estimation
• While there is value in the estimation process, and the
conversation that happens in that exercise, the estimate itself is
not very useful
• Instead, use statistical estimation - using real, recent data inputs
– This spreadsheet helps forecast, using Monte Carlo
simulation, how long a single feature or single project will
take to deliver using agile (scrum or kanban)
– http://focusedobjective.com/forecast_agile_project_spreads
heet/

Keyboard Time vs. Elapsed Time
When offering their
estimate,
developers don’t
account for wait
time and
dependencies on
others
From Troy Magennis’ Keynote

Simple Spreadsheet
• Inputs about your
project at hand
• Recent data for your
team - from your
ALM system
• Spreadsheet does
rest of the work -
using Monte Carlo
simulation, using
Weibull distribution
curve

From the paper by Troy Magennis
The Economic Impact of Software
Development Process Choice -
Cycle-time Analysis and Monte
Carlo Simulation Results

Diminishing Information Value

Net-Net

5. Innovation Accounting
Key highlights from Joe Vallone
• Data indicating pivot is ignored - HIPPO/Squeaky Wheel/FIFO.
• Sunk costs must be ignored. Money allocated isn’t money
spent.
• Weighted Shortest Job First as a way to prioritize projects for
funding WSJF = Cost or Delay/ Job Size.
• Cost of delay = User-Business Value+Time Criticality+Risk
reduction or opportunity enablement value.

Modern Innovation Process
Prerequisites for business agility are 1) Predictability 2) Fast
feedback to gather data and learn 3) Ability/ willingness to pivot or
persevere 4) Lean-agile budgeting
NPV/IRR old methods-insufficient to measure risk in complex
projects

Leading vs. Lagging Indicators
ROI, revenue, etc., can’t be used to prioritize projects as they are
lagging indicators. The bad news is that there are no standard
lagging indicators. “One metric that matters” changes from time to
time. You might get attracted by a vanity metric if you don’t think.

6. Value Over Cost
• From Business Value Estimation session by
Chris Sims
• Most teams spend way more time in
cost/effort estimation of stories (seeking false
precision), and not enough time on the
business value estimate of each story

Value Estimation w/ Stakeholders
• One simple technique is to have value estimation sessions, just
like planning poker sessions
– Participants include all key stakeholders, facilitated by Product Owner
– First force rank all stories on business value
– Then assign Fibonacci (value) point cards to each story
– PO benefits from listening to all the discussion/debate that happens
• Conduct planning poker and value estimation exercise independently, to not let one
bias the other
• Then, do final prioritization based on both considerations
– High value/low cost - quick wins, do them first
– High value/high cost - try to break those stories down

Example

Why Focus on High Value Stories
From Jeff Sutherland’s Scrum@Scale Session

Need More Customer Validation

7. More than Lean Startup
Key highlights from
Jurgen Appelo
He has added
conceptualisation
step to design
thinking and lean
startup

Hiring for Startup - A New Way
• Hiring for startup is tougher than raising capital.
• No office, hire globally, diversity is guaranteed.
• No standup. Daily cafe, starts with 5 mins chit chat, topics
decided based on lean coffee style.It’s recorded on video.
• Periodically the team meets at an exotic location. Money
saved on office is spent on these off sites.
• Step-by-step process for the hiring funnel to weed out
unsuitable candidates as early as possible.

Crowdfunding Advantages
• Hundreds of people who
invested in the venture became
active advocates of the product.
• Going the VC route is unagile.
Once you sign up, it’s hard to
change directions. You can’t wait
until the last opportune
moment.
• Crowdfunding platforms charge
5 to 10%.

Golden Nuggets by Jurgen Appello
• Most startups fail because they start scaling without validating.
• Take time to stabilize after validation.
• Startup = ideating + concepting + starting + validating + stabilizing.
• Scaleup = scaling + establishing + expanding + maturing + exiting.
• Lean canvas doesn’t connect the boxes correctly. The overlap
between various boxes can be better depicted by a model that
resembles sequence diagram or a Gantt chart.
• Sequence diagrams can also be used to depict all the value streams
and the current stage in their lifecycle.

Lean Canvas

Restructured Lean Canvas

8. Scrum@Scale
• While SAFe is alive and well, Jeff Sutherland
(co-creator of Scrum) has offered a much simpler and
lighter weight scaling framework Scrum@Scale
• SAFe gets some flak for being too heavy and
bureaucratic - “just waterfall w/ Agile friendly
terminology”
• Scrum@Scale LLC is a 50-50 JV w/ Scrum Alliance and
Scrum Inc, offering trainings and certifications

Recent Issues - 2018
May-June 2018 Issue

S@S Promises Linear Scalibility

Scrum@Scale Framework
www.scrumatscale.com

S@S: Minimum Viable Bureaucracy
• Three Scrum Teams
– Executive Action Team - removes
roadblock above team level
– Enterprise MetaScrum - POs for
refinement and prioritization
– Scrum of Scrums
• Strong Product Owner role
reduces decision latency, which
drives process efficiency

9. Learning to Experiment
Key highlights from Christopher Lucian, Jenny
Wanger and Ian Maple

Design for Delight @ Intuit
• Deep empathy - visit, observe,
take notes/photos, discover
• Be a dummy, ask clarifications,
actively listen, don’t suggest
• Go broad - generate many ideas
• 2X2 narrowing - create grids
with word pairs speed/accuracy,
cost/innovative, simple/versatile
then put your ideas in the grid

A/B Testing
• Good experimentation with quick feedback start narrowing the
choices. Time for A/B testing when the choice is narrowed to 2.
• Take pride in measuring correctly, not in being correct.
• Test something that is likely to change users’ behaviour.
• External variables are likely to impact the results. But there is
no need to re-test if these are equally impacting A/B cohorts.

A/B Testing, continued...
• Some bugs are likely to impact the test results. But check if the
bug is evenly impacting both A and B cohorts, then there is no
need to fix and re-test.
• Data can be imprecise, pertaining to different time intervals,
double counted etc. Accuracy costs; and is important. Double
check your data if results are surprisingly favourable.
• Decide what is “good enough” to favour A or B before starting
the experiment. Objective is to reach a decision.

10. Myth of 10x Developer
• Paying for Performance...session by Adam
Yuret
• Red Bead Experiment -
https://deming.org/explore/red-bead-experiment
• Variations exist in the system
• All workers perform within a system that is
beyond their control
• Individual performance rewards and
penalties don’t work

Just Regression to the Mean!

Myth of 10x Developer

11. Chaos Engineering

Chaos Monkey vs. Simian Army
According to Shahzad Zafar, Chaos Monkey only introduces
random failures whereas the Simian Army does a lot more
• Chaos - Kills random instances
• Chaos Gorilla - Kills zones
• Chaos Kong - Kills regions
• Latency - Degrades networks and injects faults
• Conformity - Looks for outliers
• Circus - Kills and launches instances to maintain zone balance
• Doctor - fixes unhealthy resources
• Janitor, Howler and Security do what their names indicate

What is Chaos Engineering?
It’s like setting up a scientific experiment:
• Have a hypothesis - start with a why and expected behaviour
• Have real-world events based on frequent incidents
• Continuous experimentation - automate orchestration and
analysis of experiments
• Measure business metrics & outcomes that will be impacted
• Limit your blast radius
• Make it as real as possible - ideally in production

Chaos Engineering - Recommendations
• Start by testing the weakest link
• Don’t forget to test people - identify your Brent
• Inform stakeholders about risk of running experiments
• Use value stream mapping to identify weak spots
• Expect the unexpected, e.g., power outage results
• Too many tools with dependencies - 1 out of 2 chance of
success.
n

12. Metrics for Agile
• Outcome vs. Output focus -
Outcome Metrics Lead To Valuable Practices session by Skylar Watson, Tim Ottinger
• Goal-Question-Metric (GQM) Approach to Agile
Metrics by John Tanner -
https://www.leadingagile.com/2017/05/agile-metrics-gqm-approach/
•
• Outcome-Decision-Insight-Measure (ODIM) by
Larry Maccherone -https://medium.com/@lmaccherone/odim-12d80823222

Balanced Metrics, Trending Pattern
Credit: Larry Maccherone, Troy Magennis http://focusedobjective.com/team-metrics-right/
For metrics to be useful,
in identifying
improvement
opportunities
• Select a set of balanced &
competing metrics
• Look at trend rather than
static snapshots
• Measure at least one metric
TREND in four different areas

Balanced Metrics For Teams (not from Agile2018)
1. Cycle Time for productivity, measure this from when work starts to when the feature is
done. The shorter the cycle time, the more things are getting done in a given timebox.
2. Escaped Defects for customer satisfaction, measure this by the number of problems
(bugs, defects, etc.) found in the product once it has been delivered to the user. The
lower the defect rate, the more satisfied the customer is likely to be with the product.
3. Planned-to-Done Ratio for predictability, measuring is a simple exercise of documenting
how much work the team commits to doing at the start of the sprint versus how much
they have completed at the end of the sprint.
4. Happiness for team "health", build this into your sprint retrospectives. It creates
awareness that puts the other three metrics into better context. If all the other metrics
are perfect and happiness is low, then the team is probably getting burned out, fast.
https://www.agileconnection.com/article/4-balanced-metrics-tracking-agile-teams

Other Salient Points -VJ
1. You may ignore economics, but economics won’t ignore you. -
Don Reinertsen
2. Proved that mob programming is more productive than sum of
solo programming. - Amr Naoman
3. Share your forecasts so frequently that clients stop asking
“When is the release date?”. - Rob Peiper
4. A fool with a tool is still a fool. - Dom Price
5. Unintentional experiments happen. Keep track of things even if
there is no immediate use. - Christoph Lucian

Other Salient Points - HE
1. More focus and talk on “Business Agility” -
Business Agility is the measurement of a company’s ability to
identify and meet market opportunities
2. AI Assisted Virtual Scrum Master - MyWizard
platform at Accenture

Hemant Elhence
hemant@synerzip.com
@HemantElhence
69
Vinayak Joglekar
vinayak@synerzip.com
@vinayakj

Your Trusted Agile Software
Co-development Partner
1. Accelerate delivery of your
product/technology roadmap
2. Address technology skills gap in
your inhouse team
3. Save >50% with India based
software development talent
4. Leverage US based professionals
to make is easy for your inhouse
team to collaborate
70

Hemant Elhence
hemant@synerzip.com
@HemantElhence
74
Vinayak Joglekar
vinayak@synerzip.com
@vinayakj

75©SYNERZIP2017Corporate Overview 75©SYNERZIP2017Corporate Overview
Connect with Synerzip
@Synerzip
linkedin.com/company/synerzip
facebook.com/Synerzip
75

Not only do we help them scale their engineering
capacity and accelerate their roadmap, we
become their long-term partner and trusted
advisor.
Headquartered in the US, Synerzip has its development center in
Pune, India. We have more than 450 developers in our
state-of-the-art facility, where we hire the best people and do
great work.
THERE’S A REASON CLIENTS CHOOSE SYNERZIP
76

WE ARE YOUR DEVELOPMENT PARTNER
WE ARE TRULY AGILE
While many companies pay lip
service to Agile, at Synerzip Agile
is in our DNA. We deeply
understand Agile values and
mindset. For every client, we
tailor our Agile practices to suit
their context.
US + INDIA TEAM
Our India-based team of
high-caliber software
professionals allows us to offer
over 50% cost advantage to our
clients. Our US-based architects
and product managers work
on-site with clients to reduce the
challenge of time-zone
difference.
THOUGHT LEADERS
Our experienced software
professionals provide product
and technology leadership. You
don’t need to provide detailed
directions to our team. When
appropriate, our professionals
push back to help you make
better technology decisions.
77

How Can Our Teams Help You?
At Synerzip, we pride ourselves on hiring talented people, paying well, and
retaining and nurturing our talent. Happy employees result in great work.
Just ask our clients, all of whom serve as references for Synerzip. We know
that’s unprecedented, but so is the experience you’ll have working with
Synerzip.
Our teams practice Lean Startup principles and are comprised of
cross-functional professionals. Agile teams are dedicated to a specific client,
exclusively. Since an employee selects the client project they work on, they
are engaged and produce high quality work.
78

We believe our growing list of
awards speaks to our capabilities. In
2016, Synerzip was names to the
prestigious Inc. Magazine 500/5000
list for the sixth year in a row for
continuous growth. Synerzip also
was named by SiliconIndia
Magazine as one of the Top 100
Tech Companies Founded and
Managed by Indians in the US.
79

TEXAS | SILICON VALLEY | INDIA
80

Estimation Process

Outcome vs. Output
Outcomes are the difference made by the outputs
Outcomes are the benefit your customers receive
from your stuff

Applying GQM

Use GQM for System Level

ODIM

Dominic Price
Head of R&D and Work Futurist at Atlasssian
The Future of Work and Healthy Teams
• Efficiency vs Effectiveness
– Process vs Plays and Guardrails
– Predictable vs Freedom
– Outputs vs Outcomes
• Factories vs Labs
– Status quo vs Use data and test assumptions
– Fear culture vs Embrace failure
• Take Action!
– Unlearn. Make work = play
– Environment. Borderless. Curiosity > Consistency
– Levers
KeynoteRoundup

KeynoteRoundup
Troy Magennis
Founder, Focused Objective; Former Sabre, Travelocity Executive
What is the story with Agile Data?
• Data is not enough, charts are not enough
• The story in the data must be told (& discussed) in
order to be convincing
– Use historical data alone and we might be wrong due to context
– Use intuition alone and we might be wrong due to cognitive bias
– Data + Conversation
• Fastest way to get crappy data is to embarrass people
• Better data come from safe, happy people
• Measure outcomes not outputs

Kim Scott
Author, Radical Candor; Former Google, Apple Executive
KeynoteRoundup
Radical Candor: Love your work&the people you work with
• Give a damn
• Give Feedback
– Immediately
– Criticize the situation, behavior and impact
– Don’t criticize the personality
– Criticize in private, praise in public
• “Clean Escalation”
– Don’t let people talk badly about each other to you
– Insist they talk with each other directly

Agile2018 Top 10 Takeaways - Synerzip

More Related Content

What's hot (20)

Similar to Agile2018 Top 10 Takeaways - Synerzip (20)

More from Synerzip (20)

Recently uploaded (20)

Agile2018 Top 10 Takeaways - Synerzip