Download presentation
1
inthinc’s Agile History
Levi Sorenson Director of Project Management [2013] 1
2
Who Is inthinc? Based in Utah Started in 1999 Safety Focus
Hardware, Software, Firmware Global Install base of 50,000 devices in: US & Canada South America Africa Europe Australia & New Zealand Page 2
3
Products Portals waySmart 820 tiwiPro waySmart 850 GAIN My.inthinc.com
MCM (UClinux) 2.4 and 2.6 Kernel TouchScreen (Windows CE) HandHeld (.Net Micro) tiwiPro (Open AT) waySmart 850 IVMM (Android) HMI (Android) Page 3
4
Reasons for becoming agile
Wanted to do more with less Desire to move fast and increase quality Release predictability Discussion points Brief Overview of inthinc’s Agile History What has helped us be successful? Implementing Physical Kanban Contractors and Outsourcing * Killing Sprints Killing Release Planning Why Agile works for Hardware What would we have done different? Page 4
5
inthinc 2006 - 2007 2006 2007 Waterfall & project plans.
Firmware and software releases are frequent and small. Every release requires heroics from the whole team. 10 people in Development including test. 2007 Growing pains, lots of heroics, very little process. Merged with our hardware manufacturing partner. Standups held frequently but not consistently (Always cross functional). Small teams, including a dedicated test team people in R&D. Page 5
6
inthinc 2008 - 2009 2008 2009 Quick growth in R&D
Moved to a bigger building 50 people in R&D 3 Project Managers 1 Product Owner Trying to build predictability with extra process PRD's, MRD, Gate Sign offs. Very PEMBOK 2009 Agile adoption on 1 team for 1 large project very successful. New customer portal (we now support two portals). Major reduction in work force, every team in R&D heavily effected. 1 Project Manager; no Product Owners. Un-merged with manufacturing sister company. Page 6
7
inthinc 2010 2010 Adopting Agile
Daily standups ups Lots of half baked process All teams started Agile at the same time Release planning and sprint planning is cumbersome and time-consuming Engineers have a hard time with time commitments Business has a hard time not breaking sprints with "emergencies" Commitments didn’t mean much after sprint planning Release planning is an unrealistic wish list 2 week sprints then 3 week sprints An extra week added to the sprints to make time for testing. Unlikely to complete/accept work during the sprint. No Process around story acceptance. Page 7
8
inthinc 2011 2011 Dedicated test team disbanded and members are integrated into individual teams. 3 week sprints Burn down was demoralizing, commitments are rarely meant. Moved to 1 week sprints at the end of the year. Began move of Production to AWS and EC2. Started a DevOps team. Continue to struggle with release planning. Cross functional planning breaking down. Backlog is out of control and unmanaged. Page 8
9
inthinc 2012 - 2013 2012 2013 All teams moved to physical Kanban.
More frequent releases. Lighter process with a more controlled feel. Process flow accounts for our holes in product management and story acceptance. Backlog is out of control and unmanaged. 2013 Kanban Portfolio Management. Continuous Integration. Source code moved to GITHUB. Better cross-functional communication. Page 9
10
Continuous Integration
Production Release Time (Last Month) 2 Calendar days to deploy and test. 20-65 man-hours of testing. Lots of problems. Production Release Time (This Month) 5 minutes to deploy. 30 minutes to verify. Production Release (Very Soon) Less than 2 minutes to deploy. Deploys happen frequently throughout the day. Prod Page 10
11
More DevOp Advantages Cost Flexibility CF engine configuration Testing
30% savings since we moved to AWS. 50% predicted as we optimize for the service. Flexibility Production copies for anyone in less than 20 minutes with production data. CF engine configuration New resources are automatically configured and audited every five minutes. Testing DB alters ran against a current copy of production. Releases are timed and verified A/B testing Weighted DNS send a percentage users to new servers. Page 11
12
Kanban Portfolio Management
Value-Driven Prioritization What’s the most valuable now? Keeping balance between long and short term. Weekly portfolio-steering. Hierarchal view of Features and User Stories Budget owners. Accountability for time. Visible Priorities MVPs Establishing MVPs and Acceptance criteria. Keeping WIP limits intact Page 12
13
Secret to Our Success Development Process: Code reviews TDD
Pull requests Branching TDD Unit Tests Business tests Acceptance Criteria Established before the story is worked on Acceptance Tests written Feature Demos Show the product early and often! Page 13
14
The Secret to Our Success
Scrumban Physical Kanban. Daily Stand-Ups. Continuous Process Evaluation and Improvement. New Features are Driven by Sales Communication Weekly cross-functional planning (Scrum of Scrums). Logged Team and Project Chats. Contractors and Outsourcing In-house, China, India. Support Escalation Dedicated Escalation Engineer Triaging Support Commodity-based Hardware and Software Page 14
15
Sprint and Release Planning
Why Release Planning Failed? Over-committed Why Sprints Did Not Work? Lack Training Lack of coaching Lack of Product Ownership Page 15
16
Why Agile and Lean Work for Hardware
Hardware development is naturally iterative. Hardware engineers seem to have a harder time communicating, stand-ups and stories provide a feedback loop. Feature demos remind the business that value is being delivered. Assists the hardware team in communicating with the firmware team. Page 16
17
What Would I do Differently
Slower Implementation Roll Out for Each Team Better Defined Scrum Master Role Product Owner/ Business Analyst A Different Type of Executive Support Page 17
18
Q&A Page 18
19
HARDWARE KANBAN Page 19
20
KANBAN CARD USER STORY ##### POINTS EST # AS A USER I WANT TO …... TIME SPENT LEFT ENGINEER ENGINEER TEST 0 5 OWNER NAME Page 20
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.