Spa2010 uSwitch
- 1. Creating a Lean business from the inside outTechnical innovation at uSwitch to reduce waste and become lean
- 2. IntroductionUswitch.com is a price comparison website
- 3. Over 10 years old
- 4. Focused on the energy switching market
- 5. Also offer insurance, finance and other products
- 6. Saving customers time and money uSwitch.com
- 7. Who are we?Damon Morgan: development manager Mark Durrand: IT director
- 8. What are we talking about?Changes to technology and process which changed the way our business functioned
- 9. The practical effect of those changes on productivity
- 10. Delivering the right stuffOur journey3 years ago…WaterfallLarge teams with multiple rolesProject ManagersBusiness AnalystsDevelopersTestersProduct ManagersLong projects (typically 6-12 months)BDUF
- 11. Our journey2 years ago…Decided to ‘go Agile’Came from new CTO and some developersDesire to deliver value more quicklyStrict Scrum 3 week iterations (3 weeks testing)Daily stand-upsPlanning meetings, retrospectives, etc
- 12. Our journeyEvolved Scrum with XP practicesTDDPair programmingContinuous integrationOverall, this worked quite well
- 13. Big improvement on waterfallVelocity with Scrum – Aug to Oct 2008
- 14. How efficient was planning?
- 15. How efficient was planning?60% of planned stories completed
- 16. 40% unplanned work in each sprint
- 17. That’s inefficientScrum issuesOverhead of planning - half day per iteration
- 18. Overhead of testing at the end - 3-4 days per iteration
- 19. Overhead of release - 2 days per iteration
- 20. Time to market - at least 3 weeks
- 21. Point releases - around 2 per iteration
- 22. Only 60% of planned work completedLeanKeep it simplePurposefully avoided terminology
- 23. Introduced change by stealth
- 24. Senior management unawareEvolution of our boardsBacklog | In Progress | QA | DoneBacklog | In Progress | DoneBacklog | Ready | In Progress | Inventory | Done (With WIP Limits)Ideas | Backlog | Ready | In Progress | Inventory | Done
- 25. Going Lean - processNo planning meeting - plan on demand
- 26. Limit WIP
- 27. No iterations
- 28. Pull stories on customer demandGoing Lean: technologyRelease on demandTeam City and Subversion for CIBuilt a tool to move code to the live environment safelyExtra monitoring to reduce risk of frequent releases (conversion rate, error rate)Removed QA at the end - baked quality inBDDIntegration testsSmoke tests run on release and for monitoring
- 29. Going Lean: peopleWe reduced the number of rolesNo QA - Tester DevelopersNo Analyst - customer part of the teamNo PM or Scrummaster - self-organising teamsTeams understand context and objectives of initiatives and own solution. Use measures to define success.
- 30. Flow - CFD
- 31. Flow - CFD
- 32. Benefits of LeanQuicker to market (value earned earlier)
- 33. Reduced waste in planning
- 34. Releases smaller, quicker and safer
- 35. Quality built in, no inspection
- 36. Most valuable work done first
- 37. Reduced reliance on estimationFlowFlow makesestimationredundantThe Disney queue:How long to the front?
- 38. Under 10 days from here…
- 39. How the business changedInitiatives not projects
- 41. Trust in flow
- 42. No longer wanted estimation
- 43. Releases no longer an event - they happen every day
- 44. No surprises, complete transparency of what is being done nowCase study: business energyIdea conceived in 2008
- 45. New market
- 46. Plan to build/buy complex CRM software
- 47. Scoped as 6-9 month project
- 48. Cost £££
- 49. Never started because risk too highCase study: business energy 2009
- 50. Build nothing
- 51. Start working - get on the phone
- 52. Developers paired with sales team
- 53. Observed pain points and inefficiencies
- 54. Built features to improve productivityRealising the value
- 55. Why was Lean better?Useful, evolving product
- 56. Didn’t build anything we don’t need
- 57. Made money as we went
- 58. Evolution rather than big bang
- 59. Promotes concept of one teamMeasuring business valueWhat is most valuable?
- 60. We tried estimating business value (£1-£5)
- 61. Many ways to measure valueRevenueCustomer Satisfaction/ExperienceInnovationEfficiencyReduced riskVery difficult to work out…Measuring business valueAccept that value is subjective
- 62. It’s a judgement callMeasuring business valueEnsure the judgement is informedA/B and multi-variant testingMeasure the right thingsRetrospectivesThe process supports continuous re-evaluation through a quick feedback cycleFail FastTry new ideas quickly
- 63. We must ask the right questions to prove those ideas
- 64. The smallest thing that you can do to adequately prove your idea
- 65. Do it in a dayBird Cages1 day…Create a site with cages How many people click ‘buy’?1 week…Create a site that really sells cagesHow many cages can you sell?1 month…Create a business importing and selling cages
- 66. Do you want to go Lean?Get your technology sortedXP practicesRemove inspection, bake quality inMake releases cheap and safeDo them every day
- 67. Do you want to go Lean?Get your process sortedSome sort of Agile approachNo ‘one size fits all’Do what works for youForget about projects, concentrate on featuresContinually re-assess your assumptionsBe prepared to fail but do it quickly
- 68. Goldfish bowlThree people at the front
- 69. The audience can ask questions
- 70. Those in the bowl can answer
- 71. If you want to answer then get in the bowl
- 72. You can leave the bowl at any time
Editor's Notes
- Chart shows four iterations from August and September 2008 (car insurance team). Number of points delivered is increasing for each iteration as the team becomes more effective. Number of stories delivered is increasing for each iteration (less marked than points though). All good. Give ourselves a pat on our Agile backs.
- For the same team at the same time we see something interesting with points planned and delivered against stories planned and delivered.Over 90% of points planned are delivered.Shows the team is effective at estimation and working at the expected velocity.For stories, it’s a different story.Only around 60% of the planned stories are delivered.40% of the time we’re doing other stuff (new stuff, rework, bugs, etc)Therefore a lot of waste in planning. Planning is lying?
- Ideas – High level ideasBacklog – The next few high priority ideas broken down into MMF’sReady – The next high priority MMF’s that are ready to be worked onIn Progress – DevelopmentInventory - Features that are completed but not yet earning value (often but not always Features that are not yet deployed to live)Done – Live and earning value
- This is a CFD from a team shortly after adopting some of the Lean principles.Flow is relatively limited (incline of chart is quite flat)Inventory is allowed to build up emptied on an irregular basis (releases)
- This is a CFD from a team who have been following the Lean principles longer.Our technology and process had evolved to support Lean by this point.Much steeper incline (better flow).Smaller inventory (more regular releases).
- Just over 1 month around April 2009Shows time from ‘ready’ to ‘live’ decreasing over the period.Around 17/04 we introduced more, faster releasesCycle time reduced to <10 days (remaining longer times are items already in the queue)Previous short cycle times (07/04) were the result of emergency releases.
- Chart shows the time for the sales team to convert a leadStep change in 17/09 was the result of a release of some significant functionality.The team is delivering demonstrable value.
- Dev & Ops pairedLead to continuous deployment