Case Against Scaling (Scrum Gathering Berlin)
- 1. Copyright Reaktor 2013 Luottamuksellinen
- 2. Case Against Scaling Back to basics with your enterprise transformation Sami Lilja Reaktor Twitter: @samililja Copyright Reaktor 2014
- 3. How big is big? Pick a piece of paper and a pen. Think about software / IT projects you have been working on. Based on your experience fill in the blank: A project is a big project when it has more than ___ people working on it. Copyright Reaktor 2014
- 4. Copyright Reaktor 2014 The Beef
- 5. Solving the wrong problem Source: http://www.medscape.com/viewarticle/806573 Copyright Reaktor 2014
- 6. Finding the right question The way we formulate the problem, limits our space of potential solutions Meetings take too much time People lack motivation Work environment, the system, Copyright Reaktor 2014 vs. We have too many meetings vs. decreases motivation Coordinating between parallel projects is not effective vs. We have too many parallel projects
- 7. Scaling Agile? The problem is not that we lack ways to scale Agile. The problem is not that we fail with Agile in large organizations. The problem is that we are large. Size does matter. Copyright Reaktor 2014
- 8. Copyright Reaktor 2014 Economy of Scale Scalability Size Cost of overhead Sublinear = Scales well Highly repeatable “How many?”
- 9. Labor-intensive work Copyright Reaktor 2014
- 10. Knowledge work? Copyright Reaktor 2014
- 11. When doing knowledge work in Large Scale, the key question is not “How?” – it is “Why?” Copyright Reaktor 2014
- 13. Organizations getting bigger Copyright Reaktor 2014
- 14. Pear-shape organizations Copyright Reaktor 2014
- 15. But, hey.. • Do you say companies should not grow? – No, I am not saying that • Do you say companies should only do small things? – No, I am not saying that – But.. small batches are better than large batches • Are you against the frameworks that promise Agility in large-scale? – No, I am against doing large-scale development – However, the frameworks take “large scale” as given and do very little to reduce that Copyright Reaktor 2014
- 16. What makes us Big? Large organization or project Fear of transparency Copyright Reaktor 2014 “We need lot of different competences” “Relative overhead is smaller” Complex systems Separating action, feedback, knowledge and decision making “Adding more people will speed us up” Big projects need lot of people need big projects need … Silos in organization (Conways’s Law) Lot of unfinished work (WIP) Belief in Economies of Scale Failure Demand Short-term (Project) thinking
- 17. The root of all Evil I Work-in-Progress Copyright Reaktor 2014
- 18. Work-in-Progress • How many things your organization is currently working on? • How easy it is to get dedicated people / team to deliver customer value? Copyright Reaktor 2014
- 19. Hey, but.. • Work-in-Progress and Little’s Law are about time through system • What does this have to do with project size? • Large amount of WIP helps to create unnecessarily large projects – When time-through-system gets long, some organizations add more people to gain speed – People are split to work on many parallel projects. Getting X people worth of work takes 10x more people Copyright Reaktor 2014
- 20. Work in Progress • Most organizations have too many things going on at one time, because – People are costly: Fear of <100% resource utilization – It is easier to start things than complete things – Large projects require lot of people require large projects require lot of people.. Copyright Reaktor 2014
- 21. Work in Progress • We underestimate the overhead that Work-in-Progress causes • In reality, large WIP causes huge and costly problems Copyright Reaktor 2014 – Delays • time-through-system = Work-In-Progress / Velocity – Queues and synchronization problems – Internal Failure Demand • Meetings, coordination effort, waiting, re-work, …
- 22. Sami’s Test #1 • Think about predictable demand that comes to your organization – Support request, creating a new service, ramp-up of a project, fixing a bug, … • What would be the fastest completion time for such work in your system, if you could do anything to make it happen? • If your current performance is lower, why is that? What causes delays? Copyright Reaktor 2014
- 23. Quick review • Pick a person near you, form pairs or triplets • In your small group, have a 2 minutes discussion about – What topics have been covered so far? – How do you feel about these topics? – What concepts have been significant to you? – How could you use these in your work? Copyright Reaktor 2014
- 24. The root of all Evil II Failure Demand Copyright Reaktor 2014
- 25. Value demand Adds value from customer point of view. Something customers are willing to pay for. This type of demand we want. Copyright Reaktor 2014 Failure demand Failure to do what customer needs Bad quality, wrong product or service, delay. No product or service. Missing either what or how customer wants Can account up to 80% of work
- 26. All demand is considered work Source: http://www.limebridge.com.au/page/Learning_Centre/Cartoons/ Copyright Reaktor 2014
- 27. Failure demand: Not only bugs NEW IT SYSTEM! Copyright Reaktor 2014 Value for user Fail CU ST O M E R S UP P O R T “PRESS 1 IF YOU CALL ABOUT..” “PRESS 2 IF YOU CALL ABOUT..” “PRESS 3 IF YOU CALL ABOUT..”
- 28. Internal Failure demand Do we need this process? What thinking created this? Copyright Reaktor 2014
- 29. Hidden Failure demand Copyright Reaktor 2014 Value Demand (Project work) Failure Demand (Bug fixes etc) Other The Plan Value Demand (Project work) Failure Demand (Bug fixes, meetings, waiting, coordination) Other The reality
- 30. Copyright Reaktor 2014 Dysfunction Something in the design and management of work that is causing problems.
- 31. Institutionalized Dysfunction Problem that is resolved by adding processes or management actions and then focusing on actions rather than the original problem. Copyright Reaktor 2014
- 32. Institutionalized Dysfunction and Agile Manifesto We are uncovering better ways of developing software by doing it and helping others do it. Through this work we have come to value: Individuals and interactions over processes and tools Working software over comprehensive documentation Customer collaboration over contract negotiation Responding to change over following a plan That is, while there is value in the items on the right, we value the items on the left more. Copyright Reaktor 2014 Slippery slope to institutionalized dysfunction
- 33. Sami’s Test #2 • Assume you could freely choose the smallest possible number of people to implement the product or service. • How large would that group be? • If such group is significantly smaller than your current development project, why is that? Copyright Reaktor 2014
- 34. Quick review • Pick a person near you, form pairs or triplets • In your small group, have a 2 minutes discussion about – What topics and issues have been covered so far? – How do you feel about these topics? – What topics or concepts have been significant to you? – How could you use new learning in your work? Copyright Reaktor 2014
- 35. THE WAY OUT? Copyright Reaktor 2014
- 36. What makes us Big? Large organization or project None of these are Laws of Nature. None of these are imposed on you. Fear of transparency Copyright Reaktor 2014 “We need lot of different competences” “Relative overhead is smaller” Complex systems Separating action, feedback, knowledge and decision making “Adding more people will speed us up” Big projects need lot of people need big projects need … Silos in organization (Conways’s Law) Lot of unfinished work (WIP) Belief in Economies of Scale Failure Demand Short-term (Project) thinking These are the results of thinking. And we can get rid of these if we want.
- 37. Putting things in perspective • Up to 80% of work in organization is Failure Demand – What if you could reduce it, just a little bit? • Significant amount of work in project is caused by large amount of WIP – What if that improves as well? • Keep in mind the pear-shape organization and super-linear cost of scaling… Copyright Reaktor 2014
- 38. Copyright Reaktor 2014 Size Cost of overhead Sublinear = Scales well Reducing project size by X% decreases costs by a lot more than X% X%
- 39. Copyright Reaktor 2014 But hey, … • “You are overreacting. We all know large scale may not be the best solution. But it is usually unavoidable. And it works”
- 40. It is not perfect but it works If it works, don’t fix it - American Car manufacturers, 1970s Copyright Reaktor 2014
- 41. Sami’s Test #3 • OK, let’s assume you’ve done everything to limit WIP, remove failure demand and reduce complexity • Still your project is Large(-ish) .. At least 3x bigger than “by-the-book” Agile project • Doing things in large scale is the only option. And you want to do it Agile. • Have you done a very successful small end-to-end Agile project before attempting a large scale Agile project? Copyright Reaktor 2014
- 42. SCRUM AT LARGE SCALE Copyright Reaktor 2014
- 43. What is missing from this picture? I can not see the customer here.. Copyright Reaktor 2014
- 44. What is missing from this statement? What if the demand comes from end user? What if we are creating a service? Copyright Reaktor 2014
- 45. Asking the right question • “How can I make Scrum work in my organization and my projects” – Coaching, Teaching, Inspect and Adapt • “How can I fulfill value demand from customer using Scrum” – Coaching, Teaching, Inspect & Adapt, … – Study and understand demand – Design and manage work against Value Demand! Copyright Reaktor 2014
- 46. Forgetting customer • Forgetting customer may lead to tunnel vision – Creating backlogs for teams rather than for customer need – Having unnecessary dependencies between teams (and product owners) – Internal failure demand: synchronization, prioritization, waiting – Lack of purpose: Backlog represent “work” instead of “Customer need” Copyright Reaktor 2014
- 47. Copyright Reaktor 2014 Customers, users Customers, users Demand Demand Value for customers and users Value for customers and users
- 48. BEYOND SCRUM: " SOLID FRAMEWORK" " HTTP://SOLID.REAKTOR.FI Copyright Reaktor 2014
- 49. • Three different levels and frames for thinking – Note: These boxes are not separate organizations • Purpose of Solid: Help organizations design and manage work so that value demand from customer is fulfilled • Solid helps organizations to find the right questions Copyright Reaktor 2014
- 50. Operational Level looks at value delivery flow in order to maximize the Return on Investment. Thinking frames • True Agility • Kanban, Scrum, XP Tactical Level aims to find the right products and services. It also provides tools to manage investment (project) portfolio. Thinking frames • Customer development & Lean startup • Data science Strategic Level provides understanding about markets and customer needs. It helps to design workflow and value delivery system. Thinking frames • Systems Thinking • Outside-in perspective Copyright Reaktor 2014
- 51. Summary • Attempts to do knowledge work in large scale are likely to fail – …or at least suboptimal way to create products and services • Main reasons for being large – Lot of parallel work-in-progress – Inability to see and remove failure demand – Fear of fast feedback and transparency • Large Scale is a System Condition • In order to change System Conditions, we need new thinking – We need to find better questions, not just good answers Copyright Reaktor 2014
- 52. solving the right problem? We are engineers. We are trained to solve problems. Copyright Reaktor 2014
- 53. Ways to deal with a problem Dissolution: redesign the system or its environment in such a way that it eliminates the problem or the conditions that caused it solution: discover or create behavior that yields the best, or approximately the best, possible outcome, one that "optimizes" the situation. Resolution: employ behavior previously used in similar situations, adapted if necessary, so as to obtain an outcome that is good enough. Absolution: ignore a problem and hope it will solve itself or go away of its own accord. http://results2match.com/ackoff-again-4-different-ways-of-solving-a-problem Copyright Reaktor 2014
- 54. Dissolution: redesign the system or its environment in such a way that it eliminates the problem or the conditions that caused it Copyright Reaktor 2014