際際滷

際際滷Share a Scribd company logo

Reach for the stars

Dan Brown
Dan Brown

This document discusses challenges to Kanban principles and provides examples to illustrate how those principles can be applied. It begins with an example of modeling lead time and throughput at a drive-thru to show how these metrics can be used. It then discusses how NASA successfully applied the principles of limiting work in progress and validated learning incrementally to achieve major goals like landing on the moon. The document argues these principles can be used in any organization to improve predictability and outcomes.

1 of 32
Reach for the stars
Lets Challenge Preconceptions Estimates inform us when things will finish Kanban uses difficult maths You cant limit WIP in big organisations Validated learning doesnt deliver anything 息 2013 ripplerock Dan Brown @KanbanDan
How good an estimate would you have in 30 seconds? 30 minutes? 5 minutes? 2 weeks?
Reach for the stars
Everything that doesnt have a cross on it is a whole galaxy Your friends cant help you now?
Lets Substitute Predictability for Estimation This presentation contains Maths. I will be asking you some questions. But I have an example to make it as painless as possible... 息 2013 ripplerock Dan Brown @KanbanDan
Lets go to the Drive-Thru What is a Drive-Thru Typically found in fast food You stay in your car You drive around the building You: Order your food Pay for your food Collect your food All through your car window After you collect you drive away with your food 息 2013 ripplerock Dan Brown @KanbanDan
Drive-thu example Let me define my terms to be clear Lead Time - the time from a particular customer driving up, to driving away with a burger Throughput Rate - how frequently customers drive away with food Original Drive-Thru only had 1 window So if it takes 90s to get served with 1 window: Avg Lt is 90 seconds Avg Tr is 1 customer per 90 seconds
Fast Forward in time Some people worked out it could be improved 2 Window system order & pay at first window (45s), collect at second window(45s) How does that affect our measurements? Avg Tr is now 1 every 45s, Avg Lt is now 90s 息 2013 ripplerock Dan Brown @KanbanDan
Pop Quiz What happens with 3 windows? 30 seconds to order 30 seconds to pay 30 seconds to collect What is Lead Time? 90s What is Throughput rate? 1 every 30s 息 2013 ripplerock Dan Brown @KanbanDan
But who cares? Your customers care! Throughput Rate: How frequently new Features come off the line Lead Time: when will this Feature be done if we started now Allows us to predict when whole Product will be done 息 2013 ripplerock Dan Brown @KanbanDan
Dan, when will the product be done? If you deliver 1 work item every 2 days Your Tr = 0.5 items per day (units must match) If your Lt is 11 days If you have 100 work items to finish Your total Product Time = 11 + ( 100 / 0.5 ) Pt = 211 days Product Time for a new project is: Lt + ( Number of Features / Tr ) But take note of variance to the averages of Lt & Tr to give tollerances! 息 2013 ripplerock Dan Brown @KanbanDan
But Dan, how can I use maths to help me? You can use Littles Law (for stable systems) to link Tr, Lr and WIP in a simple equation but We dont have time for that right now. You could always come talk to me afterwards Or attend an LKU Accredited Kanban Course - Real Kanban for example www.ripple-rock.com/training/real-kanban.aspx 息 2013 ripplerock Dan Brown @KanbanDan
Back to the drive-thru 2 windows are open, but Window 2 actually takes 50s Window 1 takes only 40s What is the Tr? WIP is 2, Tr = 1 per 50s, so Lt = 2 * 50s = 100s ( thanks to Littles Law) Why is this not 90s? 息 2013 ripplerock Dan Brown @KanbanDan
But in the real world we get a queue between windows of 3 cars (limited by space) WIP isnt 2 then, its really the 2 at windows plus the 3 queuing, so what is the WIP now? 5! 息 2013 ripplerock Dan Brown @KanbanDan
So what difference does that make? With WIP of 5 Tr is still = 1 per 50s Lt = Tr * WIP Whats the new Lt? 250s! Increasing the WIP without reducing the Tr increases the Lt! Maths done 息 2013 ripplerock Dan Brown @KanbanDan
Oh, and by the way
How do you make a footprint on the moon? You finish One small step at a time! NASA says: Do one thing at a time, with supreme excellence. A colleague once told me: As soon as our clients work out that all they have to do is put everything into an ordered list, then finish them one at a time well be out of a job We keep saying it, but were still in jobs 息 2013 ripplerock Dan Brown @KanbanDan
NASA Limited WIP in Action Do One Thing At A Time Weve seen the maths and we can measure why it works In the 1950s and 1960s NASA were living it And they still are 息 2013 ripplerock Dan Brown @KanbanDan
What is their one thing now Who supplies the International Space Station? 息 2013 ripplerock Dan Brown @KanbanDan
So what are NASA doing? 息 2013 ripplerock Dan Brown @KanbanDan
All of NASA? They have a separate division called the JPL They do the space telescopes like Hubble Now they are doing James Webb SST 息 2013 ripplerock Dan Brown @KanbanDan
With Supreme Excellence Not just about showing off Focus on QUALITY! Post launch bugs mean something different to NASA Remember the fuss about Hubbles focus? James Webb will be out of reach of humans 息 2013 ripplerock Dan Brown @KanbanDan
What about us? Isnt everything Safe to Fail? Yes and No. Yes before launch, No after launch. There are situations where the blue screen of death isnt just a phrase But even when its not, fixing bugs in production is the most expensive place 息 2013 ripplerock Dan Brown @KanbanDan
If you love it, let it go One of the key Kanban lessons: If you focus on Throughput, quality drops, but then what? Bugs, Technical debt, slow throughput 25 Throughput Tech Debt Bugs 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 -5 (Faked exaggerated data to illustrate the point) 息 2013 ripplerock Dan Brown @KanbanDan
With a quality focus Focus on Quality what happens? Bug counts & Tech debt drop What happens to throughput? 25 Throughput Tech Debt Bugs 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 (Faked exaggerated data to illustrate the point) 息 2013 ripplerock Dan Brown @KanbanDan
How did we get there By finishing one small step at a time NASA started manned space flight with Mercury Gemini was about learning how to go to the moon 2 weeks in space for the first time Docking spacecraft Then came Apollo 息 2013 ripplerock Dan Brown @KanbanDan
Apollo 1 landed on the moon right? Not quite Apollo 8 round the moon Apollo 9 test out the LEM Apollo 10 drop the LEM within 9 miles of the moon Apollo 11 one small step 息 2013 ripplerock Dan Brown @KanbanDan
Incremental steps They did it by doing it. The POC was real launches with real Validated Learning Each step moved NASA forward and enabled the next step The Moonshot started with Wernher Von Braun and the V2! 息 2013 ripplerock Dan Brown @KanbanDan
What can we learn? If NASA can limit WIP, so can we all Even the biggest of big bangs can be delivered incrementally Validated Learning leads to success Tackle Risk by using Collaborative Experimentation 息 2013 ripplerock Dan Brown @KanbanDan
My Challenge Estimates inform us when things will finish Only when things arent complex Predictability based on real metrics is much better Kanban uses difficult maths Simple maths gets you most of the value You cant limit WIP in big organisations If NASA can so can we What is really stopping us? Validated learning doesnt deliver anything It gets us through to where we need to be 息 2013 ripplerock Dan Brown @KanbanDan
Any Questions? 息 2013 ripplerock Dan Brown @KanbanDan

More Related Content

Reach for the stars

  • 2. Lets Challenge Preconceptions Estimates inform us when things will finish Kanban uses difficult maths You cant limit WIP in big organisations Validated learning doesnt deliver anything 息 2013 ripplerock Dan Brown @KanbanDan
  • 3. How good an estimate would you have in 30 seconds? 30 minutes? 5 minutes? 2 weeks?
  • 5. Everything that doesnt have a cross on it is a whole galaxy Your friends cant help you now?
  • 6. Lets Substitute Predictability for Estimation This presentation contains Maths. I will be asking you some questions. But I have an example to make it as painless as possible... 息 2013 ripplerock Dan Brown @KanbanDan
  • 7. Lets go to the Drive-Thru What is a Drive-Thru Typically found in fast food You stay in your car You drive around the building You: Order your food Pay for your food Collect your food All through your car window After you collect you drive away with your food 息 2013 ripplerock Dan Brown @KanbanDan
  • 8. Drive-thu example Let me define my terms to be clear Lead Time - the time from a particular customer driving up, to driving away with a burger Throughput Rate - how frequently customers drive away with food Original Drive-Thru only had 1 window So if it takes 90s to get served with 1 window: Avg Lt is 90 seconds Avg Tr is 1 customer per 90 seconds
  • 9. Fast Forward in time Some people worked out it could be improved 2 Window system order & pay at first window (45s), collect at second window(45s) How does that affect our measurements? Avg Tr is now 1 every 45s, Avg Lt is now 90s 息 2013 ripplerock Dan Brown @KanbanDan
  • 10. Pop Quiz What happens with 3 windows? 30 seconds to order 30 seconds to pay 30 seconds to collect What is Lead Time? 90s What is Throughput rate? 1 every 30s 息 2013 ripplerock Dan Brown @KanbanDan
  • 11. But who cares? Your customers care! Throughput Rate: How frequently new Features come off the line Lead Time: when will this Feature be done if we started now Allows us to predict when whole Product will be done 息 2013 ripplerock Dan Brown @KanbanDan
  • 12. Dan, when will the product be done? If you deliver 1 work item every 2 days Your Tr = 0.5 items per day (units must match) If your Lt is 11 days If you have 100 work items to finish Your total Product Time = 11 + ( 100 / 0.5 ) Pt = 211 days Product Time for a new project is: Lt + ( Number of Features / Tr ) But take note of variance to the averages of Lt & Tr to give tollerances! 息 2013 ripplerock Dan Brown @KanbanDan
  • 13. But Dan, how can I use maths to help me? You can use Littles Law (for stable systems) to link Tr, Lr and WIP in a simple equation but We dont have time for that right now. You could always come talk to me afterwards Or attend an LKU Accredited Kanban Course - Real Kanban for example www.ripple-rock.com/training/real-kanban.aspx 息 2013 ripplerock Dan Brown @KanbanDan
  • 14. Back to the drive-thru 2 windows are open, but Window 2 actually takes 50s Window 1 takes only 40s What is the Tr? WIP is 2, Tr = 1 per 50s, so Lt = 2 * 50s = 100s ( thanks to Littles Law) Why is this not 90s? 息 2013 ripplerock Dan Brown @KanbanDan
  • 15. But in the real world we get a queue between windows of 3 cars (limited by space) WIP isnt 2 then, its really the 2 at windows plus the 3 queuing, so what is the WIP now? 5! 息 2013 ripplerock Dan Brown @KanbanDan
  • 16. So what difference does that make? With WIP of 5 Tr is still = 1 per 50s Lt = Tr * WIP Whats the new Lt? 250s! Increasing the WIP without reducing the Tr increases the Lt! Maths done 息 2013 ripplerock Dan Brown @KanbanDan
  • 17. Oh, and by the way
  • 18. How do you make a footprint on the moon? You finish One small step at a time! NASA says: Do one thing at a time, with supreme excellence. A colleague once told me: As soon as our clients work out that all they have to do is put everything into an ordered list, then finish them one at a time well be out of a job We keep saying it, but were still in jobs 息 2013 ripplerock Dan Brown @KanbanDan
  • 19. NASA Limited WIP in Action Do One Thing At A Time Weve seen the maths and we can measure why it works In the 1950s and 1960s NASA were living it And they still are 息 2013 ripplerock Dan Brown @KanbanDan
  • 20. What is their one thing now Who supplies the International Space Station? 息 2013 ripplerock Dan Brown @KanbanDan
  • 21. So what are NASA doing? 息 2013 ripplerock Dan Brown @KanbanDan
  • 22. All of NASA? They have a separate division called the JPL They do the space telescopes like Hubble Now they are doing James Webb SST 息 2013 ripplerock Dan Brown @KanbanDan
  • 23. With Supreme Excellence Not just about showing off Focus on QUALITY! Post launch bugs mean something different to NASA Remember the fuss about Hubbles focus? James Webb will be out of reach of humans 息 2013 ripplerock Dan Brown @KanbanDan
  • 24. What about us? Isnt everything Safe to Fail? Yes and No. Yes before launch, No after launch. There are situations where the blue screen of death isnt just a phrase But even when its not, fixing bugs in production is the most expensive place 息 2013 ripplerock Dan Brown @KanbanDan
  • 25. If you love it, let it go One of the key Kanban lessons: If you focus on Throughput, quality drops, but then what? Bugs, Technical debt, slow throughput 25 Throughput Tech Debt Bugs 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 -5 (Faked exaggerated data to illustrate the point) 息 2013 ripplerock Dan Brown @KanbanDan
  • 26. With a quality focus Focus on Quality what happens? Bug counts & Tech debt drop What happens to throughput? 25 Throughput Tech Debt Bugs 20 15 10 5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 (Faked exaggerated data to illustrate the point) 息 2013 ripplerock Dan Brown @KanbanDan
  • 27. How did we get there By finishing one small step at a time NASA started manned space flight with Mercury Gemini was about learning how to go to the moon 2 weeks in space for the first time Docking spacecraft Then came Apollo 息 2013 ripplerock Dan Brown @KanbanDan
  • 28. Apollo 1 landed on the moon right? Not quite Apollo 8 round the moon Apollo 9 test out the LEM Apollo 10 drop the LEM within 9 miles of the moon Apollo 11 one small step 息 2013 ripplerock Dan Brown @KanbanDan
  • 29. Incremental steps They did it by doing it. The POC was real launches with real Validated Learning Each step moved NASA forward and enabled the next step The Moonshot started with Wernher Von Braun and the V2! 息 2013 ripplerock Dan Brown @KanbanDan
  • 30. What can we learn? If NASA can limit WIP, so can we all Even the biggest of big bangs can be delivered incrementally Validated Learning leads to success Tackle Risk by using Collaborative Experimentation 息 2013 ripplerock Dan Brown @KanbanDan
  • 31. My Challenge Estimates inform us when things will finish Only when things arent complex Predictability based on real metrics is much better Kanban uses difficult maths Simple maths gets you most of the value You cant limit WIP in big organisations If NASA can so can we What is really stopping us? Validated learning doesnt deliver anything It gets us through to where we need to be 息 2013 ripplerock Dan Brown @KanbanDan
  • 32. Any Questions? 息 2013 ripplerock Dan Brown @KanbanDan

Editor's Notes

  • #7: EVERYTHING is a 5!
  • #11: Need to give them some figures here
  • #12: What about a project with already in flight stuff? = 200 in this example remove the priming initial lead time