Organized for Speed

In this article I will explain how to improve product development flow and deliver value faster.  I will describe how a few small changes in the way we are organized can make a significant difference in our value delivery potential.  These insights are based on some widely held theories and – more importantly – practical experience gleamed from working with numerous companies, in an array of industries. I will also explain the place and importance of uncertainty in our working world and how a more Lean mindset can help us view this uncertainty as as opportunity to better understand – and exploit – our options.

The world is Full of Uncertainty

Figure03-3Bc Before we think about delivering faster, we must make sure we are developing the right product. A useful map for navigating your way into the concepts and field of complexity is « The Stacey Matrix ». It is important to clarify our assumptions on where we are before choosing one approach. Agile has a proven record in complicated and complex zones. Lean is excellent for Complicated situations. And Lean Startup is probably the best approach to date in matters Chaotic. The temptation to oversimplify when we need to explain complicated or complex problems is understandable, even to some extent our attempts to describe things that are chaotic in nature.  Unfortunately in adopting this mindset – this oversimplification – we devalue and reduce the likelihood that we can fully understand this inherent uncertainty, and the potential to use this to our advantage.

Exploiting Uncertainty

Cubical dice Imagine you’re a gambler, a professional gambler. Is uncertainty your enemy? You’re playing in your favorite casino.  Are you ready?  Then let’s play! Go on  : http://www.roll-dice-online.com/

  • select 2 dices of 6 sides :
  • Let’s assuma we have a gain opportunity of 40€ if sum of drops is 7 or more. 0€ if 6 or lower.
  • Cost of 10€ per roll and per dice.
  • Even if amount is small same player can play as quick as he wants.
  • The casino take 17% off of all gain for services.

Standard: you roll the two dices at same time:

Average gain: 23€. 19,9€ after casino charges, Average cost: 20€.
Return On Investment: 0,99  Loss of 1%(*)

(*) this is how casino makes money. What if you could use feedback loop ? i.e rolling one dice, look at the result and decide to roll or not to roll the second one. What would we intuit about our chance to gain 40€?

Average gain: 22€. 19€ after casino charges. Average cost  18€
Return On Investment: 1,036  Benefit of 3,6% ( 1€ )

Calculation at the end of the article.

  1. Even if the second strategy sounds more complex. It is still more valuable. Despite the immediate feeling of failure failed (here by getting an ace at fisrt roll) we should use it as learning to take decision.
  2. Small decision sometime give small benefits but this is still a value for the business if decisions are taken and executed with speed: An experimented gambler could play one game every 5 second i.e 720 games in one hour.  His/her average gain could be up to 720€ per hours.
  3. Even if casino is of course based on uncertainity, Such a gambler has no risk of failure with so huge number of drops. A system able to take very fast small decision transform uncertainty into more predictible results. We could say it reduce uncertainty but it would be not acknoledging that uncertainty is the issue to solve not the one to underestimate.

Of course, taking many fast decisions will be a competitive advantage if  those are in average good decisions. Here, gamblers are experts in both analysing feedback and executing the right corrective actions (i.e do not roll the second dice if an ace occurs at firt roll) when appropriate. This intense use of feedback and multiple fast decisions close to the floor plant is one of lean principles. All Agile teams I know also make sure to have maximize  feedback prior considering a job as finished. It is why test skills are so important on side of coding building and architectural skills inside a single scrum team.

Centralized strategic decision, decentralized execution

Now let’s say you own a company employing up to 1000 gamblers. As soon as you heard about this casino and their ‘generous’ rules, you made the strategic decision to deploy your gamblers to this high lucrative game. This is centralized decision making. You then explained the principle of feedback to your employees and how it can be used to help them make better tactical decisions – you decentralized the execution.
If only life were so simple … 
Very soon the casino discovers your strategy and changes the rules.  News of the changes will quickly propagate through your network of gambler employees. 
 
Now, let’s imagine your Gamblers are not allowed to take the initiative to stop gambling or even to look for new ways to make money, at the same time they inform you that the dice are no longer loaded in their favor.

This information flow – from ground to governance – is the second feedback loop and is exactly how firms should operate. Non aligned decisions are avoided by having a clear system of governance. The aim of governance is to only take strategic decisions that will inform all linked actions – execution details and policies are decided on the ground, with support from ‘enablers’ to bridge the gap and provide operational support. In his book about organizing for complexity Niels Pflaeging  describe it by concentric circle . The market is at the periphery.  At scale some centralization might be needed at the center. Of course the more direct the contact is between people at the center and those at the periphery, the better.  Role of people at the center is to take right decision to help those who serves the market. They also always try to reduce cognitive distance between them and people close to the circumference. decentralizeFinding the right balance is not easy: The famous motorbike manufacturer Harley Davidson® took a radical approach to delegate all « non strategic » decision to the factory floor and also to share all business information, including cost margins. Those figures used to be confidential in traditional enterprise but not in Freedom Inc. They also encouraged/required companies in their supply chain to do the same thing.

And the result was quite remarkable – customer satisfaction went up and so did profits. 

Balance Demand against throughput. Cost of delay and visual management

What Theory of constraint say is simple:

  1. identify the constraint
  2. Exploit the constraint
  3. Subordinate the constraint
  4. Elevate the constraint
  5. Do not let the Inertia become the constraint ( Aka Goto step1)
Many tools exist to deals with bottleneck issue. This is an area for creative thinking
Another example is Value Stream Mapping, which focuses on the interdependence of the twin flows of material and information. This is not about scientist calculation of what would be better, but instead is an exercise in learning through observation and visualizing what actually happens.
My 2 cents tip is just to make all task done by the team visible. To do so, there is nothing better that using the room wall.
The most interesting part is to show where work or information is waiting, how long it has waited, and what is the consequence of that wait.  Some teams move a vertical bar to track the number of days expended in an iteration/sprint.  On the factory floor, this would be inventory. In product development, this is just queues.
On the factory floor, this would be inventory. In product development, this is just queues.
One operating principle that Agile uses is to avoid any internal queues inside the team, or limit as much as is practically possible.  This requires an ongoing diligence from the team.  And even with the best intentions there are thousand of reasons to have queues of either information or partially finished products.  The main question is which queue should I take care of first.
Even in agile there are thousand of reason to have queues or either information or partially finished products. The main question is which queue I should take care of first.
Don Reinertsen gives a very good principle to to do that. He suggest to attach an economics indeed a cost to each queue. name for that is Cost Of Delay.
Then the game sounds like the Dice early game of that article. Gamers just tackle the highest cost of delay first.
In Agile, this is I think the best way to order any queue. From initial idea, to funneled idea ( that came with some business and rough estimate assumption) to idea read to be executed, to idea under execution and deployment.

Sized adequately

The book This is Lean describe 4 quadrant:

ResourceVSFlowMatrixLean

Flow has been studied for long on highway.   a simple approach is called the Greenshield’s Model.

Greenshield made the assumption that, under uninterrupted flow conditions, speed and density are linearly related. This relationship is expressed mathematically and graphically below.

Diagram of Speed versus Density

This makes as a result

Diagram of Flow versus Density

Mary Poppendick show that batch size has also a tremedous impact on the Flow, smal bacth are less sensitive to load.

Only limitation is the transaction cost, which are fixed cost. In Lean this is called SMED ( Single Minute exchange of Die by analogy to changing die in car parts manufacturing going from days to few minutes in Toyota Produc System). Optimum value is where transaction cost is about the same amount than the waiting cost.

The graph below show how it affects the flow. Here transaction cost is about 1/4 of smaller batches.

When operating over 80% we mainly pay tax, indeed jobs spend far more time in queue than running.

At 80% of load, Big batch cost about 60% more.

Even surprising, I had the chance to see that many times and this is used in all computer congestion algorythm.

Capture d'écran 2015-04-13 08.07.47

All good Agile teams practice this by defining work items that bring value to a customer ideally sized about 1/10 to 1/3 of their fixed iterations duration.

Reduce Work In Process, Deliver Often

Reducing the size of the work is only beneficial if we agree not to do too much work at the same time. This is done by limiting the work in process. This is what Kanban brings to Agile.
Even if this seem obvious, this is the most difficult part of team works. This could mean working with others on something we do not know perfectly because this is the highest priority and not something I know the best. Teams generally need to practices.
One tip here is to look for the numerous kanban games and play some with the full team.
Management could also affect negatively Kanban. Even if they understand the need to have less centralized decision, they might still push for urgent things without going through the Cost of Delay. Just by the fact that management is too often setting goal for the team instead of the customer, many companies has a huge cultural biased against organization for speed. Good news is that more and more company get rid of goals and objective set by management for benefits of goals set by the customers. Jurgen Appelo develops a good vanilla compilation of management tools and techniques that could help having management back into the race of improving the speed.
There is a funny things on limiting WIP, in fact it is not so important to choose precisely what should be that limit. Don Reinertsen show this is indeed following a U curve:
Capture d'écran 2015-03-29 19.07.33
I would rather recommend to look at kanban simulation tool and see how this also reduce dramatically the variability.

Challenges

We are not used to limit the work in process. Instead we love being busy occupied. Instead we should really take care about developing a Discipline Of Delivery.

  • Our business assumption are not always easy to validate. I will not detail here since Lean Startup would be a far better answer.
  • Due to Cultural and historical biased we tend to over-centralize decision even multiply number of layers. This is the main management roadblock and slow down the full system.
  • Do not start something else before you have reached the quality the end customer will need on the previous items. This is done by making sure there is an agreement on what to do , how much ( Acceptance criteria ) and how well ( Definition of done).
  • Make daily things visible so that you will see bottleneck(s).  There are many ways to do visual management but using a physical wall is my favorite.
  • Spend 15 minute a day describing on that board what’s happened since last day, what ‘ next and what could be done to be faster on the highest cost of delay priority.
  • Make sure you set/collect measurements such as cumulative flow diagrams and have dedicated time at regular intervals to look at how to improve the process/change the WIP limit..

Definition:

Cognitive distance Cognitive distance is used to describe how two individuals are different, at the knowledge level, but also in the way they perceive and interpret external phenomena, that is to say somehow in their modes of reasoning. Nooteboom (2000, p. 73) defines the concept: « a difference in cognitive function. This Can Be a Difference in domain, range, or mapping. « The more individuals have different functions or « mental models », the more the cognitive distance between them is important.

References Stacey complexity graphs http://www.gp-training.net/training/communication_skills/consultation/equipoise/complexity/stacey.htm

PDCA http://en.wikipedia.org/wiki/PDCA

Principles of product development flow  second generation lean product development.by Don Reinertsen

Introduction to product flow development by Yuval Yeret:  http://fr.slideshare.net/yyeret/introduction-to-product-development-flow?qid=53b5a67d-24ac-4e49-9044-bcca7416097f&v=default&b=&from_search=2

Queuing theory simulation: by Tero Parviainen  http://ss15-teropa.divshot.io/

Organized for Complexity by Niels Pflaeging« >http://fr.slideshare.net/npflaeging/slideshelf

More Than a Motorcycle: The Leadership Journey at Harley-Davidson (Harvard Business School Press. By Rich Teerlink, former CEO and chairman of Harley-Davidson Motor Company and coauthor, with Lee Ozley, of 2000) http://freedomincbook.com/2010/09/03/strategybusiness-presents-freedom-inc/

https://hakanforss.wordpress.com/2014/08/25/waste-of-overproduction-greater-than-waste-of-highly-skilled-people-idling/

http://rei.revues.org/3814#tocto1n2 http://fr.slideshare.net/jchyip/balance-demand-against-throughput

https://www.toc.tv/TV/video.php?o=p-16245&ac=sus&mtr=p-16245&open=excerpt&id=17#.VRgfoWY3m1U http://blackswanfarming.com/cost-of-delay/

Phoenix project http://en.wikipedia.org/wiki/The_Phoenix_Project:_A_Novel_About_IT,_DevOps,_and_Helping_Your_Business_Win

Calculation of the Dice Game:

  2 dices per drop stop if first dice is 1
Fisrt Dice average gain per game average spent per game average gain per game average spent per game
1 40 6,7 20,0 0,0 10,0
2 80 13,3 20,0 13,3 20,0
3 120 20,0 20,0 20,0 20,0
4 160 26,7 20,0 26,7 20,0
5 200 33,3 20,0 33,3 20,0
6 240 40,0 20,0 40,0 20,0
23,3 20,0 22,2 18,3
19,94 18,99
ROI ROI
1,167 1,212
casino of course take part of the gain : here 17 %
ROI after taxes: 0,997 ROI after taxes: 1,036
Publicités

Laisser un commentaire

Entrez vos coordonnées ci-dessous ou cliquez sur une icône pour vous connecter:

Logo WordPress.com

Vous commentez à l'aide de votre compte WordPress.com. Déconnexion / Changer )

Image Twitter

Vous commentez à l'aide de votre compte Twitter. Déconnexion / Changer )

Photo Facebook

Vous commentez à l'aide de votre compte Facebook. Déconnexion / Changer )

Photo Google+

Vous commentez à l'aide de votre compte Google+. Déconnexion / Changer )

Connexion à %s