am 11. März 2011
This book is a must read for everyone interested in product development flows and especially for all product development managers.
Summary of the book
Don Reinertsen describes in his book 174 principles how to develop an improved, more profitable development flow.
He uses ideas from lean manufactoring, economics, queueing theory, statistics, the internet, computer operating system sesign, control engineering and maneuver warfare and adapts them to product development.
The author states that todays problems in product development are caused by failure to correctly quantify economics, blindness to queues, worship of efficiency,hostility to variability, worship of conformance, institutionalization of large batch sizes, underutilization of cadence, managing timelines instead of queues, absence of WIP constraints, inflexibility, noneconomic flow control, and centralized control.
He emphasizes that it is important to treat development decisions as economic decisions. This is in his opinion much better than chasing whichever proxy variable is popular at the moment.
Developing an economic framework and decisions rules help to quickly process the small, perishable economic decisions. The economic models and rules do not have to be perfect, because even imperfect rules improve decision making.
A decision rule that is missing in most companies today is a quantified cost of delay.
Reducing Queues is the key to improve product development performance. The hidden cost of queues are often overseen. They are the hidden source for most development waste, because they increase cycle time, expenses, variability and risk and they slow feedback, reduce quality, and decrease motivation.
Queues are linearly increased by variability and exponentially by high capacity utilization.
The structure of the queueing system affects the queues, e.g. sharing a queue for multiple servers. The Optimum queue size occurs when we balance the cost of capacity agains delay cost. The cost of queues can be reduced by a queuing discipline.
The commulated flow diagram (CFD) and Little's formula are tools to visualize and analyze queues and to relate queue size to cycle time.
Variability is, in contrast to manufactoring, not always bad in product development, because here the economic payoff-functions are frequently asymmetric.
It is important to focus on the cost of variability instead of worrying about the amount of variability.
On the other side variability can be reduced using variability pooling, creating covariance, exploiting reuse and especially by reducing batch size.
Small batch sizes are desirable to product developers, because they reduce queues, accelerate feedback and reduce risk. Decreasing transaction costs, and thus allowing smaller batch size, has an enormous economic value. "Although batch size is a mature concept in manufactoring for almost a century, today, most product developers do not even recognize batch size as an issue."
Work in Process (WIP) constraints can be used to control cycle time, since they influence queue size which determines cycle time. They are powerful because they are inexpensive, incremental and reversible. They prevent queues from reaching dangerously expensive levels of congestion. "Unfortunately, today less than 1% of product development processes are managed with WIP constraints."
Cadence and synchronization are known from lean manufacturing. Time based cadences have big advantages. Synchronization can reduce queues without changing either batch-size and capacity utilization.
By borrowing concepts form computer Operating system design, we can develop network based development processes, that use flexible sequencing, tailored routes, alternate routes, late binding and some other methods and do much better fit the needs in PD than the linear processes used today.
The goal in managing the PD process is to increase the economic outcome. We can design control systems that must deal with multiple objectives and moving targets to achieve this goal.
The concepts of leading indicators and balanced set points helps to improve the process. Product development presents us with unplanned economic opportunities that we should exploit.
Fast feedback reduces queues and accelerates learning. It can be achieved using small batches and control systems that handle small batches of information.
Accelerated feedback loops improves efficiency and gerate urgency. Slow feedback undermines urgency in PD.
It is important to have additional fast local feedback loops. They work best when de decentralize control.
There are some metrics that support flow based product development, e.g. measuring design-in-process inventory, cost of queues, aging of items in queue, batch size trends, feedback speed, decision cycle time, number of multipurpose resources, etc.
Robust systems combine centralized and decentralized control e.g using layered control and virtual centralization.
Centralized control works well for problems that are infrequent, large, or those with scale economies.
Decentralized control permits us to adapt to uncertainty. It works well for problems and opportunities that are perishable.
Decentralized control requires a set of technical skills that can be trained and practiced. It can be supported by a set of management choices that enable lateral communications and the creation of cohesive teams. Decentralized control is important and decisions rules and mission statements are necesary for it.
am 24. April 2014
Eine sehr gute Hilfe für alle, die ihre Produkt-Entwicklung vollständig und flexibel, optimiert auf Entwicklungsziele ausrichten wollen.
Ein paar Erkenntnisse aus dem Buch zum Anfüttern:
(1) Entwicklungsergebnisse sollen Profit bringen. Alle Aktivitäten sollten darauf ausgerichtet sein und sich daran messen. Stellgrößen sind insbesondere Entwicklungskosten, Herstellkosten, Produkt-Features, Time-to-Market und Risko/Versicherungen. Nicht die Optimierung einer Stellgröße bringt den Erfolg, sonders das aktive stetige Einstellen des richtigen Arbeitspunktes für alle Stellgrößen zusammen.
(2) Entscheidungskurven für a oder b ergeben U-Kurven, die in der Regel ein breites Minimum haben. Das bedeutet, solange die Entscheidung nicht auf dem Rand der U-Kurve liegt, liegt man nicht weit vom Optimum weg. Die Entscheidung kann also mit einer beruhigenden Ungenauigkeit getroffen werden.
(3) Mehr Auslastung von Entwicklern führt bei Änderungen und Taskwechsel zu relativ mehr Overhead. Die effiziente Ausnutzung von Entwicklern liegt eher in der Größenordnung von (z.B.) 65-75% als bei 90-100%. Das genaue Optimum muss im eigenen Unternehmen gefunden werden.