The Swiss outlet of the Limited WIP Society today hosted Donald Reinertsen, one of the thought leaders of the Lean community. He gave a speech on second generation lean product development.
Event hosted by the Limited WIP Society Switzerland.
Developers and their Economic Choices
Developers can make better economic choices more quickly, if we provide them with good decision support information.
Developers should be provided with an even more and better decision base. Non value added activity should be left out. But which are they? It’s hard to compare 90 to 80% utilisation versus extended elapse time for a given feature. All elements of value must be converted into comparable economic units of measure, e.g. profitability: If all goes according to plan, how much profit will the feature generate? If we are 60 days late, how would that hamper profitability?
But how accurately do we measure items today? Developers should be confronted with the economic impact of a delay. Developers› guesses vary wildly from 10:1 to infinity (when no income is to be made any more because a company may be out of business). Delays are gauged very differently. Hence it’s important to communicate the targets in common units of measurement.
Queues and their Impact on Economic Performance
Invisible and unmanaged queues are the root cause of poor economic performance in product development.
HP measures in bits on a hard drive, i.e. invisible inventory. Car manufacturers can easily determine their work in progress WIP. Invisible inventory is not managed easily. Wrongly, queues are rarely measured as a performance indicator.
In the telecom industry of 1909 it was crucial to learn how many phone lines were necessary for say 10’000 inhabitants. The question sparked Danish engineer Agner Krarup Erlang to come up with a queuing theory: The closer you get to full utilisation, the longer the queues get. 85 to 90% is the top utilisation a manufacturing site can achieve.
It’s actually unproductive to overload the engineering workforce since the costs of queues will go up quickly.
Product Development and Economic Impact
Rather than trying to eliminate product development variability, we should try to alter it’s economic impact.
Innovation is one of the key drivers of many businesses. But it also comes with uncertainty of outcome. If you take risk out of the process you may as well not move at all.
Stock prices, options, probability to loose money and an asymmetric payment function offer interesting models that can be adapted to product management.
In product development we have asymmetric payment functions, e.g. Boeing has to figure out whether an aluminium-lithium alloy will help their business or not. The cost for research is way smaller than the potentially high payoff area.
In manufacturing the Taguchi function says that you are losing money as soon as you deviate from the optimum.
Queue and Batch Size
The fastest and easiest way to reduce queues is to reduce the batch size in our product development process.
Reducing the batch size is the most effective measure to take as a product manager. It will not only shorten cycle times but give engineers feedback earlier in the process. Bad assumptions are eliminated out of the equation. There are fewer defects but more meetings. If you do the meetings on a scheduled basis (as promoted in Scrum) the overhead is limited and the added communication becomes helpful along the planning process.
Smaller batch sizes avoid the ‹elephant going through the boa restrictor› phenomenon: Don’t overload the system at all gates involved.
Coming up with smaller batches allows for batch planning. Critical batches should be tackled first in the sequence. Also venture capitalists cut their investments into smaller chunks in order to limit risks.
Process cycle time and work in progress
The easiest way to control the cycle time of a process is to control the amount of work in progress in the process.
The waiting time in the queue equals to the length of the queue divided by the process rate.
It’s important to understand that these values take averages into account.
When the queue is full you may close for new input as in ’shut the door› flow control. But first in first out doesn’t work for all businesses, either, say a hospital emergency room. The cost of delay has to be considered along the way in lean product management.
Synchronised Cadence vs. Asynchronous Processing
A synchronised cadence offers performance advantages compared to processing asynchronously on demand.
Microprocessors used to trigger based on demand in the beginning of computer hardware design. Eventually they changed to periodic polling for more efficient processing. The same is true for email. It can be processed every time an email comes in or based on a schedule. Minimising context switching costs is the key here.
Sequencing of Work Products based on Economics
At both the project and portfolio level, the sequencing of work should be based on economics.
Computer operating system design deals a lot with queue management. Constraints are an important concept, e.g. time-out features.
Fast Feedback Loops for Better Performance
In the presence of uncertainty fast feedback loops permit us to achieve better economic performance.
Truncating unproductive paths early is important not only in the project but also at the bottom line. Speed of feedback should be measured and controlled.
Donald Reinertsen
Twitter: @dreinertsen
www.ReinertsenAssociates.com
http://twitter.com/#!/WalterSchaerer/status/93021456312705024
Swiss Limited WIP Society’s next briefing: David J. Anderson, ‹inventor› of Kanban, and Klaus Leopold will give a training in Zurich on September 15 – 16 2011. See their kanban training website for more details.
Ein vorbereitendes Kanban Kickoff Training wird Klaus Leopold am 9. September auf Deutsch geben.
