Move over Toyota. Software is leading the Lean world
Forty years ago, Toyota started eating its way through the auto industry with new tactics that we called “lean manufacturing”. Compared with the older American auto makers, they had higher quality, lower costs, and faster delivery of new products. For a generation these ideas have been spreading to help companies worldwide make things better, faster and cheaper. But the flow of ideas has reversed. Toyota is still playing an acoustic guitar from the 60’s, while software companies have amplified lean tactics to the point where they are rocking with 50,000 watts. Pull over, Toyota. Here comes a Tesla over-the-air update. Software is eating the world, and tech companies like Google and Amazon are leading the way with continuous delivery.
My “lean” education started in 1990 with an academic study of car development in Japan called “Product Development Performance”. Reading between the lines, I understood that parts suppliers would bring in new parts and bolt them to a prototype car. This idea directly inspired continuous integration in software. Other software developers were picking up a similar inspiration for what became “Kanban” continuous development.
Thus the monster received a spark of life. And the monster grew. Now, four of the five most valuable American companies are powered by continuous delivery — Amazon, Google, and Facebook, who have used this process from birth — and recent convert Microsoft. These companies are going global with advantages in scale and product cycle time that are similar to the advantages that helped Toyota bankrupt GM. They can produce a new product in months, not years, and they can make all of the little improvements that increase quality at the rate of thousands per day.
The rise of continuous delivery has a double impact on manufacturing. The first impact is inspiration. Just as agile software development got a boost from manufacturing ideas, manufacturing and all other production processes are picking up ideas from software. The second impact is very direct: More and more of what manufacturers do is software. More of the functionality of items like cars and even light bulbs is implemented in software. And, the production, sale, delivery, and servicing of these objects is controlled by software. And, as continuous delivery makes software development more productive, producers have an incentive to wrap more software around each person they hire and each piece of metal they bash.
Cruising to Continuous
It’s time to return the favor, bringing software continuous delivery to lean manufacturing, and a little bit of Silicon Valley to Tokyo. In 2015, I had the opportunity to go to Tokyo and present a full picture of American continuous delivery to the Japanese industrial establishment. Some of my slides are included here.
I learned some things from that 1990 book. For example, when development is faster, it is also more productive. Here is the high-tech chart:
Also, you should overlap your phases for development, testing, and production. Here’s the chart:
There are also lessons about dividing products into components and upgrading the components, producing in small batches, ramping up to improve quality, and not shutting down production lines.
In an ideal lean world, everything you do is about delivering value to the customer. In successful Japanese companies, the authors found “heavyweight” product managers who take responsibility for designing, building, and marketing a car. Their job matches the lean idea of a “value stream manager” who delivers a complete package of value. It stands in contrast to the less successful approach where a “lightweight” product manager acts as a coordinator for functions that deliver things like design, manufacturing, and marketing, without necessarily delivering value.
Continuous delivery amplifies these tactics.
Strong product managers: The leading digital companies use strong product managers who have the responsibility and the authority to create high-value products. They use all of the other tools — continuous change, testing, and measurement, to relentlessly find customer value.
Continuous product change: Change is the raw material for increasing value. A modern online service changes every day. Even if you are not producing an online service, you might be delivering banking services, or running a factory, or delivering parts, or doing a lot of other things that are actually using online services. And, the basic tactics of frequent change are seeping through into more and more types of physical products.
Improvement at the component level: The software architecture that goes with continuous delivery is called “Web services” or “microservices”. This is a technique for chopping up a big piece of software into smaller components. Each component is a computer or “Web service” that talks to other computers. Since these components are small, it’s easy to modify and test them. A company like Amazon has thousands of these components, each tended by a small team, and can release thousands of improvements every day.
Testing overlaps with everything: As engineers change components, they test them. This is called “unit testing”. If unit tests work, they immediately put their components into a bigger test system to see if the whole system will work. This is called “continuous integration.” There are actually many other layers of test, running up through selective customer trials (beta testing) and measurement of the modified product in use (monitoring). I call this “test layering.”
Measurement and big data: Every automated process and component and customer interaction produces data. This is a vast amount of data that we can use to figure out the efficiency of our operation, and the value of our products. This data is the necessary ingredient for product management — figuring out how to improve the product — as well as operational improvements.
This measurement provides a direct connection between engineers and customers. Engineers do not need to wait for a marketing person to report back on usage, or a sales person to report back on customer reactions. They can directly see the usage and comments in real time.
So, continuous delivery approaches the lean ideal where production turns immediately into measured value, without lags for storing stuff (inventory) or communication.
Organization
Top software companies also amplify some of the good effects of lean management on the organization. An organization that has continuous testing and measurement actually requires less management. Some management challenges can be automated away. A centralized executive doesn’t need to figure out what to do, because teams can take guidance from their test results and measurements. So, it’s less centralized, and it can adapt faster to changes.
Engineers are important, just as they are important in a lean manufacturing operation. Engineers products, and they also build automation that makes the product. And, they are directly connected to customers and customer data. It’s not an accident that a lot of the big new tech companies are run by engineers.
Product management becomes a critical skill. Product managers have to work fast, because they have to keep up with the engineers. And, they have to produce great products that always get better. Fortunately, they have data. They have to absorb a lot of data about what customers are doing. And, they have power to pull value out of the entire company. They work with front-end teams, and those teams talk to back-end teams, and this pulls components and skills from the whole company.
Strategy
Those product managers are not aiming for Toyota-style quality. They are optimizing something different. They are optimizing for usage, and value in use. It doesn’t help them to produce a high quality product with low usage. They can get more usage (and deliver more value) if they run experiments that sometimes look broken.
They are aiming to keep customers for a long time and maximize customer lifetime value. This means that if a customer throws out one iPhone or iDevice, you want them to get another. If they get out of one Chevrolet, you want them to get into another that’s optimized for them.
They use cloud services to make this connection. You throw away your phone or your car, but your cloud account lasts forever. Apple builds services around the appstore, and itunes, and backup, and migration, and voice recognition, so that customers are always connected and they can move smoothly from one device to another. If you are Ford, you sell a “mobility” account that outlasts a car and moves customers even if they don’t have access to a car. If you are Cisco you have to lift your network services out of individual boxes and provide an evolving, secure, connection with a rotating set of devices. If you are GE you sell a lifetime subscription to machinery with monitoring. All of these things happen with cloud services.
When you build cloud services, you move to continuous delivery. You also face the challenge of integrating all of your engineering groups that each make a great device or service. You have to put them into one big test system, so that your overall customer experience is continuously working and evolving, even while your product groups run their own release cycles. This is the challenge of “life cycle” and “omnichannel” and “360 customer experience” and “customer relationship management” and a multitude of other business buzzwords.
The ultimate effect of this tactic is astonishing. Consider the difference between the Toyota version of a car and the Google future version of a car. The Toyota version of a car is a shiny device delivered to a dealership. After you buy it, you deal with it while it gets older and less shiny. The Google future version of a car doesn’t have to get routed to a dealership, because they already know where you want to go and when you want to be picked up. My phone tells me “you should leave now to be at your meeting in 10 minutes,” and it won’t be long before it says “and your car is waiting”. They don’t have to sell maintenance, because if they detect a problem with the car, they will bring it back to a warehouse where parts are forward-cached in the exact amounts that will be statistically required. Other computers are ordering, building, and shipping the replacements — which contain continuous improvements.
Lean manufacturing ideas have supported a lot of consultants. Those days are done. I expect the top consultants in the field to move over to the study of continuous process. We’re already bringing some great management consultants over to our focus on “full stack” continuous delivery that includes strategy and organizational change.
