The Starfish and the Spider
O. Brafman & R.A. Beckstrom
This is a book about decentralized human systems and how they differ from centralized systems. Centralized systems struggle to dominate decentralized systems, which can be more robust, agile and innovative than would be expected. The advent of the internet has created a new venue, in which decentralized systems can emerge and the consequences of this are not yet clear.
The book provides a number of historical examples to illustrate the nature and impact of decentralized systems – and three deserve description here.
In the early 1500s, the Spanish arrived in Mexico and encountered the Aztec civilization. The Spanish were well organized with a military hierarchy. The Aztecs were also well organized with a strong hierarchy. The Spanish found their way to the capital, met the Emperor and basically said “give up or die”. He gave up, was killed, and the Spanish had conquered Mexico. They pulled the same trick with the Inca. In the face of overwhelming force, the Spanish took over by decapitating rival organizations. Hearing rumors of gold to the north, the Spanish worked their way towards modern day United States until they reach modern day northern Mexico where they encountered the Apache Indians. Using their usual tactics, the Spanish attempted to enlist the Indians as farmers or to overcome and enslave them. The Apache escaped and began to resist. The Spanish began to raid villages, killing leaders, and trying to intimidate the Apache. The Apache regrouped, spread out, and began to raid Spanish towns. Over time, the Apache drove the Spanish out of northern Mexico. The Apache did not have chiefs, though they had leaders. There was no central coordination. If some Apaches wanted to go on a raid, they went on a raid. They were everywhere, all the time. The Apaches started out decentralized, and under pressure decentralized further. Different groups could try different tactics and share information with other groups. The Apache goal was not to conquer the Spanish, but to avoid being conquered. They succeeded for centuries.
Shawn Fanning wanted free music, so he wrote some software that allowed people to upload their music to a server and download other music – and Napster was born. The music industry soon understood the threat of Napster to their profitability and took Napster to court. Eventually, the US Supreme Court ruled that Napster violated the rights of the music industry and shut Napster down. Almost immediately, Kazaa and eDonkey appeared. These did not collect music on a server, but put people directly in contact with each other to swap files. This was much harder to break up than Napster, but the music industry made it very hard for Kazaa to make money and eventually made it irrelevant. Hackers modified Kazaa into K+, and got rid of the advertizing that supported Kazaa – making it even more decentralized. Almost immediately, innovation led to an even more decentralized form of file sharing appeared called eMule. Nobody knows who started eMule or where it came from. A key feature of eMule is that it is open source. Now almost anyone could improve the file sharing software and there was nobody in charge of evolution from then on.
Bill Wilson was told he had 6 months to live unless he quit drinking. He had seen experts before, but they had not helped him. Wilson realized that the people best suited to help him were other alcoholics, so he founded Alcoholic Anonymous. Nobody is in charge of AA. Anybody can start another AA group. AA has the 12 steps, and a number of other norms – but they do not have rules or hierarchy. Leaders lead by example – not appointment or election.
Starfish and spiders look similar. There is a central body and a bunch of legs. But there is a big difference. Cut off a leg or two from a spider and it will probably die. Cut a leg or two off a starfish and you will have two or three starfish. A spider has a small head (compared to the body), and if you cut off the spider’s head – it dies. Centralized organizations are like the spider. The parts depend on the center to tell them what to do. Centralized systems are coercive. Central authorities tell you what to do – and you do it or else. Decentralized organizations are like the starfish. Starfish can act as a single organism, but each leg is autonomous. Decentralized systems are open. There are no central authorities, nobody issues orders, and nobody must follow them. The central story of the book is that new technology will enable decentralized organizations to form and out-compete centralized organizations under a variety of circumstances. In effect, decentralized file sharing totally changed the economics of the music industry. Decentralized software development has already displaced most “commercial” software from servers. Understanding the nature of decentralized organizations is increasingly important.
The book makes some generalizations about decentralized systems including the following.
- When
attacked, a decentralized system becomes even more open and decentralized*.
- Decentralized systems have an “ideology” that participants defend by decentralizing further.
- It is
easy to mistake starfish for spiders.
- Some decentralized organizations resemble centralized organizations. You need to understand the interactions between people.
- An open system doesn’t have central intelligence; the intelligence is spread throughout the system.
- Open systems can easily mutate.
- The
decentralized system sneaks up on you.
- It is easy to overlook a decentralized system and mistake its reach and growth rate.
- As industries decentralize, overall profits decrease.
- Put people into an open system and they’ll automatically want to contribute.
- When
attacked, centralized organizations tend to become more centralized.
- Centralized systems also have an “ideology” that participants defend by centralizing further.
Decentralized systems do have structure. A key point is to understand that every organization has an ideology that binds it together. Many businesses have ideologies related to efficiency, risk taking, interpersonal behavior, and making money. Charities have ideology around service. The ideology is what usually attracts people to the organization and provides the underpinning of the organization’s actions. The ideology of the Apache might have been related to independence, while that of Alcoholics Anonymous might be help people save themselves. The ideology of Ebay might be help people trust that online sales with strangers are safe. The most important aspect of a decentralized system is the ideology. This is what the organizations defend when challenged by another system.
Stable decentralized systems have people in a number of roles. One group of people might be called catalysts. The role of catalysts is to organize or connect people who already are inspired around a topic. In many ways, catalysts are leaders, but they rarely provide obvious direction. They often catalyze the formation of a group then leave the group to carry on. A second role is that of champion. The champion is the person who expands the network. Small networks may not contain a champion by large networks do. These peoples are often extroverted, high-energy people who spread the message and advocate for the ideology of the network. They look more like the conventional concept of a leader, but they provide less organizing or connecting than catalysts. Champions gather recruits for catalysts. Generally, catalysts are active in a “movement” before champions get involved, and usually less prominent. The women’s suffrage movement in the United States is associated with Susan B. Anthony – not with Elizabeth Stanton who began organizing three years before Anthony became involved.
The structure of decentralized organizations is associated with circles. This is a group of people who have a common interest in the ideology (perhaps communicated by the champion) and who self-organize (with help from a catalyst) in support of the ideology. These circles are relatively flat and leadership can be flexible. Often, one circle can grow, break apart and retain essentially all its functionality. Catalysts and champions help circles link together, but they are not the links themselves. These links might be strong or weak, but they tend to be disintermediated. Circles are bound by norms rather than rules. In other words, circles are controlled by culture. Different circles may develop different norms, while maintaining the overall ideology. AA meetings differ from group to group, but are all identifiably AA meetings.
The final structural aspect of decentralized organizations is that they often form of the backbone of a pre-existing network. The abolitionist and women’s suffrage movements began on the backbone of the Quaker church. The Apache had a pre-existing loose tribal structure based on family groups. But the Quaker church did not take up abolition as an organization for some time. However, people in the church mobilized other people in the church to join and spread the ideology.
Perhaps the most unique and important role is the catalyst. Study of catalysts suggest that the following “tools” are frequently used to influence and connect people.
- Genuine
interest in people
- Catalysts really listen and demonstrate their interest in people
- Loose
connections
- Catalysts have a wide range of loose connections – not a small group of strong connections. They are comfortable connecting people that they barely know.
- Mapping
- They are constantly trying to understand how people are connected. Did they go to the same school, both know Bob T., work at the same company, or live in the same city as anybody else that they know. These mental maps can be very extensive and detailed. Some people keep records of this to help their memories.
- Desire to help
- Meet
people where they are
- Catalysts don’t give advice. Advice implies superior knowledge or power. They lead by example and show people a possible path forward. They do not push people onto a path, but show them a path they can use.
- Emotional Intelligence
- Trust
- Inspiration
- Tolerance
for ambiguity
- The nature of a decentralized organization is less structured and predictable. Individuals and groups are not tightly coordinated, so what’s going on can be unclear for sustained periods. Catalysts do not fight this – they are not controlling the action but catalyzing it.
- Hands off approach
- Receding
- When the group forms itself, they do not need the catalyst, so the catalyst withdraws. The catalyst is not a role for people who seek credit.
All of the preceding discussion would suggest that decentralized organizations will always overcome centralized organizations, but this is not true. Three approaches have proven success.
- Given the importance of ideology, if a centralized organization can replace/displace the decentralized organization’s ideology, the decentralized organization will disband. This is the idea behind “hearts-and-minds” campaigns and most business change initiatives. Replace “old” norms with new, and you destroy the old structure. Done right, you create an improved structure.
- Get the decentralized organization to centralize. The Apache were “defeated” by giving some of them cows. The valuable cows gave some members of the organization property rights which disrupted their independence ideology with a wealth ideology. People with cows wanted to keep them, and people without cows wanted to get them.
- Decentralizing your own organization can remove much of the tension with a decentralized organization. IBM was originally a proponent of proprietary software, but switched to open-source software. This created a contrast with Microsoft’s centralized approach and made IBM less vulnerable to disruption on the software dimension. In fact, they assigned a couple of hundred software engineers to work on the software to reinforce their commitment.
Centralized organizations are very successful in situations where efficiency, coordination and complication must be managed. Decentralized organizations are very successful when adaptability, innovation and complexity must be managed. Each type of organization can develop an ideology that conflicts with the other. This probably has negative effects on both. In fact, there are a wide range of hybrid organizations. Ebay has a set of centralized systems and procedures that enable buyers and sellers to operate very independently. Techniques like “appreciative inquiry” can help alter vertical structures by creating personal bonds between people at different levels of the hierarchy. Toyota has a centralized operation with a strong brand and production ideology (just-in-time, Lean, Kaizen, etc). It also has a strong commitment to decentralized production groups where work groups (quality circles) control their own production. The book notes that workers are asked for suggestions about improving production and 100% of suggestions are accepted – because the ideas are accepted by their peers. If the idea does not work, the team replaces it with another idea. The workers know that they have authority and responsibility. By combining structures related to quality and inventory control with decentralization about process, worker creativity is allowed to fuel continuous improvement without much direction from management.
If the world comprises both centralized and decentralized organizations, the following observations might almost be thought of as the “new rules”
- There are now significant diseconomies of scale. Small flexible organizations can deliver services quicker and easier that highly centralized organizations. Their ability to change direction in the face of changing customer desires may out maneuver more conservative “brands”.
- Creating a network can be hard, but it is getting easier with modern communication technology. These loose connections increasingly influence customer decisions.
- The environment is increasingly chaotic and unpredictable. Companies must be more agile to adapt to this chaos, because it can be predicted.
- The most important information in an organization may be found at the edge of the organization. Successful organizations can access and use that information, and not discount it because it is from the edge.
- Given a chance, everyone wants to contribute. Successful organizations create space for people to contribute in their “own” way.
- Successful organizations will not attack decentralized organizations because they understand that they risk creating even more diffuse decentralized competitors. This is different from changing their own practices or the incentives for people, which may be very useful and effective.
- Catalysts dominate decentralized organizations but not because they control them. Successful organizations will recognize the catalyst role and not give them formal leader roles. They will allow the catalysts to connect and inspire, then get out of the way.
- Organizations will use ideology to instill the right balance of centralized control and decentralized ambiguity to be adaptable without being uncoordinated.
- Measurement will change from detailed, precise metrics to directional metrics for many activities because it is better to be vaguely right than precisely wrong.
- More organizations will become flatter and more decentralized, at a minimum forming hybrid organizations that combine both centralized and decentralized structures.
Hybrid structures are quite common in businesses, but companies may act as if they are centralized on some matters. The book highlights a number of cases where centralized thinking at a company level cost the company in the long run. One of the bigger unstated implications of this book is that some decentralization within an organization is necessary but not sufficient to deal with a world containing decentralized organizations. Some amount of a company’s outward facing behavior needs to have a “decentralized mindset” to adapt to the chaotic real world.
My comments and interpretation:
- This book was written before social media became commonplace. It was before the Arab spring, where protestors used social media to organize spontaneous demonstrations. It was before companies began to introduce social media into their workplaces to improve connectivity. Decentralized organizations determine the fate of restaurants (Yelp), dispose of goods (Craigslist) and supply our facts for almost any subject (Wikipedia). A decentralized group created the software that runs many servers (Apache) and computers (Linux). The full effect of decentralized emergent social groups will not be obvious for some time, but the effects are already accumulating.
- By the 1990s, neuroscientists realized that memories were not stored in cells, but somehow in the interactions between cells. Some pattern of interactions comprised a memory. I wonder if an organization’s capability is essentially the same thing. It is not what people “know”, but the interactions between them that matter. In the brain, scientist realized that this approach created a more resilient memory. The death of a single cell did not eliminate a memory because the memory was many interactions instead. Similarly, most organizations maintain their capabilities over time despite the turnover of people.
- The book mentions analysis that shows that the internet maintains almost all of its function even when 95% of the nodes in the network are disabled. The integrity of the system does not depend on a few vulnerable nodes. Consequently, even a weak network may be more robust and agile than a good centralized network.
- If you think of a culture being a network of ideas, practices, interactions and the like, it is easy to see that a culture could resist rapid change. Most organizations have a mix of centralized and decentralized structure. According to this book, the centralized structure may be quite effective in many ways, but it can’t compete with the decentralized structure also present. Centralized control will succeed when it aligns with the decentralized culture, and fail when it works at cross purposes. It seems to me that the opposite won’t happen though.
- I have an uncle who became very involved with Alcoholics Anonymous. It was an important part of his life and he was very involved as a sponsor in the organization. When he died, many people came to see us as we were attending to his affairs to express their thanks to us for his effort. He had left us some money and we thought that we would make a donation on his behalf. Nobody could tell us where to send the money or what they would do with it. A charity that won’t take you money is a different kind of organization.
- When I began my career in business, management gurus of the day advocated employee empowerment and organizational flattening. This was amplified through the development of the web-based economy. Since 2000, more traditional companies have been returning to more centralized control. This book observes that each extreme creates opportunities and issues. I think it is interesting that the extremes are described as ideologies. What this illuminates is that they have quite a bit of hidden meaning for people. In 1917, the communists had defeated the republicans are established the Soviet Union. While western countries were in the early stages of deployment of telephone networks – which supported a highly decentralized communication network – the Soviets deployed a network of loudspeakers. Anytime the government wanted to make an announcement, they employed their network. It did not allow the government to listen (that was not a priority) or people to communicate with each other (another non-priority). It allowed the Soviets to exert control. It does not take much effort to find business articles that emphasize control or much more effort to find articles that complain about lack of organizational flexibility. As ideologies go neither extreme meets the needs of modern companies and both ideologies may need to be displaced.
*text in italics is directly quoted from the book
Recent Comments