Thinking about decentralized systems means thinking about how an organization can agree upon what it holds to be true and how this relates to the actual ground truth.
A single source of truth is a concept from information science, but it’s so universal that it can be applied to a whole variety of scenarios and organizations.
It means that whenever participants disagree about the state of the world, or a particular aspect of it, a designated participant gets to decide which state is the true state.
Note that this concept uses the word truth not as “what is true” (that would be the ground truth), but as “what is held to be true” by one participant — the source — and in turn accepted by all others.
It is, in a word, a key building block of hierarchy and a defining feature of centralized organization. Having a single source of truth has many benefits. The most obvious one is that it quickly resolves disputes and avoids interminable disagreement.
But that’s only a benefit if we can be sure that the truth the single source offers is close enough to the ground truth most of the time. So it has to be both infallible and honest.
If that’s not the case, and one can easily think of many scenarios where it isn’t, the drawbacks of such a setup become quickly apparent. We don’t want anyone in an organization to be incompetent or lying, and certainly not the boss.
There are two clear alternatives to such a centralized setup. The first is many truths coming from many sources of truth, leading to inevitable goal conflict.
The second is what we call a single truth without a single source.
To us, this is the shortest definition of a decentralized system. The reason why we have been paying close attention to decentralized systems for the last years should be clear from this.
An organization with a dysfunctional single source of truth cannot succeed. An organization that cannot agree on a shared truth is not an organization.
In other words, an organization has to come to a consensus, and it has to come to the right consensus. Where truth doesn’t diverge from ground truth.
A single truth without a single source then becomes an essential building block for a whole range of organizational and information architectures. But what are they, and how can we make it work?
The deceptive allure of centralized systems
Centralization of command — bureaucratized hierarchy — has a long track record as an organizational form that gets things done. It’s hard to imagine building large-scale organizations without it.
But then again, it’s also a popular target for trenchant criticism. And this is where we should start when we want to think about where building decentralized organizations makes the most sense.
Centralization has a powerful quality of being able to cut off discussion. This might not sound intuitive, but it should be readily apparent to anyone who has ever sat through interminable strategy meetings.
Distilling the available information into the right strategy is important, but when faced with the choice between the wrong strategy or endless discussion, picking any action tends to be better than no action at all.
Relegating this task to the participant assumed to be most knowledgeable and most qualified — in one word, the most senior — is an obvious next step. It is the underlying idea of all hierarchical systems.
But there are a whole number of drawbacks. The most senior participant might simply not be the source that’s closest to the truth — a common scenario whenever the ground truth is literally on the ground rather than in the boardrooms — or the most senior member might have very selfish reasons to distort the truth, and all information flow that is connected to it.
Designing decentralized systems
Designing decentralized systems means creating information architectures that mitigate these single source of truth problems. In other words, they have to make sure all decisions are made by all participants, and all participants can access the same information.
In other words, a decentralized system is one that is not owned and not controlled by an individual participant.
This is of course tricky to pull off since organizations have a tendency to gravitate towards centralized hierarchy. Decentralization is costly and cumbersome.
So this additional effort has to produce an additional benefit for everyone involved. This benefit has to come in the form of improved access to information for everyone, improved guarantees that this information is timely and accurate — the ground truth — and not distorted before or after by a single participant or a coalition with shared interests.
And it has to come in the form of all participants being able to converge on the same shared perspective — the consensus — that is closer to the ground truth than a centralized system can guarantee.
In short, a decentralized system has to be able to quickly resolve disputes about what is the truth, and limit the influence of a single participant on the accepted truth to be useful to everyone.
This is why most real-world organizations are hybrid forms. Managerial hierarchies to resolve the decision paralysis problem; supervisory boards to resolve the “going off in the wrong direction” problem.
Using both tools correctly is the secret to organizational design.