How we make decentralized decisions
in a self-organized company
Decision-making is a — if not the — core discipline of any organization. If we want to make progress, we need to make decisions. Decision-making is about aligning our actions with our goals. Enabling decentralized decision-making is one of the main reasons we structured our company as a network of autonomous teams. Our intention is to move decision-making authority to the information and enable those that feel a sense of urgency to take the initiative. This avoids decision bottlenecks and allows for decisions to be made quickly where they are needed.
We’ll answer common questions like: Can anyone make any decision? How can we be inclusive without getting stuck in endless discussions? Who do I need to consult before making a decision? Can I no longer make decisions without consulting others first? How do we proceed when we can’t agree as a group?
To clarify what we mean when talking about a decision to be made, we use the 7 levels of delegation as categories. We specify who makes the decision (!) and if anybody needs to be asked (?) beforehand. We use different practices on levels (2) Sell, (3) Consult, and (4) Agree in different situations as described below.
Why we don’t use coercive power (1)
Management, as invented 100 years ago by Taylor, has served us well in the industrial age. The division between thinking by a few on the top that tell the many on the bottom what to do, afforded us massive efficiency gains. However, when applied to knowledge work instead of physical work, these advantages mostly disappear and are counteracted by disengaged knowledge workers that get paid to think.
The coercive power hierarchy operates through delayed or broken feedback loops that prevent decision-makers from experiencing and personally feeling the impact of their decisions. This leads to slow and ill-informed decisions, that are too static for markets that keep getting more dynamic and complex.
We avoid these drawbacks by distributing decision-making to autonomous cross-functional teams instead of functional departments. When distributing decision-making, we consider these two principles.
- People who have to carry out a decision, make or delegate the decision
- People who are impacted by a decision can give input for it
This structure and these principles promote the following ways of making decisions.
We aim for a decision distribution of roughly
- 90% by individuals: safe-to-fail and mandate,
- 9% by groups: group consent process,
- 1% by individuals deciding for groups: advice process
The remainder of this post explains how and when we use each of these decision-making practices.
Deciding yourself (2) when it is safe to fail
When faced with a decision, we first consider reversibility and consequence before escalating to another decision-making practice. For decisions that are easily reversible or have little consequence, we choose to make decisions on our own. Many operational one-off decisions fall into this category. In these cases, “asking for forgiveness, instead of permission” is the best option because it is fast and creates momentum.
Decisions that are irreversible but have little consequence are a good training ground for learning. Making mistakes here is cheap, and the lessons learned can be used to avoid more costly mistakes. For instance, you may learn who has a strong opinion and valuable input for hardware after purchasing an adapter (50€) on your own. Now that you know, you may consult them, before making more a significant purchase, like a laptop (2000€).
Decisions that have big consequences but are easily reversible are great for experimenting. You may revert quickly after gathering evidence that your decision was a poor choice. Congratulations. Now you can argue about facts instead of opinions. You may, for instance, choose to do a project with a candidate as a freelancer before making them an offer as a full-time employee.
When decisions are both (highly) consequential and (nearly) irreversible, we choose to make them carefully and methodically by using one of the following practices.
Deciding yourself (2) with a mandate
The simplest case for making decisions is when you have a mandate to do so — either given by a group you are deciding for or by a role you fulfill. Most operational case-by-case decisions fall into this category.
We avoid creating role descriptions that are detailed checklists of specific tasks because we think they lead to “dumb” behavior along the lines of “Not my fault. I just followed the rules.” However, we find it very helpful to make the mandates of roles explicit with high-level decision areas that enable leadership and accountability.
Example: This could be you in your role as a recruiter rejecting a candidate after a first screening. You wouldn’t inform anyone about this decision you make daily even though it is not easily reversible and has a significant consequence. If someone asks you about a specific decision, we expect you to explain your reasoning — to sell your decision, if you will. This helps others understand and accept your choices, as well as enabling them to learn from them.
Agreeing (4) in a group with consent
When there is no mandate for someone to make the decision, we default to a group decision by the affected group using the group consent process. We prefer consent over majority vote or consensus. While consensus is commonly interpreted as everyone is for a proposal, consent is about no-one is against a proposal. At first glance, this may sound like the same thing and a mere play with semantics. But only until you realize that next to “I prefer” and “I object” people have a third position: “I tolerate”.
In a group, “No-one objects” (consent) is equal to “We tolerate”. “We tolerate” leaves a much bigger solution space than “We prefer” (consensus).
Making positions explicit and visual helps discussions to move forward. It is much easier to go with someone else’s position when I feel like I was heard and understood.
For consent to work well, it is essential to understand what a reasoned objection is — and what it is not. A reasoned objection is not “I’d prefer another option.” but “I think, doing this will harm us and risk achieving our goal.” Objections are not vetos that stop a proposal, but identified risks to be integrated into an amended proposal. They are the start of a discussion, not the end of it. When we phrase objections as suggestions for how to alter the proposal to make it acceptable, discussions become a lot more solution-focused.
Sociocracy uses the slogan “good enough for now, safe enough to try” to describe the experimental nature of consent decision-making. With that, consent tends to reach quicker agreements and leads to smaller experiments.
Example: Teams use the group consent process to decide on things that affect everyone in the team, like their goals for the next three months. Most long-term strategic and policy decisions on team level fall into this category.
If we can’t agree as a group we escalate: We choose a decision-maker that follows the advice process. Choosing a decision-maker tends to be as simple as someone suggesting a person from the group, and the rest of the group agrees. Technically, this is another (meta) consent decision, but it is usually much easier to make.
Decide yourself after consulting others (3) with the advice process
For cases that are not covered above, we default to the advice process. We do this for decisions in big groups where we deem group consent too inefficient or after we tried and failed to reach consent within a group.
We draft proposals in a shared place (e.g. a wiki or a file-sharing drive) where people can see the progress and comment throughout the decision-making process. We use a template to communicate our intention to make a decision on a topic. With a message to the affected group, the initiator proposes who is deciding what by when and why this is important now. This makes decision-making transparent and enables the affected group to contribute.
We expect decision-makers to consult experts and those affected actively. It’s not enough to send out a proposal and passively wait for feedback. This has a subtle side effect: Asking for someone’s advice is a way to show appreciation — to show that their opinion is valuable to find a great solution.
The greater the impact of a decision, the more people we consult. Advice received must be taken into consideration. But advice is simply advice. Ownership of the issue stays clearly with one person: the decision-maker.
Decision-makers make sure everybody is committed to the decision, including those that disagree. They talk to them before they publicly communicate their decision. They explain that they heard them, that they understand their perspective, and why despite that they decided differently. There is no need to make watered-down decisions to please everyone. People respect decisions if they feel they were heard and understood.
If we disagree with a decision, we practice forgiveness and bring it up with the decision-maker. We follow the conflict resolution process if necessary.
Example: We use the advice process for decisions that make changes to our organizational structure, like creating a “bench” group for people that are temporarily without a team, or decisions that affect people across the company, like switching to a new recruitment tool.
Who makes decisions with the advice process?
To be clear: We don’t have dedicated decision-makers. You become a decision-maker by identifying an important and urgent decision to be made and taking initiative or by being chosen by a group of people unable to decide something with the group consent process.
Decision-making is not evenly distributed among everyone, and some voices carry more weight than others. People with more social capital influence decisions more than others. Any group of people has hierarchies of reputation and merit. An even distribution of decision-making was never the goal. The goal is inclusion and openness and to enable those that feel a sense of urgency to take initiative.
Expertise in a particular area often influences who makes a decision, but it’s not the sole criteria that makes someone a good fit for decision-making. An understanding of all perspectives, the skill to thoughtfully weigh options, or social capital with the affected group may be more important in some circumstances.
Traps to avoid
Decision-makers need to be willing to question their initial assumptions and ideas for a solution and zoom back out to the problem space. This is difficult because our brains function with natural biases. We tend to pick sources of information that are not representative (selection bias). And to make matters worse, we tend to distort the information we receive to seek confirmatory evidence instead of trying to disprove our assumptions (confirmation bias). We cannot entirely avoid these biases, but being aware of these biases and the difference between problem and solution space helps.
The advice process can initially feel uncomfortable. While it’s easy to appreciate intellectually, it is emotionally challenging. It requires separating oneself from one‘s ego’s attachment to being right, and instead, trying to change perspective to see what one has a hard time seeing.
Adoption pains to overcome
The adoption of decentralized decision-making doesn’t happen overnight after making the switch from a coercive power hierarchy with supervisors to autonomous teams. Three phenomena that we experienced were permission seeking, slow decision-making, and resistance.
It has been some effort to get rid of the reflex to seek permission. It helps to have leaders that resist the urge to “just make the call” when someone “gives” them a decision to make. Instead, these are teachable moments to ask: “What do you need to make this decision yourself? Who’s going to be affected? Who’s an expert?” This, in turn, requires trust in the positive intention and the competence of each of us to be able to let go of the decision-making power. This is difficult because uncertainty feels uncomfortable — more so than being wrong.
Another source of friction is the speed of decision-making with the advice process. Consulting people takes time. Considering their advice takes time. Making the process transparent takes time. Consequently, more time passes between the intent to make a decision and actually making it. However, decision-making is a means to end. We want something to happen after we make a decision. If part of the desired effect is that people change their behavior, it is overall more effective to include them in the decision-making process than trying to convince them after the decision has already been made. Involvement creates commitment.
The built-in reflection time and diverse perspectives lead to more diverse proposals and ultimately to better decisions, that are more in line with our company values and the future we want to create rather than individual belief systems.
Resistance is a natural part of decision-making because people change at different rates. Collaborative decision-making makes this resistance more transparent, which often leads to frustration on the side of the decision-maker. However, making the resistance more visible makes it also easier to address and overcome.
If we want to make progress, we need to make decisions. By decentralizing decision-making we enable decisions — and hence progress — in any place in the organization while getting more engaged colleagues at the same time.
Decentralized decision-making is more effort compared to someone making the call because they are higher up in a power hierarchy. We know that. That’s the trade-off we make for enabling higher autonomy of teams and individuals, which ultimately leads to more innovative products and more fulfilled people. Being clear on what decision-making practice to use in which context helps us mitigate the drawbacks of this trade-off.
As with so many things, distributed decision-making is simple — just 2 principles and 4 practices — but not easy. For it to work well you also need to become at least decent at things like empathy, nonviolent communication, conflict resolution, mediation, leadership, responsibility, accountability, and personal time management.
In my mental models, I try to pragmatically synthesize what works in both traditional and new ways of working. Writing is both how I communicate my mental models and how I crystalize them. My writing is exploratory and evolving. I hope to revise this post over the next few months with newly gained clarity. To support this, I very much appreciate your feedback. Be it a critique, uncovered blind spots, different perspectives, or agreement. An earlier version of this post was originally published on blog.gini.net and here is a recording of a 30-minute talk I gave on the topic.