We’ve heard them so much they’ve become clichés: “Fail fast; fail often!” “Failures are gifts.” “Leaders must learn to tolerate failure.” “There’s no shame in failure, the shame is when we fail to learn from our failures.” These are powerful statements that urge new attitudes and new ways of operating in improving organizations. But the stakes are high in the world of practice — especially for leaders. The new perspective on failure cannot mean we actually want lots of it, can it?

Surely there must be more thoughtful guidance to be offered about when, how, and what makes failure acceptable, even productive. After all, if learning and improving from failure was inevitable, then the world would be a much better place. Indeed, there is a real need for thoughtful guidance, and we would do well to be clear about the circumstances that make failure acceptable and productive.

We see three conditions that are necessary for failure to actually enable improvement, thus three conditions under which failure is acceptable. We identify these three with brief exposition here.

“Of course!” one says. “We should always be working on things that matter.” But what this means in an improvement context is quite particular. What we are trying to do when we fail is important. Trying to do something that “seems like a good idea” is good, but not enough.

How, then, can leaders select a matter or initiative on which to focus from the many competing possibilities begging for their attention? Improvement science would urge that the initiative have three key attributes.

First, it should be something with a meaningful place in our theory of improvement. As part of the chartering of a Networked Improvement Community (NIC), we study the problem that the NIC focuses upon; we study the system that gives rise to the problem; we review available research and evidence (from both scholarship and practice); and we develop a theory of improvement (often represented as a driver diagram) addressing the NIC’s problem. This driver diagram articulates a coherent theory of action that identifies the high leverage drivers that should improve performance against intended aims. It also ensures the coordination of collective effort across an often widely dispersed network. Working appropriately on drivers in the shared theory of improvement is working on things that (given the best available wisdom, research, and evidence) are thought to maximize performance against the aims. It means working on things for which others in the NIC are awaiting our contributions — just as we are awaiting theirs. That’s working on things that matter.

Second, we should prototype and test ideas that are consistent with our design principles. Design principles are the result of close study of the scholarly and practice-based evidence relevant to a given problem. Collectively they articulate the NIC’s best guidance regarding the identification of change ideas that are consistent with its beliefs about effective practice. They ensure that the invitation to adapt and innovate is not an invitation to chaos. Working on ideas that fit within the conceptual space to which the NIC is committed is working on things that matter.

Third, we should test ideas that are informed by the knowledge about reliable and effective implementation generated within the NIC. This includes knowledge derived from successful tests and from failures, as both are instructive of how best to implement ideas reliably, effectively, and at scale. When we work on adding to the knowledge accumulating within the NIC about how to implement well, we are working on things that matter.

Perhaps the single biggest problem with failure is that it can cost so much. Material investment (time, money, etc.) and political capital are two big types of cost, but there are others. Opportunity costs — the investment of considerable attention in large scale implementation of one solution to the ignorance of others — can be an issue when they do not pan out. Doing real and irrevocable harm is another formidable potential cost, especially in educational contexts in which the costs of failure can have dire consequences for the professional lives of teachers and the life chances of students. Maybe most debilitating of all is the loss of faith in the system (and for that matter the loss of the will to improve) that comes when energies are repeatedly dedicated to improvement efforts with which we become disenchanted and ultimately abandon.

The cost of failure is an important consideration involving three factors that should influence our determination of how large or small our implementation efforts should be: 1) our confidence in the programmatic ideas themselves; 2) how ready the field is to implement them well; and 3) the cost of not succeeding with the implementation. Only where our confidence is high, readiness is great, and the costs acceptable should we consider implementing broadly. As questions and concerns arise about how exactly the programmatic expression should work, or whether we are well prepared (especially in terms of knowledge, skills, or dispositions) to implement it, we would be wise to start small and learn our way into implementing well. Practically speaking, these uncertainties and doubts describe virtually all efforts at change to some degree.

Implicit in improvement research are methodologies wherein the learning process begins by starting with small tests of change. Obviously small tests mean greatly reduced costs of failure at the very time that the risk is greatest. Then we look to spread improvement ideas only as our confidence in the ideas themselves, our knowledge about how to implement well, and our ability to do so grow commensurately with our efforts to scale. Simply put, improvement research, by its very nature, manages the cost of failure as it readies both people and ideas for successful implementation reliably and at scale. While doing so, it also protects those involved from the inevitable risks that come with attempting to solve persistent problems and exploring new paths. With that, failures can be learning opportunities, and not material, human, or political disasters.

All that is said of failures is true. They have much to teach us about improving, so much so that old-timers refer to this as the “scar tissue” that comes with experience, and when it is referred to it is almost always referred to in the context of learning. But learning from failures requires more than a promise to ourselves and a lot of hope for the best. It requires that we bring to bear the right culture and dispositions as well tools and routines.

To ensure that we do learn from our failures, the first order of business is to promote a culture of trust, openness, and teamwork. This is something that only the leadership of an organization can do, as only the leaders can manifest and model a set of values to establish the proper tone and culture. Just as no one individual can know everything about a system, no one individual can understand all possible root causes of a failure in that system. To learn from failure, all must feel invited, welcomed, and safe in expressing their views of a failure.

Beyond the interpersonal and cultural conditions that permit us to learn from failure, there are tools, procedures, and social routines that discipline the learning. They involve all who know about the problem and care about the organization and its performance; they discipline the capture of information about the failure, its causes, and possible corrective solutions to test; they establish plans for testing proposed improvements that are intended to avoid the failure in the future; and they do all this to depersonalize matters so as to focus on the problem — not the blame.

All of the preceding is predicated on leaders themselves developing the knowledge, skills, and dispositions needed to cultivate, manage, and leverage “learning organizations” and “learning systems.” Doing so will require that leaders construct their practice (and learn from their own inevitable failures) in the same ways that they manage their own enterprises: by focusing their energy on leadership practices that matter in their current context; by managing the costs to themselves and others as they experiment (and solve problems) in their own practice; and by cultivating effective learning communities that feature the trust, openness, and teamwork that supports collegial growth.

If we fail while working on improvements that matter; if we manage the cost of failure such that it is not debilitating to the organization when it occurs; and if we put into place the psychological, social, and procedural conditions that allow us to learn from our failures, then we maximize the likelihood of learning the very things that enhance organizational performance for good. Now that is acceptable failure.