Process Debt
operational debt · undocumented process
The accumulated cost of doing things ad hoc. Like technical debt it is invisible until scale arrives, at which point every undocumented decision has to be made again by someone who was not there.
In practice
Six people onboard clients six ways. Nothing fails, and every new hire takes three months to become useful because the process exists only in conversations.
The common mistake
Waiting for a quiet period to write processes down. The quiet period does not arrive, and the debt is cheapest to pay while you are doing the work anyway.
Work done without a defined process is faster now and charges interest later. Like financial debt it is sometimes the right instrument and always has a rate, and like financial debt the trouble starts when nobody is tracking the balance.
The metaphor, and what it originally meant
Ward Cunningham coined technical debt in a 1992 experience report, and later objected to how it was being used. His metaphor was never about sloppy workCunningham, W. (1992). 'The WyCash Portfolio Management System.' OOPSLA experience report. In a 2009 clarification Cunningham stressed that the debt he meant was shipping code that reflects an incomplete understanding of the problem, then refactoring as understanding improves. Debt taken deliberately and repaid; not a licence for mess. — it described shipping something that reflects your current, incomplete understanding, and repaying by refactoring as understanding improves. The same correction applies here. Improvising a process the third time you do something is correct: you do not yet know what the process is, and writing it down prematurely encodes a guess. The debt begins when volume rises and the writing-down never happens.
What the interest is paid in
Fred Brooks (1975) supplied the arithmetic for the largest payment. Communication paths in a group of n people grow as n(n−1)/2, so when the process lives in people's heads rather than on paper, the coordination cost of an undocumented process rises quadratically with headcount while the output rises linearly. This is why the same ad-hoc approach that was efficient at three people is crippling at eight, with no single moment at which anything visibly broke.
Nonaka and Takeuchi (1995) describe the repayment mechanism: externalisation, the conversion of tacit knowledge into explicit statements. Their central observation is that the conversion is generative rather than merely clerical — writing a process down changes it, because articulating a decision exposes the parts that were never actually decided. The document is usually less valuable than the argument required to produce it.
The objections
Some process knowledge cannot be made explicit. Polanyi's (1966) result — we know more than we can tell — sets a ceiling that no amount of documentation effort clears, and the parts that resist articulation tend to be the judgement calls that most determine whether the work is any good. A written process can capture the sequence and lose the discrimination.
Worse, documentation can make things actively worse. Adler and Borys (1996) distinguished enabling from coercive formalisation: procedures written to help a competent person handle a non-routine case improve both performance and satisfaction, while procedures written to constrain them degrade both. The distinction is in who the document is addressed to, and it predicts why most process initiatives are resented. Process debt is therefore not simply retired by writing more down; written badly, it is refinanced at a higher rate.
What it rules out
It rules out reading the absence of failures as evidence the process is fine — the interest is paid in onboarding time and rework, not in incidents. It rules out waiting for a quiet period, which does not arrive and in which the knowledge is coldest. And it rules out headcount as the remedy, since Brooks's arithmetic means adding people to an undocumented process raises coordination cost faster than output.
It does not rule out improvising. Early improvisation is correct, and the discipline is not to document everything but to notice the transition — the second time someone asks how something is done, the debt has started accruing and is cheapest to repay while the work is being done anyway.
Sources
Adler, P. & Borys, B. (1996). 'Two Types of Bureaucracy: Enabling and Coercive.' Administrative Science Quarterly 41(1). · Brooks, F. (1975). The Mythical Man-Month. Addison-Wesley. · Cunningham, W. (1992). 'The WyCash Portfolio Management System.' OOPSLA. · Nonaka, I. & Takeuchi, H. (1995). The Knowledge-Creating Company. Oxford University Press. · Polanyi, M. (1966). The Tacit Dimension. Routledge.
Concept web
Open the full webQuestions
What is process debt?
The accumulated cost of doing work without a defined process. Like financial debt it is sometimes worth taking on and always carries interest, paid in onboarding time, rework and coordination rather than in visible failures.
Why does process debt get expensive as a team grows?
Brooks (1975) showed communication paths grow as n(n−1)/2. When the process lives in people's heads, coordination cost rises quadratically with headcount while output rises linearly — so an approach that was efficient at three people is crippling at eight.
Can every process be written down?
No. Polanyi (1966) established that we know more than we can tell, and the parts that resist articulation tend to be the judgement calls that most determine quality. A written process can capture the sequence and lose the discrimination.
Does writing processes down always help?
No. Adler and Borys (1996) distinguished enabling formalisation, written to help a competent person handle a non-routine case, from coercive formalisation, written to constrain them. The first improves performance and satisfaction; the second degrades both.