Definition of Done vs Acceptance Criteria
Two people look at the same piece of work and agree it is done. One means that it does what was asked of it. The other means that it has been tested, reviewed, documented and put somewhere the next person can use it. Both are reasonable meanings of the word and the gap between them is where a great deal of rework lives.
Acceptance criteria describe what this particular item has to do to be accepted. The Definition of Done describes what finished means for everything the team produces, whatever the item happens to be. Section 5 of the PMBOK® Guide Eighth Edition gathers the working agreements and techniques teams use for this, and the pairing matters because each one leaves a hole that only the other fills.
One is about the item, one is about every item
Acceptance criteria vary. Each item has its own, written with whoever will accept it, and they describe behaviour, performance or condition specific to that piece of work: this report reconciles to the ledger, this station cycles in sixty-two seconds, this form rejects a malformed reference. They answer the question of whether the right thing was built.
The Definition of Done does not vary. It is a single standard, agreed once by the team and revisited occasionally, listing what has to be true of any item before anybody calls it finished. It answers the question of whether the thing was built properly, and because it applies to everything, it does not need renegotiating item by item, which is the source of most of its value.
What belongs in a Definition of Done
Verification to a stated level. Not tested in general, but tested to something specific: the checks run, the coverage expected, the cases that must pass. A vague line here produces a team in which each person applies their own standard and everybody believes the standard is shared.
Review by somebody who did not do the work. A second pair of eyes, named by role, on the work itself. This is the item most often quietly dropped under pressure, and it is usually the one keeping the others honest.
Integration with the whole. The work sits in the real system, the real line, the real document set, and the whole still functions. Items that are complete in isolation and unintegrated are the raw material of a late and painful assembly phase.
Everything the next person needs. The instruction updated, the drawing reissued, the record entered, the training given. This is the part that distinguishes an output from a usable output, and it is the part that a downstream team experiences as whether the project respects them.
A station that met every criterion
A manufacturer was improving a sub-assembly line, station by station, with a project team working alongside production. Each station had acceptance criteria and they were good ones: cycle time at the duty the line ran, first-pass yield above a stated figure, ergonomic assessment passed, verified by the named quality lead.
The third station met all of them and was signed off on a Thursday afternoon. On the Monday the line ran it at about half rate and the team leader had four operators working from a printed sheet the project engineer had written out by hand.
Nothing had been missed against the criteria. What had been missed was everything that was not in them: the work instructions still described the old station, the two night-shift operators had not been trained, the fixture change had not been added to the maintenance schedule, and the spare sensor was on a six-week lead time with none held. Every one of those was somebody's job and none of them belonged to any single acceptance criterion, because they were the same for every station.
The project wrote a completion standard that afternoon and applied it to the remaining nine stations. Work instructions reissued and posted. Both shifts trained, with names recorded. Maintenance schedule updated. Critical spares on site. Those four lines did not change the acceptance criteria for any station and they changed what handover meant, and the fourth station onwards ran at rate from the first shift.
The despatch lane outside the production office made the point permanently. Two pallets of the same assemblies, one wrapped and labelled and ready to go, one finished in the sense that the parts were made. Both are done, by somebody's definition, and only one of them can leave.
Predictive work has the same thing under another name
Completion and exit criteria. Stage gates, handover certificates, commissioning checklists and the conditions for leaving a phase are doing the work of a Definition of Done, and they are frequently more rigorous than anything an adaptive team writes. The vocabulary differs and the mechanism is the same: a standard that applies to everything crossing a boundary, agreed in advance and not renegotiated per item.
The team owns it and the organisation constrains it. A team may set a higher standard than the organisation requires and cannot set a lower one, because the regulatory, safety and contractual conditions are not the team's to trade. Where a project finds its team's definition quietly below the organisation's standard, that is a conversation to have immediately rather than at the first audit.
For a PMP® candidate, the thing to carry is that the two artefacts answer different questions, so a scenario where completed work keeps returning is worth examining for the standard nobody wrote down. A response that tightens the acceptance criteria of one item has addressed a condition that was probably met. A structured PMP exam preparation course works over situations where the work satisfies everything it was measured against and is still unusable.
One question, asked of three people separately, tests this: what has to be true before we call something finished here? If the three answers differ, the team has no definition, and writing the common version down will take twenty minutes and settle arguments for the rest of the project.
Interested in going further?
The gap between built and usable is where most handover disputes live, and closing it is a matter of agreeing one standard before anybody needs it. Omega's PMP® Exam Preparation works through completion standards in predictive, adaptive and hybrid delivery.
Working agreements of this kind are among the techniques described in the PMBOK® Guide Eighth Edition.
Ad · Amazon affiliate link.
References
A104: Acceptance Criteria: The Small Detail That Prevents Big Arguments
A191: Product Backlog vs Sprint Backlog
A193: Backlog Refinement: What It Is and Why It Matters
A180: Facilitation Techniques for Difficult Project Conversations
A187: Retrospectives That Lead to Change
PMP and PMBOK are registered marks of the Project Management Institute, Inc.
