How do you decide when to strengthen a Definition of Done, and what do you do about the gap it opens?

A Definition of Done is not written once and frozen. It is a **ratchet**: the team raises it as its capability grows, and it is never quietly lowered. Knowing when to turn the ratchet, and what to do about the gap a new line opens, is the part of this subject that belongs to whoever leads rather than to the team's routine. ## When to strengthen it Two conditions must hold together: 1. **The current standard is being met consistently.** Adding a line to a standard the team already misses teaches everyone that the standard is decorative. Fix the holding problem first — a bar that yields is not a bar. 2. **Evidence says the current bar is the constraint.** Escaped defects of a recognisable kind, rework on recently finished items, a release path that needs days of manual effort, an accessibility complaint — each points at a specific missing line. Strengthen where your own evidence points, not where somebody else's checklist points. Add **one line at a time**. A single line the team can hold under pressure is worth more than six lines honoured on quiet weeks, because the standard's entire value is that it holds on the bad days. ## The gap between the standard you want and the one you can meet Teams routinely want a line they cannot yet sustain: full automated verification of a legacy area, an accessibility check nobody is trained to run, a release step that still needs a person. There are three honest options and one dishonest one. | Option | What it means | When it is right | |---|---|---| | Adopt now, invest now | Add the line and fund what makes it cheap | The gap is small and the investment is already scheduled | | Hold as a stated aspiration | Standard stays as it is; the wanted line is recorded with an owner and a date | The gap is real and the work to close it is not yet funded | | Adopt for newly touched work | The line applies to everything the team touches from now on | A legacy area cannot be brought up to the bar in one step | | Adopt and skip quietly | The standard says one thing and the team does another | Never | The distinction that matters is between an **aspiration that is visible** and an **exception that is invisible**. A wanted line recorded openly — with what is missing and who is closing it — keeps the standard honest and creates pressure to close the gap. The same line adopted and then skipped destroys the credibility of everything else in the standard, because it teaches the team that lines are negotiable. ## Strengthening under a deadline Consider a public-library catalogue team bound to a delivery date that carries a financial penalty. Six weeks out, an audit shows accessibility verification missing from most recently finished items, and somebody proposes suspending two lines of the standard to protect the date. Suspending them creates no capacity. It moves work from before the date to after it, where it costs more, and it hides how much was moved. The lever that genuinely exists is **scope**: fewer items, finished to the same standard. State that tradeoff plainly to whoever owns the penalty clause — a smaller product releasable on the date, or a larger one carrying an unquantified remainder that lands after the penalty has already been triggered. Framing it as a decision for the person who owns the consequence, rather than as an engineering preference, is what makes the argument land. If a line genuinely must be relaxed, the relaxation is explicit, bounded and recorded: which line, for which items, until when, and who closes it. That is a decision. Silently missing it is not. ## Several teams, one product When several teams contribute to one product, what ships is only as releasable as its weakest contribution — so the lowest standard among them becomes the real standard for the product, while the highest one gets the credit. The usual answer is an organisational minimum every team must meet, with each team free to make its own standard stricter for its own work. Set that minimum at what every team can genuinely hold. A minimum half the teams miss is worse than a lower one they all meet, because it reintroduces exactly the ambiguity the standard existed to remove, and it does so while looking rigorous on paper. ## What to avoid - **Raising by decree.** A standard imposed without changing what the team can actually do produces skipped lines and a quiet second standard alongside the official one. - **Wholesale copying.** Another team's list encodes their product, their constraints and their tooling, not yours. - **Retroactive changes.** Changing the standard mid-Sprint and applying it to work already underway breaks the agreement rather than raising the bar. - **Never revisiting it.** A standard untouched for two years usually stopped describing the team long ago, and everyone has quietly stopped reading it. The last thing to own is verification. A line is not adopted when it is written; it is adopted when items finished several Sprints later still satisfy it. Check that, and if the line is being skipped under pressure, either fund what makes it cheap or take it back out and say so openly.