A migration ticket that reads “move the reporting database to the new platform” is almost guaranteed to be closed twice: once when the data lands, and again, later, when someone discovers a nightly job still pointing at the old host. The remedy is not more effort but sharper acceptance criteria, the statements that describe what is true when the work is finished. This list sets out eight criteria worth attaching to any migration ticket, illustrated with one running example: moving a reporting database and its consuming services from an aging server to a managed cloud instance.
1. One sentence describing the end state
Open the ticket with a single sentence that a new colleague could read and understand: “All reporting queries run against the managed instance, and the old server accepts no connections.” Everything else on the ticket exists to prove that sentence. If you cannot write it plainly, the scope is not yet understood, and no amount of technical activity will compensate. A tight sentence also acts as a boundary: any suggestion that does not help make it true belongs on a different ticket.
2. A named list of what moves
Migrations stall on things nobody listed. State which databases, schemas, scheduled jobs and service accounts are included, and just as importantly which are excluded. In the running example, the list might name the reporting schema, its two scheduled exports and the dashboard service account, while explicitly leaving an archive schema for a later ticket. Ambiguity about inventory is the single most common reason a closed migration reopens.
3. A check that someone else can run
Write at least one verification that does not rely on the engineer’s memory. For the example: “A count of rows in each migrated table matches between source and target as of the cutover time, and a sample of ten reports produces identical totals.” The point is not statistical rigour; it is that a reviewer who did not do the work can reproduce the result. Where possible, store the check as a script or saved query attached to the ticket.
4. Connections cut over, not just data copied
Copying data is the visible half of a migration. The invisible half is making every consumer talk to the new location. Add a criterion stating that connection strings in each listed service point at the new instance, and that the old server’s access logs show no new connections over an agreed observation window. Silence in a log is evidence that a hidden consumer is not lurking, and it costs nothing to collect.
5. Access and permissions confirmed
Teams often declare victory when data arrives, then discover that an analyst lost read access or a service lacks write permission. Spell out the roles that need access after the move and require a successful login or query from each one. The criterion can be as short as “the reporting role, the export role and the analyst group each complete a test query on the new instance.” Permission gaps found during acceptance cost minutes; found on Monday morning, they cost goodwill.
6. Written in the present tense
This one concerns style rather than substance. Criteria written in the future tense (“the system will be faster”) describe hopes. Criteria written in the present tense (“exports complete within the existing nightly window”) describe states that can be observed. The grammar nudges authors toward checkable statements. Read each line aloud and ask whether you could point at something and say “yes” or “no.” If you cannot, rewrite it until you can.
7. An explicit out-of-scope line
A short list of things the ticket will not do protects the team from slow expansion. For the example: “Archive schema, historical dashboards and schema redesign are out of scope.” Out-of-scope lines feel negative, yet they are the most effective defence against additions that arrive disguised as small favours. When a stakeholder asks for one more table, the team can point to the line and discuss a new ticket instead of absorbing the work silently.
8. A date, an owner and a version
Finally, attach the date the criteria were agreed and the person or role responsible for confirming them. If the criteria change, record a new version rather than overwriting the old text. This small habit transforms the ticket from a casual note into an agreement with a history. Months later, an auditor or a curious engineer can see what “done” meant at the time, which is far more useful than reconstructing it from memory.
Before and after: the same ticket rewritten
Consider the original wording once more: “Move the reporting database to the new platform.” A reviewer reading that line cannot tell whether the job includes the scheduled exports, whether the old server should be switched off, or how anyone would confirm success. Now read the rewritten version. It opens with an end-state sentence, names three schemas and two jobs, links a saved verification query, lists three roles that must connect, and carries a version stamp. The workload is identical. What changed is that every ambiguity the engineer would otherwise resolve alone, silently and differently from a colleague, has been settled in the open.
The rewrite takes roughly fifteen minutes. In exchange it removes a stack of follow-up messages, a likely reopened ticket and at least one awkward conversation about whose assumption was correct. For teams that handle many migrations a quarter, that trade is among the cheapest quality improvements available.
Putting the eight together
Taken together, the criteria for the running example fit on a single screen:
- The reporting queries run against the managed instance; the old server accepts no connections.
- Named schemas, jobs and accounts are listed; the archive schema is excluded.
- Row counts and a report sample match between source and target at the cutover time.
- Consumers point to the new instance and the old logs stay quiet through the observation window.
- Three named roles each complete a test query.
- Every line is stated in the present tense.
- Out-of-scope items are written down.
- The ticket carries a date, an owner and a version.
None of these demands exotic tooling. They demand only that the team decide, in advance and in writing, what finished means.
Connecting criteria to wider project planning
Acceptance criteria sit downstream of requirements. If the requirements are fuzzy, the criteria will be as well, so it pays to invest at the front of a project. A thorough engineering-level treatment of web project requirements shows how detail at the start feeds sharper tickets later. And when a team is deciding whether to build a custom application in the first place, a rundown of the questions worth asking before custom development helps ensure the project deserves a ticket at all.
A short aside on how other blogs talk about operations
Planning language shows up in surprising places. As a lighter example, an operations-themed blog whose archive spans rental guides, event listings and travel tips uses the same everyday vocabulary of scope and completion for very different subjects. It is a handy reminder that the habit of naming an endpoint is portable, which is precisely why it is worth building into engineering tickets.
Who should read the criteria before work starts
Criteria only help if the right people see them early. The engineer doing the work should read them first, since an unrealistic line is cheaper to challenge on day one than on day ten. The owner of each consuming service should read them next, because they hold knowledge about hidden dependencies that never appears in documentation. Finally, whoever will sign off should confirm that the wording matches what they intend to check. A five-minute read by these three audiences frequently surfaces a missing job or an unlisted account while fixing it is still trivial.
Common pitfalls when applying the list
Three traps recur. First, copying criteria from a previous ticket without revisiting them; the old text rarely fits the new scope. Second, writing criteria after the work is mostly done, which tends to bend the wording toward whatever happened. Third, treating the list as a checklist to be ticked mechanically rather than as a description of an end state. If every box is ticked but the opening sentence is not true, trust the sentence.
Closing thoughts
A migration ticket becomes dependable when it carries a precise end-state sentence, a named inventory, a reproducible check, proof that consumers moved, confirmed access, plain wording, a stated boundary and a dated owner. Eight short lines, written before the first command runs, convert “we think it is done” into “we can show it is done.” Adopt them on the next ticket that crosses your desk, and watch how many reopened tickets disappear.

