Good data starts with an owner.
Two start dates. Three systems. One person arriving on Monday. How to make the records behind everyday work dependable.
Start with the decision
A new colleague is due to join on Monday. Recruitment has one start date, the people record has another and IT has already scheduled access. Every field is filled in. The information still cannot support a reliable decision.
Choose one workflow before trying to clean every record. For a new starter, list the facts that affect the work: confirmed start date, manager, working location and the equipment they need. Each fact should answer a practical question.
Know who can confirm the fact
The most recently edited value is not necessarily the right one. Agree which source owns each fact and who can resolve a disagreement. A signed arrangement may establish the start date while a manager confirms which equipment the person needs.
Keep the source and effective date with the value. When someone changes teams, the current manager and the manager responsible for an earlier decision may be different people. A useful record preserves that distinction.
Make conflicts easy to resolve
In this example, pause the work that depends on the start date. Show both dates to the responsible person and ask them to confirm. Then identify which tasks need to change. An unexplained overwrite simply moves the confusion elsewhere.
Give each issue an owner and a consequence. Incorrect access dates deserve attention before an inconsistent display label. A short queue of decisions that are actually blocked is more useful than a long list of imperfect fields.
Keep the correction connected
Correcting the people record is only part of the job. Check whether the equipment request, welcome meeting and access schedule used the old value. Record which dependent tasks were reviewed, including those that did not need changing.
The EDPB explains accuracy and data minimisation as data protection principles. In daily work, apply them by collecting relevant facts and giving people a route to correct errors. Start with one workflow and test a missing value, a conflict and a correction before adding more automation.