Münster's Kita-Navigator Algorithm — A Cautionary Case After Sibling-Matching Failures Cost 144 Children Their Daycare Places

The city of Münster, Germany, has long used its “Kita-Navigator” software to allocate scarce daycare (Kita) places; a 2017 court ruling had already found the city unable to demonstrate that its allocation criteria were being properly applied. In February 2022, vendor ITK updated the system to use the Gale-Shapley deferred-acceptance algorithm, intended to make allocation fairer and more efficient.

The Gale-Shapley algorithm, in its standard form, has a known limitation: it cannot jointly match applicant pairs such as siblings applying together, and ITK had reportedly flagged this limitation before the update went live. A separate configuration error further mismatched the childcare hours parents had requested with the hours the system actually offered them. Parents and daycare managers began reporting that siblings were being assigned to different facilities.

The city's youth welfare office suspended the Kita-Navigator allocation process on 13 February 2023. A joint review with ITK confirmed that the priority ordering used to compare requested care hours had been incorrectly configured. In the aftermath, 144 children who had initially been accepted lost their guaranteed daycare places. Münster officials publicly stated they were “surprised and shocked” by the failure and confirmed the algorithm had not been programmed to keep siblings together.

The city noted a trade-off it had not adequately communicated in advance: allowing siblings to be split across different daycares can, mechanically, raise each child's individual chance of getting a place at all, compared with requiring a joint placement — but parents had not agreed to that trade-off and had not been warned it was possible.

The case is a documented, named cautionary example of deploying a well-known matching algorithm without adapting it to a real operational constraint (sibling placement) and without adequate testing of a configuration change, resulting in direct, quantified harm to 144 children in the population it was meant to serve — a useful counterweight to more successful matching-algorithm deployments in education administration.

Read the full analysis: https://www.muenster.de/pressemeldungen/web/frontend/output/standard/design/standard/page/1/show/1121983