Games announce dates and then move them, usually late and usually by months. The pattern is consistent enough to be structural, and it comes from what the last phase of development actually contains.
The unpredictable work is scheduled last
Building individual systems is comparatively estimable, because a team has done similar work before and can size it against experience. Connecting those systems is not.
Integration reveals conflicts that only appear once everything runs together, and the number of interactions grows much faster than the number of features. Two systems have one relationship; ten have many.
Because integration sits at the end, the least predictable work is also the work with the least remaining schedule to absorb it. A slip there has nowhere to go.
Dates are committed before the game exists
Release windows are set early to coordinate manufacturing, marketing, retail partners, platform slots and localisation, all of which need long lead times of their own.
That means the date is chosen at the point when the least is known about the project. It reflects business scheduling rather than an engineering estimate.
When the two diverge, the date is the thing that has to move, because the alternative is shipping a build that does not exist yet in a usable form.
Certification and localisation cannot be compressed
Console approval, age ratings and translated builds all sit after the game is otherwise finished, and each has a fixed duration that no amount of staffing shortens.
A failed certification submission adds a further cycle. A studio that finishes late does not lose the days it overran, it loses those days plus the queue.
This is why delays are frequently announced in multiples of weeks rather than days. The schedule downstream of development moves in blocks.
Adding people late makes it worse
New staff need context, and the people who can supply it are the ones already fully occupied. Onboarding consumes exactly the capacity it is meant to add.
Late-stage work is also concentrated in specialists: performance engineers, certification testers and audio implementers. Their tasks cannot be parallelised across a general workforce.
Studios that respond to a slip by hiring generally find the date moves anyway, which is why experienced production leads prefer cutting content to adding staff.
Announcing the delay is itself a calculation
A studio that knows in autumn it will miss a spring date may wait, because an early announcement extends the period during which the project has no news.
Waiting is expensive in a different way. Pre-orders taken against a date that leadership already doubts damage trust when the change arrives.
The most common compromise is announcing a delay alongside new footage, which converts the news into a beat rather than a gap. The timing is a communications decision layered on a production one.