Trapped library study

Pre-registration

Published before we opened a single package. The point of writing predictions down first is that it makes the result capable of embarrassing us.

What we predicted

  1. A substantial minority of packages will ship with no editable source. We expected somewhere between a quarter and a half.
  2. SCORM 1.2 will dominate, despite being superseded for two decades.
  3. Dead runtimes will cluster by publisher and era rather than being evenly distributed.
  4. Tracking will mostly work. We expected the failure to be in recoverability, not in the packaging standard itself.

What would have proved us wrong

  • Fewer than 15% of packages lacking editable source would have falsified the central claim.
  • An even spread of dead runtimes across publishers would have falsified prediction 3.
  • Widespread tracking failure would have falsified prediction 4 and meant we were measuring something else entirely, badly built packages rather than trapped ones.

What we found

Predictions 1, 2 and 4 held. Prediction 3 held, and more sharply than expected, dead runtimes were almost entirely concentrated in a single publisher's older material.

The headline figure landed at 46.6%, at the top of our predicted range.

Why this is published

A number you cannot check is not a finding. Writing the predictions and the falsification conditions first is what separates a study from a marketing statistic, and it is why we will keep publishing them even when the result is less convenient than this one was.


The study · Methodology