What does it mean that schema shock when reality contradicts your model?

The discomfort of a failing schema is data not damage.

Example: You ship a feature you're certain users will love. Product analytics show 4% adoption after three weeks. Your stomach drops. That drop is not a sign that you are wrong about everything — it is your schema ('I understand what users need') colliding with reality. The discomfort is a signal: your model of the user was inaccurate in a specific, discoverable way. If you defend the feature, you protect the schema and learn nothing. If you investigate the gap between your prediction and the outcome, you upgrade the model.

Try this: Identify a belief you hold with high confidence about your work, a relationship, or a skill. Write it as a concrete prediction: 'If I do X, Y will happen.' Now actively search for one piece of evidence that contradicts or complicates that prediction. Write down what you find. Notice the emotional response — the flinch, the urge to explain it away, the desire to add qualifiers. That flinch is schema shock. Label it, sit with it for thirty seconds, then ask: what does this evidence tell me about the limits of my current model?