Don't Reward 4: Category Hopping. Switching categories mid-stream looking for traffic.

The pattern: streamer starts in Just Chatting. Sees low CCV. Switches to a trending game. Sees no boost. Switches to Cozy Games. Switches back. The algorithm reads the volatility as inconsistency.

What the algorithm does:

• Splits the broadcast's engagement data across each category visited
• None of the category-specific browse pages get a strong placement signal
• Recommendation engine can't connect the streamer to coherent audience archetypes
• Sidebar recommendations become diluted across multiple weak categories

The recovery: commit to one category for 90 days minimum. Even if that category isn't producing immediate results, give the algorithm enough consistent data to optimize against.

The legitimate exception: updating category to match the current activity (move from Just Chatting to a specific game when you start playing it). This is not category hopping — it's accurate categorization. The bad pattern is switching categories within the same activity for traffic-seeking reasons.

Pick the lane. Hold the lane. The data the algorithm needs to optimize you takes 6-12 weeks of consistent category presence to build.