two & a half gamers – The Attention Budget Rule: why tutorials should teach maximum 3 things at a time

Summary

We sit down with Katie Madding (CEO, emhanceAI, ex-Adjust CPO) and John
Hobson (founded user research at Bungie and Blizzard, 20+
years on Halo, Destiny, WoW). The conversation walks through
two real case studies — a Skullgirls Mobile playtest and a
creative comparison — using emhance AI’s facial-emotion
analysis (0.89 ROC AUC, basically “lie detector” precision).

The findings are uncomfortable. Skullgirls’ tutorial pauses
combat to teach mechanics players don’t need yet. Every
level ends with 5 passive screens stacked back-to-back. And
the creatives with “near-death experience” hooks within the
first 4 seconds genuinely outperform — not because Chinese
UA teams said so, but because the emotion data shows it.

If you build games or ads for mobile, this is required
viewing.

The link to the report is here: https://eu1.hubs.ly/H0v1sLM0

Peaksel case study: https://www.emhance.ai/success-storie…

📌 KEY TAKEAWAYS

 

— emhance AI uses facial-emotion analysis (0.89 ROC AUC, “lie 

  detector” precision) to identify exact moments of high 

  and low engagement during playtest sessions. Engagement 

  = emotion + attention combined.

 

— Skullgirls Mobile tutorial findings: pausing combat to 

  teach optional mechanics players don’t need creates 

  frustration. Players come away thinking the game is 

  shallow because the tutorial fights are too easy and the 

  text walls killed momentum.

 

— John’s “attention budget” rule: teach maximum 3 things 

  at a time, then give players space to digest. From his 

  Halo days: pop up a tooltip → if used, snooze 5 minutes 

  → if not, repeat. Snooze 15 minutes after second use.

 

— The “5-stack of boring screens” problem is universal. 

  Win → victory animation → reward screen → loading 

  screen → cut scene → talking head → finally back to 

  gameplay. Nobody designed this stack on purpose. Real 

  playtesting reveals it.

 

— Why 6 out of 7 AI ads still suck: the concept is right 

  (“save the character”) but the small execution details 

  matter enormously. Emotion data lets you identify which 

  4 of 15 things in a creative actually worked rather than 

  throwing out the whole concept.