Progressive coaching needs a product-wide frequency budget
Aliases: coaching cadence · intro quota · just-in-time frequency cap
What it is
Once features may speak at first relevance, those speeches still stack inside one product. A global budget caps how often progressive coaching may fire: per session, day, or week, automatic intros added together must not exceed it. Without a cap, each feature independently decides “relevant now,” and someone can collect three timely lessons in one morning. This entry is about cadence of just-in-time intros inside a single product. It is not about hints going dead after habituation, and not about how several teams rank claims on one quota.
Why it happens
Relevance is locally true and globally conflicting. First filter, first share, and first shortcut can land in the same hour, even the same task chain. Each sentence is justified alone; chained, they become another front-loaded class that merely uses contextual triggers. Recovery time after an interruption is a shared resource, not partitioned by feature. The budget turns that resource into countable slots: spend one, and other relevant intros wait for the next slot. Whether slots reset per session or per calendar day decides whether a heavy workday gets taught to a halt. The budget must also be global—per-page or per-feature caps are no cap, because people live in time, not in page accounts. A just-in-time policy that optimizes “when relevant” without “whether a relevant intro still has a slot” teaches more, not less, to active users, because they trip more relevance points.
Where it stops holding
A sentence on blocking failure (“it didn’t save; retry here”) is not feature teaching and must not spend the coaching budget, or incident days go mute. User-opened tutorials and help centers do not spend the automatic budget. Legally required notices follow their own cadence and cannot be blocked by a teaching cap. The first session after install may have a slightly wider allowance, but the width must be a hard count, not “say everything relevant in week one.” Returning after long inactivity suddenly densifies triggers; a loose weekly budget turns the return day into a makeup class.
Applying it
- Set a product-level cap on automatic progressive intros (for example one per session, two per day). All features share that counter; none may fire merely because they are relevant.
- When a second relevance point lands in the same task chain, use silent copy at the entry instead of a second popup.
- Persist the counter per user and device; write the reset period into the rule, and watch actual intro counts on an internal dashboard.
- Verify by sampling a week of active-user sessions and measuring gaps between automatic intros. If two sentences land before one task can finish, or a day exceeds the cap, queue to the next slot instead of firing on every relevance point.