Engagement— is not the outcome.
A 2024 meta-analysis of mental health apps found the correlation between user engagement and clinical outcomes was r = 0.16 — statistically significant but small. A large share of the engagement industry is optimizing a metric that only weakly predicts what users came for.
Engagement-as-goal works in the short term — higher DAU, longer sessions, more subscriptions in the first quarter. It produces user distrust, app store backlash, and long-term churn when users notice what has been done to them.
Three layers.— only one of them is a goal.
Outcome metrics are the transformation — often hard to measure directly, validated periodically. Behavior metrics are activation and retention at the right cadence — the working metrics for product decisions. Engagement metrics — DAU, MAU, session length — are diagnostics, never goals.
Outcome
Behavior
Engagement
Seven decisions— in order.
Each decision constrains the next, so order matters. Most failures we see in the field happen because teams skip to step 5 (reinforcement mechanics) before answering step 1 (what outcome are we actually designing for).
Define the user transformation, not the engagement target.
An outcome is a real-world change in the user’s life: conversational Spanish in nine months, a lower resting heart rate, a funded emergency account. An engagement target is a behavior inside the app: opens, lessons completed, minutes in session.
- Outcome metrics (top) — the transformation itself.
- Behavior metrics (middle) — activation and retention at the right cadence.
- Engagement metrics (bottom) — DAU, MAU, session length. Useful diagnostics, never goals.
Seven cadences.— one fits your app.
Every consumer app has a primary behavior type. The archetype determines everything that follows — natural cadence, where stretch lives, and what good recovery design looks like. If no archetype fits cleanly, that itself is a strategic signal worth investigating.
Exposure
Repeated small-dose encounters with content. Skill compounds through repetition.
Effort
High-intensity action requiring recovery afterward. Each session is significant.
Reflection
Accumulated experience being processed; needs material to reflect on.
Maintenance
Biologically or medically required tracking. Cadence set by the body or the protocol.
Decision
A periodic choice that shapes the near future.
Social
Communication or presence with others; cadence emerges from social context.
Discovery
Exploration, inspiration, or search.
Baseline & ceiling
— design lives in between.
The baseline is what the outcome requires. The ceiling is where the behavior becomes harmful, performative, or pointless. Stretch lives in between — typically baseline +20–50%, varying by archetype.
Illustrative example. A real Cadence Ladder is produced by the Cadence Diagnostic using portfolio data and the archetype-specific marginal value curve.
The anti-pattern— promise.
Sitting underneath the framework is a commitment. It is not a marketing line. It is an operational constraint that disqualifies specific patterns from Flywheel client work — patterns that produce short-term metric lift and long-term user distrust.
Manufactured urgency
Countdown timers that invent pressure to act. Real deadlines fine; invented ones aren’t.
Weaponized loss aversion
Guilt-trip notifications, shame UI, irrecoverable streaks, paid streak recovery as monetization.
Variable rewards on trivial actions
Slot-machine reinforcement on behaviors that produce no outcome — pull-to-refresh, infinite scroll as a feature.
Social pressure manufacturing
"Your friend just beat you," forced visibility, leaderboards the user didn’t opt into.
Predatory monetization
Loot boxes, obscured currency, pay-to-skip punishment loops, retention that only resolves through purchase.
Dependency-creating triggers
Time-based notifications as the primary engagement mechanic where event-based cues would serve the user.
Infinite surfaces
Feeds, scrolls, or loops without a natural stopping point. Every session should have a "you’re done for today" moment if cadence calls for it.
How we answer— Step 02 rigorously.
The Cadence Diagnostic produces a Cadence Plan — a specific, defensible prescription for how often the app should ask for the user’s time, how much stretch to apply, and how cadence should vary by user segment.
Classify the Cadence Archetype
Every consumer app has a primary behavior type. The archetype determines everything that follows. If no archetype fits, the product probably has an unclear core behavior — which is a bigger problem than cadence design.
Identify the Cadence Source
Stretch logic only applies when cadence is a design choice. For many apps it isn’t — biology, calendars, or life events set it. Before stretching, we ask what determines the cadence in the first place.
Five Questions for Cadence & Stretch
For user-determined cadence, five questions produce the Cadence Plan — a defensible prescription for frequency, stretch zone, and cadence ladder by user segment.
Two ways— to put it to work.
The framework lives in two distinct places inside Flywheel’s service architecture — a bounded standalone engagement for post-launch teams reaching for gamification, or a methodology layer inside a full Design & Build.
The Cadence Diagnostic
For post-launch founders who are reaching for gamification without having answered the upstream questions. We produce a Cadence Plan you can implement yourself, or contract us to. Lives inside the Understand tier of our service architecture.
Inside Design & Build
For full engagements, the framework is part of how we think, not a separate SKU. The Cadence Plan is an artifact produced during Discovery and referenced throughout Design and Build — how it becomes genuinely proprietary, not a document we sell once.