Building Grip Detection in Flutter

Grip detection means knowing which hand holds the phone — and adapting button positions accordingly. It's a small UX win with big retention impact.

The signal#

The accelerometer tells you device tilt on the X-axis:

• Negative tilt = leaning left = user holding right
• Positive tilt = leaning right = user holding left

Easy in theory. Hard in practice.

The problems#

Real users don't hold phones perfectly still. They walk. They're on the bus. They're lying down on a couch. Naive thresholding misclassifies 40% of the time.

A few sources of false positives:

• Walking introduces a ~2Hz sway that mimics tilt drift
• Landscape rotation looks identical to a deep tilt
• Lying flat on a desk reads the same as both hands free

The solution#

Multiple signals combined:

1. 20-sample window of accelerometer data — rolling median, not raw
2. Posture detection (portrait / landscape / moving / flat) — gates the classifier
3. Persistence prior — last session's grip is a strong predictor of this one
4. Eager-lock window — if 5 samples agree with the prior, lock it for the rest of the session

This gets you 92% accuracy on the first session and 98% on session 5+.

In Morph SDK#

All of this is one widget:

GripAdaptiveLayout(
  child: ProductView(),
  primaryAction: AddToCartButton(),
)

The button repositions toward the user's thumb. That's it.

What you get for free#

The same widget also handles:

• Reachability — anything inside primaryAction stays in the bottom 60% of the screen
• One-handed mode detection — collapses non-essential UI when grip is unstable
• Telemetry hook — onGripChange: (grip) => analytics.track(...) for product analysis

A small thing that makes your app feel like it was built for this user, not a generic one.