FROM MEASUREMENTS TO LIFETIME

Understand how performance fades.

Explore aging data from devices and materials. Find measured thresholds and compare simple lifetime models.

Local processing · Your measurements stay in your browser

01 Your data

Workflow

Open CSV

Open saved JSON

Resume a single-series Results JSON export with raw measurements. Calculations run again locally.

Two columns: time / cycle, performance

Optional header. Comma, tab or semicolon separated; decimal points and scientific notation supported. Maximum 20,000 rows / 2 MB.

Download example CSV

New here? Follow a worked example

X-axis unit

Threshold (%)

Units label your data; changing units does not convert x values.

Normalize to

Fit start

Fit end (optional)

Use x-axis values. The fit interval does not reset time or change measured crossings.

02 Analysis

Example dataset

Measured T80

—First downward crossing

Measured T90

—First downward crossing

Linear decay slope

—percentage points / h

Performance retention

MeasuredLinear fitExponential fitDashed: threshold

LINEAR MODEL

—

EXPONENTIAL MODEL

—

Exploratory fits, not validated lifetime predictions. A good fit does not establish a degradation mechanism. Curves are drawn only over the fitted measurement interval.

Know what the numbers mean.

Measured lifetime

Performance is normalized to the first measurement or the maximum observed value. The first downward threshold crossing is linearly interpolated between neighboring points. “Not reached” means it was not observed. If data starts below a threshold, the earlier crossing is unknown.

Two transparent fits

Linear: y = mx + b, ordinary least squares. Exponential: y = A exp(−kx), least squares on ln(y). All selected y values must be positive for the exponential fit. R² and RMSE use the normalized original scale, although the fitting objectives differ. R² is undefined for constant measurements.

Scope & privacy

All calculations run locally; StabilityKit does not upload measurement files or analysis results. Cloudflare Web Analytics collects page-visit and performance information. The hosting provider also receives normal web requests. This tool does not implement burn-in detection, TS80, uncertainty intervals or certification standards. Workflow presets set labels and defaults, not physical models.

Input, units and exported results

Provide one series with unique, nonnegative x values and nonnegative performance. Rows are sorted by x. Aggregate replicates before importing. Time values should represent the same test origin; this tool does not convert timestamps or units. CSV exports include raw values, retention and fitted values inside the selected interval. Results JSON includes raw measurement points, settings, actual fitted range, measured crossings and model parameters. Sharing that file also shares its measurement data. SVG exports preserve vector graphics for further editing. Do not interpret a prediction beyond the fit interval as an observed lifetime.