R² (Coefficient of Determination)
How well does a linear trend explain the data?
Definition
R² measures the proportion of variance in the data explained by a linear regression model. It quantifies how "trendable" a metric is — a high R² means the data points lie close to the fitted line, a low R² means the data is noisy.
Formula
Where:
is the predicted value is the mean of observed values
Interpretation
| Range | Assessment |
|---|---|
| > 70% | Strong trend — the linear model fits well |
| 40% – 70% | Moderate trend — some noise present |
| < 40% | Weak/no trend — data is mostly noise, slope is unreliable |
Key insight: A high slope with low R² is dangerous — it suggests the trend direction may be real but the magnitude is uncertain. A moderate slope with high R² is more trustworthy.
In Finsco
Used in all audit commands as part of the summary statistics table, always paired with Slope. Also used as an advisory indicator in audit returns (ROIC Trend Reliability).
| Command | Advisory Use |
|---|---|
audit returns |
ROIC Trend Reliability: R² > 70% = strong, 40-70% = moderate, < 40% = noisy |
audit moat |
Revenue Growth R² in summary table |
| Others | Summary statistics table only |
Example
Two companies with the same ROIC slope of +2%/yr:
- Company A: R² = 85% → consistent improvement, reliable trend
- Company B: R² = 25% → volatile, the +2%/yr is driven by one outlier year
Related Metrics
- Slope — the trend direction this metric validates
- Coefficient of Variation — stability (complementary to trend)