> For the complete documentation index, see [llms.txt](https://docs.thousandeyes.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.thousandeyes.com/product-documentation/internet-insights/provider-intelligence/scoring-methodologies/trend-analysis.md).

# Trend Analysis

## Overview

**Trend Analysis** shows whether a raw metric, viewable in [single-destination views](https://docs.thousandeyes.com/product-documentation/internet-insights/provider-intelligence/view-results/destination-views), is **increasing, decreasing, or staying flat** over your selected time frame. This is critical for identifying:

* **Degradation**: Is latency creeping up over months?
* **Improvement**: Is a provider's stability improving after infrastructure upgrades?
* **Stability**: Is performance holding steady (ideal for long-term contracts)?

Provider Intelligence uses an algorithm that ensures that the trend reflects a statistically significant movement over time rather than just temporary fluctuations.

## Trend Calculation

1. **Create a Time Index:** Assign a sequential index to each hour in the time series:

   | Hour | Time Index |
   | ---- | ---------- |
   | 1    | 1          |
   | 2    | 2          |
   | 3    | 3          |
   | ...  | ...        |
   | 720  | 720        |
2. **Measure the relationship (correlation)**: The algorithm analyzes the relationship between the performance metric values and the linear time index. It looks for how consistently the two variables move together:

   * **Positive correlation**: If the metric values tend to increase as the time index increases, it suggests an upward trajectory.
   * **Negative correlation**: If the metric values tend to decrease as the time index increases, it suggests a downward trajectory.
   * **No correlation**: If the metric values fluctuate randomly or stay the same regardless of time, there is no strong correlation.

   The algorithm applies a single value to the correlation (a coefficient) in a range from −1 (perfect negative correlation) to +1 (perfect positive correlation).
3. **Map to trend lines**: The system applies specific thresholds to this coefficient to provide a user-friendly line on the results page:

   | Coefficient Range | Trend Line (most metrics)         | Trend line (throughput)\*         |
   | ----------------- | --------------------------------- | --------------------------------- |
   | -1.00 to -0.50    | Down, green (improving)           | Up, red (improving)               |
   | -0.50 to +0.50    | Flat, gray (no significant trend) | Flat, gray (no significant trend) |
   | +0.50 to +1.00    | Up, red (worsening)               | Down, green (worsening)           |

   \* Most metrics perform better at lower values. Throughput alone performs better at higher values.

### Example: Latency Increasing

| Hour | Latency (ms) |
| ---- | ------------ |
| 1    | 10           |
| 2    | 12           |
| 3    | 15           |
| 4    | 18           |
| 5    | 20           |

**Coefficient**: +0.95

**Trend**: the raw metric trend line points up (worsening: lower values are better).

## Trend Variability

Trends are calculated separately for each [percentile](https://docs.thousandeyes.com/product-documentation/internet-insights/provider-intelligence/view-results/summary-statistics#understanding-percentiles) (P5, P50, etc.) and for each time frame (1, 3, 6 months). This means:

* You might see the latency trend point down (improving) at P95 but flat at P50, indicating that worst-case performance is improving while typical performance remains steady.
* The latency trend might point up (worsening) over 6 months but flat over 1 month, indicating latency worsened in earlier months, but has been steady over the last month.