> For the complete documentation index, see [llms.txt](https://docs.console.zenlayer.com/welcome/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.console.zenlayer.com/welcome/cloud-networking/overview/network-health.md).

# Network Health Score

The health score is a 0–100 score that measures the overall condition of the tunnels in your Fabric network along two dimensions: **latency** and **capacity**. The higher the score, the healthier the network.

The score is calculated in three steps:

1. **Establish a latency baseline**: tunnels that connect the same two sites and sit in the same latency range are grouped together, and the median latency of the group is taken as the "normal level".
2. **Score each tunnel**: the latency score and the capacity score are each 0–100, and each contributes 50% to the tunnel's score.
3. **Roll up into the account health score**: the scores of all scorable tunnels are averaged.

## Data source and statistical period

The health score is calculated from the tunnel performance data in the time range you select: **last 1 hour**, **last 24 hours**, or **last 30 days**.

| Metric                    | Meaning                                                             | Used for       |
| ------------------------- | ------------------------------------------------------------------- | -------------- |
| P95 latency (ms)          | 95% of the latency samples in the period are at or below this value | Latency score  |
| Bandwidth utilization (%) | The bandwidth actually used as a share of the tunnel's bandwidth    | Capacity score |

{% hint style="info" %} <mark style="color:blue;">**Note**</mark>

<mark style="color:blue;">When a metric has no data in the period, that dimension does not take part in scoring — it is not treated as 0.</mark>
{% endhint %}

## Latency baseline

Physical distance varies widely from route to route: 80 ms may be perfectly normal on a transoceanic route, while 20 ms may already be abnormal on a metro route. The latency score therefore does not look at the absolute value — it compares the tunnel against the baseline of comparable routes.

**Grouping**: tunnels are grouped together when they meet both of the following conditions.

* **Same site pair**: the tunnel's two ends connect the same sites (for example, Amsterdam ↔ New York), regardless of direction.
* **Same latency tier**: a P95 latency below 10 ms falls into the **local** tier, and 10 ms or above falls into the **transport** tier. This keeps tunnels of very different magnitudes from distorting each other's baseline.

The baseline is the **median** P95 latency of all tunnels in the group. The median is used instead of the average so that a few outliers cannot skew it.

To keep the baseline trustworthy, a group needs at least **3 tunnels** before a baseline is generated. A tunnel in any of the following cases is not given a latency score for now (**it is not scored 0**):

* Its group has fewer than 3 tunnels.
* It has no latency data in the period.
* The sites at its two ends cannot be determined.

## From tunnel score to account health score

### Score of a single tunnel

Latency and capacity are weighted 50% each. When one dimension has no data, only the dimension with data is used — missing data does not cost points.

| Available data            | Tunnel score                         |
| ------------------------- | ------------------------------------ |
| Both latency and capacity | (Latency score + Capacity score) ÷ 2 |
| Latency only              | Latency score                        |
| Capacity only             | Capacity score                       |
| Neither                   | Not included in the roll-up          |

### Account roll-up

The page shows three scores. All are equally weighted averages across tunnels, but each covers a different set of tunnels:

| Score                                           | How it is calculated           | Tunnels included                         |
| ----------------------------------------------- | ------------------------------ | ---------------------------------------- |
| Account health score (center of the ring chart) | Average of the tunnel scores   | Tunnels scored on at least one dimension |
| Latency score                                   | Average of the latency scores  | Tunnels with a latency score             |
| Capacity score                                  | Average of the capacity scores | Tunnels with bandwidth utilization data  |

{% hint style="info" %} <mark style="color:blue;">**Note**</mark>

* <mark style="color:blue;">Scores are rounded to integers for display; when no tunnel can be scored, "—" is shown.</mark>
* <mark style="color:blue;">The period averages on the page (average latency, average utilization) are for reference only and do not take part in scoring.</mark>
  {% endhint %}

## FAQs

### Why doesn't the account health score equal (Latency score + Capacity score) ÷ 2?

The three scores cover different sets of tunnels. A Layer 3 tunnel, for example, has only a latency score: it counts toward the health score and the latency score, but not toward the capacity score. The latency score and the capacity score are therefore references for their own dimension, not a breakdown of the health score.

### Why does a newly provisioned tunnel have no latency score?

A latency score requires at least 3 tunnels on the same site pair and in the same latency tier to establish a baseline. With fewer than 3, no latency score is calculated — and it does not drag your health score down either.

### Does latency below the baseline earn extra points?

No. Latency at or below the baseline is already a full 100.

### Why does the score change when I switch the statistical period?

P95 latency and bandwidth utilization differ by period, so the baseline is recalculated as well. The same period over the same data always produces the same score.

### Which metrics are currently excluded from scoring?

Packet loss and outages are not yet counted toward the health score.
