KPIs and Analysis

A KPI (Key Performance Indicator) in Kudzu Canopy NOC is a computed value that describes a measurable aspect of your network’s health — device reachability, gateway uptime, RF coverage quality, airtime utilization, or any other operational signal. KPIs are produced by analysis algorithms, stored as analysis data, and consumed by dashboards, alarms, reports, maps, and AI workflows.

Why it matters#

Raw network traffic tells you what happened, but KPIs tell you what it means. A single uplink record is just a data point; a KPI that tracks how many devices went silent in the last 24 hours is an operational signal you can act on. By continuously computing KPIs, the platform transforms high-volume telemetry into a manageable set of indicators that surface the conditions you care about most.

How it works#

Analysis algorithms#

An analysis algorithm is a scheduled program that reads network data and emits metrics. Algorithms are written in JavaScript and run on configurable time windows — hourly, daily, weekly, or monthly. Some algorithms use rolling analysis, meaning they refresh frequently while looking back over a longer sliding window to smooth out transient fluctuations.

Each algorithm belongs to a node in the problem taxonomy, which determines where its metrics appear in the Metrics Dashboard and Alarms view.

Analysis instances#

An analysis instance is the configuration that enables one or more analysis algorithms for a specific network. When you import an algorithm through the Monitoring KPIs configuration page, the platform creates an analysis instance that controls:

  • Which algorithms run against your network.
  • Algorithm parameters and tuning options.
  • Which metrics are pinned to your Favorites section.
  • Notification preferences for the alarms tied to these metrics.

Metrics#

A metric is the output of an analysis algorithm for a single run. It can be:

  • A simple number — such as the count of silent devices.
  • A ratio — such as the percentage of gateways with no traffic.
  • A count with references — such as the number of affected H3 cells, with links to the specific cells, devices, gateways, integrations, or zones that explain the value.

Metrics carry entity references so you can always drill from a number to the underlying evidence on a map or in inventory.

Analysis data#

Analysis data is the immutable result frame stored after each algorithm run. It contains the computed metrics, their entity references, metric ratios, tags, alerts, and geotags. Analysis data serves as the evidence layer used by:

ConsumerHow it uses analysis data
Metrics DashboardDisplays current and historical KPI values with charts and map focus.
AlarmsCompares metric values against thresholds to determine alarm state.
ReportsIncludes metric evidence and trends in generated documents.
Map layersRenders geotags as hexagonal cells, points, or zone highlights.
AI InsightsUses metric history and references as context for problem identification and solution generation.

Time windows and scheduling#

Analysis algorithms run on defined time windows. The choice of window affects how responsive a metric is to change:

WindowTypical use
HourlyFast-moving indicators like gateway connectivity drops or sudden device silence.
DailyBaseline health metrics like device activity ratios or coverage scores.
WeeklyTrend indicators like week-over-week capacity shifts.
MonthlyLong-term planning metrics like deployment progress or seasonal patterns.

Rolling analyses combine frequent execution with a longer lookback period, giving you near-real-time sensitivity without reacting to every momentary spike.

From metric to action#

The analysis pipeline follows a clear path:

Network data → Analysis algorithm → Metric → Alarm evaluation
                                          → Dashboard display
                                          → Map layer rendering
                                          → Report inclusion
                                          → AI insight context

Once a metric is computed, it simultaneously feeds every downstream consumer. You do not need to configure separate pipelines for dashboards, alarms, and AI — they all draw from the same analysis data.

Before metrics appear on the Metrics Dashboard, you must enable reporting for the network and import at least one analysis algorithm through the Monitoring KPIs configuration page.

Where you see this#