The Cardinality-By-Team Dashboard

Surface cardinality contributions per team. The dashboard, the conversation it triggers, and the savings teams typically achieve.

Panels

The cardinality-by-team dashboard makes per-team metric cardinality visible. The visibility itself drives behavior change; teams that see their cardinality optimize it. The dashboard's structure determines its effectiveness.

What panels matter:

• Per-team total cardinality.: Each team's total number of unique time series. The headline number; the comparison across teams; the trend over time.
• Time series count by team.: The data behind the cardinality number. Specific counts, sortable by team, drillable to specific metrics.
• Top contributing metrics per team.: Within each team, which metrics have the highest cardinality. The list is the optimization queue; addressing the top metrics produces the most savings.
• The optimisation candidates.: The high-cardinality metrics are the candidates. Reducing cardinality on these (drop labels, reduce values, restructure) produces measurable savings.
• Trend over 90 days.: Cardinality trends matter. Growing cardinality indicates accumulation; declining indicates optimization. The trend is the direction signal.
• Growth visible; spikes investigated.: Sudden growth indicates something specific. New service launched; new metric added; instrumentation bug. The spike triggers investigation.

The panels are the visualization. Each contributes to the team's understanding and conversation.

The conversation it triggers

The dashboard triggers conversations. Teams compare; outliers ask questions; learnings spread. The conversations are what produces behavior change.

• Teams compare.: The dashboard shows all teams. Each team can compare to others; the comparison surfaces outliers and benchmarks.
• Outliers ask questions.: Teams with significantly higher cardinality than peers ask why. Investigation produces understanding; the high cardinality often has identifiable cause.
• Learnings spread.: When one team optimizes, others learn. The patterns spread across the organization; cluster-wide cardinality improves.
• Healthy competition.: Teams compete to optimize. The competition produces real value; the engineering effort goes into reducing cardinality.
• Unhealthy: cardinality becomes status.: The competitive dynamic can go wrong. Teams optimize for show; metrics that should exist get dropped because they look bad on the dashboard. The team's leadership watches for this pattern.
• Avoid the optimisation-for-show pattern.: Real value requires real optimization. The team's discipline includes ensuring optimizations preserve observability rather than just reducing numbers.

The conversations are the mechanism. The dashboard surfaces the data; the conversations drive the action.

Typical savings

Teams that adopt the dashboard typically see significant cardinality reduction. The visibility alone produces value; the explicit optimization compounds.

• Visibility alone produces 20 to 30% reduction in the first quarter.: Just making cardinality visible drives reduction. Teams identify and remove unused metrics; stale labels are removed; the cumulative effect is significant.
• Teams identify and remove unused metrics.: Many metrics exist that are never queried. The dashboard surfaces them; teams remove them; cardinality drops.
• Stale labels removed.: Labels that have grown to high cardinality are reviewed. Per-user labels, per-request labels, similar high-cardinality labels are removed or restructured.
• Compounding.: The savings continue. Year-over-year, the discipline maintains; new cardinality is checked; the platform's total cardinality stays bounded.
• Pays back in observability cost.: The vendor bill drops. The platform team's storage costs drop. The savings justify the dashboard's existence.

The savings are real and ongoing. The dashboard's investment pays back continuously.

Operating the dashboard

The dashboard operates with discipline. Auto-refresh; per-team ownership; periodic review. The operations make the dashboard sustainable.

• Auto-refresh weekly.: The dashboard updates automatically. Teams see current data without manual queries; the visibility is continuous.
• Trends visible without manual queries.: The pre-computed trends are part of the dashboard. Teams see growth or decline at a glance; investigation follows when needed.
• Per-team owner.: Each team has an owner of their cardinality. The owner sees the team's data; the owner drives optimization; accountability is clear.
• Surfacing happens; teams own remediation.: The platform team surfaces the data; the application teams own the remediation. The division of labor matches accountability.
• Quarterly review with finance and engineering leadership.: The dashboard is reviewed quarterly with leadership. Finance sees the cost trajectory; engineering sees the discipline; the review keeps the program on the agenda.

Cardinality-by-team dashboard is one of those FinOps disciplines that pays off through behavior change. Nova AI Ops integrates with metric stores and team attribution data, produces the cardinality dashboard, and supports the conversations that drive cluster-wide cardinality discipline.