Analytics Dashboards

Introduction

Analytics dashboards — for revenue, product usage, SLOs, marketing funnels — are the third-most-common workload on MongoDB. With the aggregation framework, $facet, time-series collections and Atlas Charts, you can render rich dashboards directly off operational data.

This lesson shows the shape of a real analytics dashboard pipeline and the operational tricks that keep it sub-second: pre-aggregation with $merge, analytics-tagged secondaries, and time-series buckets.

Understanding the topic

The dashboard performance stack:

• $facet — multiple aggregations in one round-trip; one query renders the whole dashboard.
• $merge into rollup collection — recompute every 5 min; dashboards query rollups, not raw events.
• Time-series collections for metrics — automatic bucketing & 10×+ compression.
• Analytics secondaries — heavy reads never touch the primary.
• Atlas Charts — built-in BI tool, no extra license, embeds via signed URL.

Informative example

Dashboard pipeline — KPIs, time-series, breakdowns in one $facet:

db.events.aggregate([
  { $match: { tenantId, ts: { $gte: from, $lt: to } } },
  { $facet: {
      kpis: [
        { $group: { _id: null,
            users: { $addToSet: "$userId" },
            events: { $sum: 1 },
            revenue: { $sum: "$amount" } } }
      ],
      perDay: [
        { $group: { _id: { $dateTrunc: { date: "$ts", unit: "day" } },
                    count: { $sum: 1 }, revenue: { $sum: "$amount" } } },
        { $sort: { _id: 1 } }
      ],
      topCountries: [
        { $group: { _id: "$country", n: { $sum: 1 } } },
        { $sort: { n: -1 } }, { $limit: 10 }
      ]
  } }
], { readPreference: { mode: "secondary", tags: [{ nodeType: "ANALYTICS" }] } });

Real-world use

Forbes' editorial dashboard, Bosch's IoT fleet view, and many fintech "deal of the day" panels run on this exact pattern — $facet + analytics secondaries + 5-minute rollups.

Best practices

• Pre-aggregate to a rollup collection on a cron — dashboards must not run heavy pipelines on every refresh.
• Pin BI traffic to analytics-tagged secondaries with readPreference + tags.
• Time-series collections for metrics; never store raw points in a regular collection.
• Cap dashboard query maxTimeMS so one slow query cannot starve the cluster.