System Architecture: Building a Real-Time Physical Analytics Dashboard with MQTT and Edge RFID

In standard web development, building a live analytics dashboard is relatively straightforward: you drop a JS tracker on the frontend, ingest the payload into a message broker like Kafka, and pipe it into a time-series database.

But what happens when you need to track physical humans in real life?

Recently, we were tasked with building the backend infrastructure for a massive 15,000-delegate B2B exhibition in Riyadh. The organizers came to us with a common industry problem, asking: "Where can I find an event reporting platform with live dashboards?"

They didn't just want post-event Excel sheets; they wanted sub-second live heatmaps of crowd density and instant sponsor ROI tracking. To achieve this, we had to ditch traditional optical barcode scanners and build a robust, edge-computed IoT pipeline.

Here is the system architecture we used to build a live physical reporting platform.

The Hardware Layer: Passive Telemetry

The first bottleneck in physical analytics is data capture. If you rely on staff to manually scan QR codes at every booth, your data is sparse, delayed, and ruins the attendee experience.

To solve this, we shifted entirely to passive rfid attendee tracking.

When vetting hardware, we evaluated several top-tier rfid providers in Saudi Arabia to source enterprise-grade UHF (Ultra-High Frequency) chips. We embedded these passive chips directly into the attendee lanyards. We then installed discreet, circular-polarized antenna arrays above every doorway and exhibition pavilion.

As attendees walk naturally through the venue, the antennas continuously energize and read their unique chip IDs (UIDs) from up to 5 meters away—generating thousands of physical data points per second without a single manual scan.

The Middleware Layer: Edge Computing & MQTT

You cannot send thousands of raw, duplicate RFID reads per second directly to a cloud API over a venue's congested 5G network. It will fail.

Instead, we deployed localized edge computing. Each RFID antenna is wired to a local industrial micro-PC (like an Intel NUC) running a lightweight Node.js daemon.

The edge node performs local deduplication and debouncing:

JavaScript

// Simplified local edge debouncing logic

const activeReads = new Map();

function processRfidScan(tagUid, antennaId) {

const now = Date.now();

const lastRead = activeReads.get(tagUid);

// Debounce: Only log if the tag hasn't been read in the last 10 seconds

if (!lastRead || (now - lastRead.timestamp > 10000)) {

    activeReads.set(tagUid, { timestamp: now, antenna: antennaId });



    const payload = JSON.stringify({ uid: tagUid, zone: antennaId, ts: now });



    // Publish to local MQTT broker

    mqttClient.publish("venue/telemetry/raw", payload, { qos: 1 });

}

}

By pushing the data to a local MQTT broker on the venue's intranet, we decouple the physical read from the cloud upload. A separate background process batches these MQTT messages and streams them asynchronously to our central cloud database whenever external network conditions are stable.

The Cloud Layer: Time-Series Data and WebSockets

Once the MQTT stream hits our AWS environment, we pipe the data into a time-series database (like InfluxDB or TimescaleDB). Relational databases like PostgreSQL are great for storing the user profiles, but for calculating rolling averages of foot traffic over time, time-series DBs are exponentially faster.

Finally, we expose this data to the frontend via WebSockets. The React frontend maintains a persistent socket connection to our Node backend, instantly updating a live SVG map of the exhibition floor.

The Result

Venue operators can now look at their screens and see exactly which halls are hitting capacity limits in real-time. Meanwhile, corporate exhibitors can open their specific portal and watch a live graph of unique visitors entering their physical booth.

When building rfid event solutions for mass gatherings, the key is assuming the external network will fail. By utilizing asynchronous MQTT streams and pushing data processing to the physical edge, you can deliver sub-second live dashboards regardless of crowd size.

If you are a developer or organizer looking for robust infrastructure in the GCC, partnering with specialized rfid companies in Saudi Arabia is crucial for handling local hardware procurement and edge deployments.

What is your preferred tech stack for ingesting high-volume IoT data? Let me know in the comments.