Skip to content

Real-Time Ramadan Operations Dashboard

Live sales and operations monitoring for 1,200+ stores during the highest-stakes trading month of the Saudi retail year.

Context

Deraah Retail Group

2025 · Ramadan season

Role

Data Platform Architect

Stack

  • ClickHouse
  • Apache Airflow
  • Power BI
  • Prometheus
  • Grafana

1,200+

stores monitored live

minutes

data latency, down from next-day

1

command view for the whole network

The problem

Ramadan compresses a disproportionate share of annual revenue into a few weeks of inverted rhythm — quiet days, intense nights. Store performance data arrived next-day, which during Ramadan means a full trading cycle too late: stockouts, staffing misses, and promotion decisions were all being made on yesterday’s numbers.

The architecture

  • Fast path — incremental extracts from the POS estate land in ClickHouse on a minutes-level cadence during trading hours, orchestrated by Airflow with an aggressive-but-idempotent schedule profile used only during the season.
  • Pre-aggregation — materialized views in ClickHouse maintain store/hour/category rollups so the dashboard reads are sub-second even at network peak.
  • Serving — a Power BI operations view designed for control-room use: network map, per-region pace vs. target, and exception surfacing (stores trading significantly off forecast get pushed to the top, not hunted for).
  • Reliability — the pipeline itself is monitored with Prometheus/Grafana; during the season, data freshness is an SLO with alerting, not a hope.

What it changed

Operations leadership ran the season from a live view for the first time: replenishment and staffing calls moved same-night, and off-pace stores were flagged in hours instead of after the fact. The pattern is now the template for other peak-season events.