Company: NorgesGruppen Industry: Retail / Grocery
NorgesGruppen is Norway’s largest grocery group, with 53,000 employees, 2,100+ stores, and 125 billion NOK in annual revenue. Its applications team manages 1 million order lines per day, 8 million items sold, and a loyalty programme — Trumf — with 3.2 million members. When Karin joined as Director of Applications three and a half years ago, her team had inherited a monitoring setup producing 12,000 alerts a month that no one was acting on, a supplier that had moved on, and a culture of reactive firefighting. Today, NorgesGruppen’s directors start every morning watching a Grafana dashboard. The company now uses Grafana for quick commerce order and price tracking, AI-powered incident automation, and in-store computer vision monitoring — across 100% of its server park.
The challenge
The problems were visible. The causes were not.
When Karin arrived, her team of around 40 people was entirely dependent on a supplier that had already moved on to building its next-generation product. There was no support, no ownership, and no one to escalate to. The team was generating 12,000 alerts a month, but because nothing was actionable, nothing was being acted on. Store staff would notice a problem and call the team to report it. The team would call the supplier. The supplier would call back. “We were blaming the supplier and the supplier was blaming us and nothing really happened,” Karin said.
The underlying cause was a spider web of legacy systems and heterogeneous stacks. When the monitoring team explored solutions, nothing integrated cleanly enough — until a new team member suggested looking at Grafana. Even then, the team had no central observability platform, no proactive alerting, and no visibility into the journeys that mattered most: orders, prices, store operations.
The gap became impossible to ignore as the applications portfolio grew. Quick commerce now connects 219 stores to consumer apps, processes over 3,000 order events on a busy hour, and synchronises prices across every product in every store before opening. A single price mismatch reaches customers. A single receipt count discrepancy breaks reconciliation. The scale made reactive monitoring untenable.
The solution
NorgesGruppen chose Grafana Cloud in part because it was the first tool that actually worked in their environment — integrating with legacy systems, installing quickly, and extending easily through the open source plugin ecosystem.
They started on the retail side: 2,000 stores, over 11,000 machines. Rollout was smooth. The monitoring team began fine-tuning alerts and building custom dashboards, and when a major incident struck the organisation and a cross-functional task force needed visibility across systems, the monitoring team was invited in to build it. The task force used the dashboards throughout. Directors had them open continuously. That was the turning point. After it resolved, the decision was made: the rest of the data centre — the other 98% of the server park — needed Grafana too. Grafana became a business-critical system.
Today it runs three distinct programmes. Quick commerce monitoring tracks order events, receipt counts, and price synchronisation at scale: every morning, the system checks every product in every app-connected store, replaces stale prices, and pushes corrections — 64,000 on a typical Tuesday. Each trace spans four systems. Each receipt is counted twice, and the two numbers must match. When they don’t — as happened on the 21st of one month, with a gap of 11 — the team knows immediately.
For incidents, NorgesGruppen built Grabot: an AI-powered automation platform triggered directly from Grafana alerts. When an alert fires, Grabot opens a ticket, runs a playbook, and dispatches an AI agent back into Grafana to pull application logs, system logs, disk and memory metrics, correlate the incident with other signals, check deployment history, and write a full incident brief for the on-call team — all before a human has touched anything. The team is now working toward agent-to-agent automation end-to-end.
For Smart Checkout — NorgesGruppen’s own computer vision product, which uses cameras and in-store AI hardware to verify what customers scan at self-checkout — Grafana Cloud monitors every machine in every store. Each AI computer sends metrics and logs automatically. New stores provision themselves. A fleet overview shows every machine’s health state at a glance; a single click opens till-by-till camera feeds, GPU metrics, logs, and the exact software version running. Alerts on unusual patterns run in shadow mode first, counted before they go live.
Underpinning all of it was a cultural programme Karin called “Empower Our People” — moving engineers out of reactive support roles and into tech lead positions for observability, quality assurance, and testing. Within 11 months, the team had insourced all support, all maintenance, and begun insourcing the code. They rolled out 890 stores on a new backend system using their own people.
The impact
NorgesGruppen moved from a reactive, supplier-dependent team with 12,000 unactionable alerts per month to a proactive observability culture running across 100% of its server park.
- 12,000 alerts per month with zero action → full proactive observability, with AI-automated incident triage via Grabot
- 64,000+ stale prices replaced automatically every morning across 219 app-connected stores — before customers arrive
- 890 stores rolled out on a new backend system in 11 months, using insourced teams
- 85,000 purchases per week monitored by Smart Checkout computer vision across 9 stores; more than half of 10,000 weekly staff alerts self-resolve
- 130 autonomous stores operating 24/7, with full observability coverage
Directors across NorgesGruppen now start every morning with a Grafana dashboard
“All the directors started the day with watching Grafana Cloud dashboards. So it was established.”
— Karin, Director of Applications, NorgesGruppen
Looking ahead
NorgesGruppen is continuing to expand Smart Checkout to more stores, with plans to move from reactive alerting to predictive monitoring — detecting a till gone quiet, an unusual alert rate, or a new software version behaving differently before anyone raises a flag. Where safe, the system will remediate automatically.
On the incident automation side, the team is replacing static playbooks with adaptive guidelines, moving Grabot toward a fully agent-to-agent workflow. And as the company continues to insource its technology estate, Grafana Cloud remains the foundation — the shared lens through which a 53,000-person grocery organisation watches itself work, one order line, one price check, one checkout scan at a time.