Skip to main content
← Back to News
Article5 min readIerum Digital

Your Systems Are Running. But Is Your Business?

Abstract editorial artwork for Your Systems Are Running. But Is Your Business?, in Ierum's green, terracotta and cream palette.

Your servers are responding normally. Your APIs return successful responses, and your database is operating without issues. From a technical standpoint, everything looks healthy.

Yet something could be going seriously wrong.

Imagine an ecommerce platform that normally processes around 100 transactions per hour. Suddenly, it receives 10,000 transactions within minutes.

The infrastructure handles the load. Nothing crashes, and response times remain within acceptable limits.

But those transactions are reserving inventory without turning into completed purchases. Within minutes, much of the available stock is locked, leaving legitimate customers unable to buy.

Is it a successful marketing campaign, a faulty integration, or an attempt to abuse the system?

Your technical monitoring says everything is fine. Your business is telling a different story.

When technical metrics aren't enough

We're used to monitoring CPU usage, memory, availability, and response times. These metrics are essential, but they don't always tell us whether business operations are behaving as expected.

For that, we need to track different signals: incoming orders, completed purchases, inventory reservations, pending invoices, rejected transactions, and inconsistencies between systems.

More importantly, we need to understand how those signals relate to one another.

A spike in orders might be great news. An extraordinary increase in inventory reservations without a corresponding increase in completed purchases deserves attention.

The same applies when invoices start piling up or an integration suddenly requires far more retries than usual.

These situations can be detected before a customer complains or someone discovers a discrepancy at the end of the day.

Dashboards help you see. Alerts help you respond.

A dashboard can bring together information scattered across an ERP, an ecommerce platform, and other business systems.

It might show transactions per minute, purchase completion rates, active reservations, and the time required for an order to move through the entire process.

But dashboards have an obvious limitation: someone needs to be looking at them.

That's why critical indicators should also trigger automated alerts.

Initially, these can be simple rules: too many pending operations, reservations exceeding a certain percentage of available inventory, or discrepancies between confirmed orders and ERP records.

With enough historical data, it's also possible to detect anomalies by comparing current activity with typical behavior for a given time or period.

The goal isn't to generate an alert for every fluctuation. It's to recognize changes that may have meaningful operational consequences.

Before adding tools, check what you already have

Many business platforms already include reporting, dashboards, and automation features.

HubSpot, for example, supports custom reports, sales performance dashboards, and certain automated notifications through workflows, depending on the subscription.

Some ERP platforms also provide inventory controls, operational reports, and alerting capabilities.

In some cases, configuring an existing feature may be enough.

The challenge usually appears when you need visibility into what happens between multiple platforms.

An ecommerce system might report 100 confirmed purchases while the ERP shows only 95 received orders. Each platform may be operating correctly from its own perspective, without necessarily detecting that five transactions never reached their destination.

At that point, you need to reconcile information rather than simply display it.

Three ways to implement monitoring and alerts

Different approaches make sense depending on the complexity of the operation and the technology already in place.

Datadog can combine infrastructure monitoring with custom business metrics, automated monitors, and anomaly detection. It's particularly useful when an organization already relies on it for application observability. Costs depend on the products enabled and the volume of metrics and data collected.

Grafana provides flexible dashboards and alerting capabilities using compatible data sources. It's a practical option when your systems can already expose operational metrics.

Another approach is to use Looker Studio for reporting without building an entire observability platform.

For example, a lightweight service could query the ERP and ecommerce APIs, match transactions using their identifiers, and calculate discrepancies across orders, inventory reservations, and invoicing.

The results could be stored in a data source connected to Looker Studio, making it easier to analyze operational trends.

Meanwhile, the same service—or a separate alerting component—would evaluate critical conditions and notify the appropriate people.

In this architecture, Looker Studio handles visualization and analysis. It isn't responsible for detecting urgent incidents in real time.

The right choice depends on what you need to monitor, how quickly issues must be detected, and the ongoing cost of maintaining the solution.

Detecting an anomaly doesn't mean automatically blocking everything

Let's return to the ecommerce platform receiving 10,000 unexpected transactions.

An alert might be triggered by an unusual increase in reservations combined with a sharp decline in purchase completion rates.

That doesn't prove fraud. There may be a legitimate explanation.

The next step could be notifying an operator, investigating the source of the transactions, or introducing additional validation.

Preventive mechanisms might also include reservation expiration rules, per-transaction limits, or specific safeguards.

But automated responses introduce their own risks.

Blocking every transaction during an unexpected surge could prevent legitimate customers from making purchases.

Detecting an anomaly, sending an alert, and taking automated action are three different decisions.

Each requires an appropriate level of confidence and an understanding of the potential consequences.

Observability is part of product design

When designing digital products, we tend to focus on what they should do when everything works as intended.

But understanding how to recognize unexpected behavior is just as important.

Not every company needs a sophisticated observability platform, machine learning models, or dozens of dashboards. In many cases, a handful of meaningful business metrics and well-configured alerts can make a substantial difference.

Because a system being available doesn't necessarily mean the operation is healthy.

How long would it take you today to discover an error that happened a week ago?

  • observability
  • monitoring
  • alerts
  • ecommerce
  • software

Share