Datadog

Datadog is a company from New York that offers programs for monitoring computer systems. Large companies today operate hundreds of servers, websites and apps at the same time. Datadog continuously collects measurements from all of these components. This includes questions such as: How quickly does the page respond? How full is the memory? How many errors occur per minute? Everything ends up in a shared interface with charts and alerts. If something fails, the system raises an alarm, often before customers notice anything at all. The company was founded in 2010 and has been listed on the US technology exchange Nasdaq since 2019.

Why companies pay for a glance at the dashboard

Modern software rarely consists of a single program. An online shop is made up of dozens of small services: one for search, one for the shopping cart, one for payment. If the site becomes slow, it is unclear which of these services is to blame. Without monitoring, the development team searches in the dark. Every minute of downtime costs revenue and trust.

This is exactly where the economic value lies. Datadog does not sell a single product but a subscription. Customers pay monthly, usually based on the number of monitored servers and the amount of data. Such recurring revenue is considered particularly reliable on the stock market. That is why Datadog is often cited as an example of the Software as a Service business model, meaning software that is rented rather than bought.

One side effect: once a company has set up its monitoring, it is reluctant to switch. Dashboards, alert rules and routines are tied to the system. This lock-in is called switching costs, and it makes such providers attractive to investors. At the same time, there is competition from Splunk, New Relic, Grafana and the monitoring services of the major cloud providers.

Agents, metrics and logs

Technically, it all starts with a small helper program called the agent. It is installed on every monitored server and runs there in the background. The agent regularly reads out measurements and sends them to Datadog’s data centers. There, the data is stored, aggregated and displayed as curves.

Roughly three types of data are distinguished. Metrics are numbers over time, such as processor utilization every ten seconds. Logs are lines of text that programs write about their activity. Traces follow a single request on its path through all the services involved. Together they form a picture of what is actually happening in the system. The technical term for this is observability.

On top of this lie rules for alerts. A team might set, for example: if response time rises above two seconds, send a message to the on-call team. Increasingly, machine learning methods are taking over this assessment. They learn the normal daily pattern and report deviations without anyone having to type in fixed thresholds. A common misconception is that Datadog fixes errors itself. It shows problems; the repair remains a task for humans.

Datadog in quarterly results and the AI boom

In financial news, Datadog mainly appears in connection with quarterly results. What is watched is revenue growth and the number of customers paying more than a million dollars a year. Because almost every larger software company needs such tools, the company is considered a barometer for the spending mood of the tech industry. If customers cut back on their cloud usage, Datadog’s revenue falls too.

The AI boom has given the business additional momentum. Anyone running their own language models wants to know how long responses take and what they cost. Datadog offers its own products for monitoring AI applications for this purpose. In everyday life, however, one rarely encounters the company directly. It works in the background — but when a streaming site or a banking app runs smoothly, such monitoring software is often behind it.