Datadog's multi-dimensional usage-based pricing causes 3-10x budget overruns that companies only discover after the billing cycle ends

Datadog charges separately across dozens of product dimensions — infrastructure hosts, APM traced requests, log ingestion, log indexing, custom metrics, synthetic tests, RUM sessions, and LLM observability tokens — each with its own pricing metric and overage rate. Teams consistently report receiving invoices 3x, 5x, or even 10x their budgeted amount, not because they misunderstood the pricing page but because usage-based charges compound unpredictably during traffic spikes, incident investigations (which generate massive log volumes), or when new microservices are deployed. Why it matters: engineering teams instrument their code with Datadog to improve reliability, so more instrumentation means more custom metrics and traced requests, so the very act of improving observability directly increases the bill, so teams start selectively dropping traces and reducing log retention to control costs, so they lose visibility into the production issues the tool was purchased to detect in the first place. The structural root cause is that observability vendors have a perverse incentive structure — their revenue grows when customers have more problems (incidents generate more logs, more traces, more metrics) and when customers build more complex systems (more microservices, more hosts, more integrations), so the pricing model systematically punishes exactly the behavior the tool is supposed to encourage.