Observability Survey

People don't want to replace one form of toil with another

“Too much manual input of required context" is the most commonly cited limitation that would prevent respondents from using AI, followed by sentiments about breaking too often or not adapting and a lack of customization. This points to a desire to focus on tools that actually free engineers to invest time and energy in higher-priority tasks.

Observability practitioners overwhelmingly want AI to show its work

Nearly everyone (95%)³ says it's important for AI to show its reasoning—a clear indication practitioners want to maintain oversight over any actions AI takes. Interestingly, those who described AI as "critical" to the previously discussed use cases are the most likely to want its reasoning explained, while those who see no value in AI for those same use cases are the least likely to see it as important.

In their own words: Why respondents value open source and open standards in observability

Nearly two-thirds (65%) of organizations are investing in both Prometheus and OpenTelemetry⁴

Prometheus, the more long-standing of the two open source telemetry standards, has a slight edge in overall investment (77% for Prometheus vs. 76% for OpenTelemetry). However, OpenTelemetry continues to show greater signs of building momentum, with more people building POCs or investigating (35% for OpenTelemetry vs. 18% for Prometheus) and a higher percentage of people saying they've increased their investment in the past year (47% for OpenTelemetry vs. 42% for Prometheus).

OpenTelemetry is being used at high rates across metrics, logs, and traces

OpenTelemetry's potential as a vendor-agnostic standard for generating, collecting, and exporting all types of telemetry is starting to come to fruition. A majority of respondents are using OpenTelemetry for metrics (57%), half are using it for traces, and 48% are using it for logs. (Only 9% use it for profiles, but this is still an emerging space in observability and OpenTelemetry.) Among those using OpenTelemetry, adoption is up significantly across the board compared to last year.

Organizations are moving to OpenTelemetry for simplicity—and freedom

OpenTelemetry promises greater interoperability across telemetry signals and platforms, so it should come as no surprise that "ease of adoption and management" and "ease of switching vendors and backend technologies" are the top two observability concerns respondents hope OpenTelemetry will help resolve.

Half of all organizations plan to spend more on observability next year, largely because they're doubling down on their strategies

Only around a quarter of those who expect to spend more say it's due to higher vendor bills. In fact, nearly two-thirds (63%) expect to spend more because of broader adoption, and nearly a third (31%) expect to invest for a higher ROI. Among those who expect to spend less, the most likely reason is more efficient operations (37%).

Cost remains the top driver when selecting new observability technologies

Yes, 90% of respondents expect to spend the same or more next year, but it's clear they're looking to spend that money wisely, with 65% citing cost as an important tool selection criteria—the most common response. Ease of use is the next most important criteria, with close to half (49%) citing it as a priority.

Half of all respondents use observability tools to track business-related metrics

This includes areas such as security, compliance, revenue, order tracking, customer conversion and more. It's just one of the signs that observability is expanding beyond the realm of engineering. As one respondent put it, "We track business-related metrics for revenue and order tracking, customer activity and conversion rates, compliance and audit events, and security-related metrics. These insights help us understand how system behavior impacts business outcomes and customer experience."

Most organizations are moving to full-stack observability; SLOs and business observability are on the rise

Nearly half of all organizations (46%) are using unified infrastructure and application observability in production, and 85% are using it in some capacity ("investigating," "POC," "in production"). The next two most popular emerging use cases are SLOs and security observability, with roughly three-quarters saying they're using both in some capacity. Observability of LLM-based applications, naturally, also has notable adoption, with 57% of respondents implementing it in some capacity ("investigating," "POC," "in production").

Upper-level management at most companies sees observability as critical to their business

The percentage of respondents that say upper-level management (C-suite, VP, or director) sees observability as business-critical is even higher for retail and ecommerce (79%), media and entertainment (77%), and financial services (72%). The automotive and manufacturing industry (47%) is the only one where less than half of respondents say it is critical to those levels of management. 

Complexity/overhead is the biggest observability concern today

The top three observability concerns this year are complexity/overhead (38%), signal-to-noise challenges (34%), and cost (31%). However, setup (SaaS vs. self-managed) appears to play a role in where teams struggle, with self-managed respondents most likely to cite complexity/overhead, and SaaS users most likely to cite cost as their top concern.

Companies save money through centralized observability

More than three-quarters (77%) of all respondents say they have saved time or money through centralized observability. They cite savings through faster incident response, improved developer experience, and general improvements from consolidation and standardization. Those who haven't saved time or money are most likely to cite challenges with their vendor and tooling strategy, followed by adoption/implementation challenges (i.e., a lack of buy-in or being in the early stages of centralization).

In their own words: What respondents have to say about the benefits of centralized observability

Alert fatigue remains the biggest obstacle to faster incident response, regardless of role

30% of respondents cite alert fatigue—the most common response by a wide margin. This is the top response across all roles, but priorities diverge from there. For example, engineering managers are more likely to cite painful incident coordination across teams as a major obstacle (22% vs. 16% overall), while directors of engineering are less likely to be concerned with an inability to create a culture and process to learn from incidents (7% vs. 14% overall).

In their own words: How practitioners convince their colleagues to embrace observability

The survey was conducted online and in-person between Oct. 1, 2025 and Jan. 10, 2026. It was promoted by Grafana Labs through its website, newsletters, and social media channels. We also solicited responses at Grafana Labs events and third-party conferences.

In total, 1,363 responses were collected through a survey platform hosted by research consultancy Censuswide, which also assisted with our analysis.

^{1 "Somewhat important," "very important," and "essential’ answers combined}

^{2 "Somewhat valuable," "very valuable," and "critical for our use case" answers combined}

^{3 "Somewhat important," "important," "very important," and "it is critical" answers combined}

^{4 "Investigating," "experimenting or building a POC," "using it across a few production workloads," "using it across the majority of production workloads," "using it in all production workloads"}

^{5 "Somewhat satisfied" and "very satisfied" answers combined}

^{6 "Mostly satisfied" and "satisfied" answers combined}

^{7 Mature, centralized practices defined as "a centralized observability team runs the observability platform and provides best practices and support to product teams, but does not directly manage observability for individual services." Siloed operations defined as "an operations team, separate from the product development team, implements observability and is responsible for the uptime and performance of the application in production."}