Current deployment

By capability

Survey respondents indicated their organizations deploy observability capabilities by as much as 58% (security monitoring) and as little as 24% (artificial intelligence for IT operations (AIOps) capabilities).

• At least half had deployed each of the core observability capabilities, including security monitoring (58%), network monitoring (57%), database monitoring (55%), alerts (55%), dashboards (54%), infrastructure monitoring (54%), log management (51%), and application performance monitoring (APM; 50%).
• More than a third had deployed key digital experience monitoring (DEM) capabilities—including browser monitoring (44%), error tracking (43%), and mobile monitoring (35%)—as well as artificial intelligence (AI) monitoring (42%) and business observability (40%).
• Less than a third had deployed each of the more advanced capabilities, including AIOps capabilities (24%), synthetic monitoring (26%), distributed tracing (29%), Kubernetes (K8s) monitoring (29%), machine learning (ML) model monitoring (29%), and serverless monitoring (30%).

By number of capabilities

Survey respondents said their organizations deploy eight capabilities on average. Three-quarters (75%) had deployed at least five capabilities, including 37% who had deployed 10 or more and 10% who had deployed 15 or more.

Those who had deployed at least five capabilities were more likely than average to experience less annual downtime, spend less on outages per year, and spend less time addressing disruptions than those who had deployed four or fewer:

• 5+ capabilities: 45% lower median annual downtime, and 24% less engineering time spent addressing disruptions
• 10+ capabilities: 74% lower median annual downtime, 32% lower median hourly outage costs, and 41% less engineering time spent addressing disruptions
• 15+ capabilities: 80% lower median annual downtime, 47% lower median hourly outage costs, and 39% less engineering time spent addressing disruptions

Full-stack observability prevalence

Based on our definition of full-stack observability, a quarter (25%) of survey respondents’ organizations had achieved it.

Notably, organizations that had achieved full-stack observability experienced 79% less median annual downtime, 48% lower median hourly outage costs, and 44% less time spent addressing disruptions than those that hadn’t achieved it. They also had a 27% lower annual observability spend and were 51% more likely to learn about interruptions with observability. And they were more likely to employ all observability best practices and experience most benefits and business outcomes.

When asked about the number of tools, not to be confused with capabilities or characteristics, they use to monitor the health of their systems, survey respondents overwhelmingly reported using more than one.

• Most (88%) were using multiple tools, including 45% who were using five or more tools (compared to 52% in 2023 and 73% in 2022) and 3% who were using 10 or more tools.
• The average (mean) number of tools used was 4.5, which is 11% fewer than in 2023 (5.1) and 24% fewer than in 2022 (5.9). Similarly, the median number of tools went from six in 2022, to five in 2023, to four in 2024. The most common answer (mode) for 2024 was three tools (18%), followed by five tools (15%).
• Only 6% used just one tool. However, the proportion of respondents using a single tool increased by 37% year-over-year (YoY).

Compared to those using multiple tools for observability, those using a single tool experienced the following benefits:

• 65% lower median annual observability spend ($700,000 compared to $2 million)
• 18% less median annual downtime (249 hours per year compared to 305 hours per year)
• 45% less on median hourly outage costs ($1.1 million per hour compared to $2.0 million per hour)
• 50% less engineering time spent addressing disruptions (about seven hours compared to 13 hours based on a 40-hour work week)

When we asked survey takers whether they were using an open-source solution in addition to a proprietary solution for each of the 19 observability capabilities listed above, we found that:

• More than half (51%) of respondents were using an open-source solution for one or more observability capabilities. But only about 1% were using only open-source.
• Of the three open-source solutions included in the survey, 38% were using Grafana, 23% were using Prometheus, and 19% were using OpenTelemetry for one or more observability capabilities.
• More than a quarter were using an open-source solution for AI monitoring (31%), synthetic monitoring (28%), distributed tracing (28%), K8s monitoring (27%), APM (27%), and AIOps capabilities (26%).

Respondents were least likely to select open-source support (portability) as the most important observability vendor criteria (16%), and only 21% said that adoption of open-source technologies is a strategy or trend driving the need for observability. However, 24% said they are most likely to use open source in the next year to maximize value from their observability investment.

When we asked survey respondents about how unified or siloed their organizations’ telemetry data (metrics, events, logs, and traces, or MELT) is, we found:

• Collectively, 38% had more unified telemetry data (increased by 2% from 2023), compared to 37% with more siloed telemetry data (decreased by 6% from 2023)—a roughly even split.
• Only 12% said they had mostly unified telemetry data (they unify telemetry data in one place), and 11% said they had mostly siloed telemetry data (they silo telemetry data in discrete data stores).
• About a quarter (24%) said their telemetry data is roughly equally unified and siloed (increased by 12% from 2023).

Those with five or more tools were 13% more likely to say they have siloed telemetry data to some extent (64%) compared to those with one to four tools (57%).

Compared to respondents with more siloed telemetry data, those with more unified telemetry data on average:

• Experienced 78% less annual downtime (107 hours per year compared to 488 hours per year)
• Spent 11% less engineering time addressing disruptions (28% compared to 32%)
• Had a 4% higher median ROI (302% compared to 290%)

To practice true business observability, organizations must integrate their business-related data with their telemetry data (MELT). In reviewing the types of business-related data they said they currently integrate:

• Most (87%) had integrated at least one business-related data type with their telemetry data, including 77% who’d integrated at least two and 35% who’d integrated at least five. Just 4% had integrated all 10.
• Operations data (43%) and customer data (41%) were the most likely to be integrated.
• Product research and human resources data (both 32%) were the least likely to be integrated.

Compared to those who had less than five business-related data types currently integrated with their telemetry data, those who had integrated five or more:

• Spent 32% less on hourly outage costs ($1.5 million compared to $2.2 million)
• Experienced 63% less annual downtime (139 hours compared to 370 hours)
• Spent 27% less engineering time addressing disruptions (11 hours compared to 15 hours based on a 40-hour work week)

We once again asked survey takers which of nine different observability best practices listed in the chart below they employ. We found that:

• Most (83%) had employed at least two best practices, but only 16% had employed five or more.
• Respondents were most likely to say their software deployment uses CI/CD practices (40%) and their infrastructure is provisioned and orchestrated using automation tooling (39%)—but less likely than previous years.
• Compared to last year, 24% more said their telemetry data includes rich metadata and business context to quantify the business impact of events and incidents, 18% more said users broadly have access to telemetry data and visualizations, 13% more said their telemetry is unified in a single pane for consumption across teams, 8% more said their telemetry is captured across the full tech stack, and 2% more said they can query data on the fly.

On average, compared to those employing one to four, those employing five or more observability best practices:

• Experienced 19% less annual downtime (239 hours compared to 294 hours)
• Spent 35% less on hourly outage costs ($1.3 million compared to $2.0 million)
• Spent 38% less engineering time addressing disruptions (21% compared to 34%)
• Were 36% more likely to say MTTD improved to some extent since adopting an observability solution (72% compared to 53%)
• Were 38% more likely to say MTTR improved to some extent since adopting an observability solution (77% compared to 56%)
• Spent 20% less on observability per year ($1.6 million compared to $2.00 million)

The median annual spend on observability per year was $1.95 million. More than two-thirds (67%) were spending at least $1 million per year on observability, and only 2% were spending $5 million or more. Just 13% were spending less than $500,000 per year.

Seven factors were associated with a lower median annual observability spend, including:

• Using a single tool for observability: Those using a single tool for observability spent 67% less on observability than those using two or more tools ($700,000 compared to $2.00 million).
• Deploying more observability capabilities: The more capabilities they deployed, the less they spent on observability. For example, those who had deployed five or more observability capabilities spent 13% less on observability per year than those with four or fewer ($1.90 million compared to $2.18 million). Those who had deployed 10 or more observability capabilities spent 30% less on observability per year than those with nine or fewer ($1.50 million compared to $2.15 million).
• Achieving full-stack observability: Those with full-stack observability spent 27% less on observability per year than those without full-stack observability ($1.50 million compared to $2.05 million).
• Learning about interruptions with observability: Those who learn about interruptions with observability spent 23% less on observability per year than those who didn’t ($1.70 million compared to $2.2 million).
• Employing more observability best practices: Those who had employed five or more observability best practices spent 20% less on observability per year than those employing four or fewer ($1.60 million compared to $2.00 million).
• Integrating more types of business-related data with telemetry data: Those who had integrated five or more types of business-related data with their telemetry data spent 10% less on observability per year than those who integrated one to four types ($1.85 million compared to $2.05 million).
• Having more unified telemetry data: Those who had more unified telemetry data spent 5% less on observability per year than those who had more siloed telemetry data ($1.90 million compared to $2.00 million).