How Datadog uses Datadog to gain visibility into the Datadog user experience (opens in new tab)
Datadog’s product designers use their own monitoring tools to supplement interviews with quantitative insight into user behavior. By applying Real User Monitoring (RUM) and Logs to design questions, they made decisions about fonts, component functionality, and time-range input based on actual usage. This “dogfooding” approach improved products while making design collaboration faster and more evidence-based. ## Choosing a Monospace Font - Datadog uses monospace text for logs, stack traces, source code, container IDs, and dense data views. - Previously, users received different system fonts through a generic font stack, including Consolas, Menlo, and Courier. - The team used the browser’s CSS Font Loading API and RUM to determine which fonts users were actually seeing. - They analyzed the results in RUM Analytics and shared them through a dashboard with designers and engineers. - Datadog selected **Roboto Mono** as the standard font and used RUM after launch to verify that the rollout worked as intended. - Matching the existing visual proportions was important because font-size and character-width changes could disrupt tightly designed tables and other interfaces. ## Simplifying the DraggablePane Component - DraggablePane lets users resize adjacent content panels, but its small handle contained several controls that created visual clutter. - Custom loggers were added to the component and its draggable handle to track how users interacted with each feature. - Logs showed that almost no users used the minimize and maximize buttons, despite the space they occupied. - The team removed those buttons and replaced their functionality with a double-click on the handle. - Datadog notes that custom actions in RUM now provide a more direct way to collect this kind of interaction data. ## Expanding Custom Time-Range Syntax - Datadog initially offered only preset ranges such as 15 minutes, 1 hour, and 1 day. - The DateRangePicker introduced text-based custom ranges, but early versions supported only limited patterns such as “{N} months” or explicit dates. - Logs captured invalid user-entered time expressions, along with the page and country associated with each request. - The most common unsupported input involved “weeks,” including phrases like “last 1 week” and “last 2 weeks.” - The team used these patterns to prioritize improvements to the parser. - After adding support for common previously invalid inputs, the error rate fell from roughly 10 percent to 5–6 percent. ## Dogfooding and Collaboration - Designers tracked behavior, analyzed results, built dashboards, and documented findings using Datadog products. - Keeping data collection, analysis, and presentation in one platform made design reviews more efficient. - Shared dashboards and documentation helped designers and engineers collaborate around concrete evidence rather than assumptions. - Using the product internally also gave the team direct experience with the platform from a user’s perspective. Datadog’s examples show that quantitative product data works best alongside qualitative research: interviews explain user needs, while RUM and Logs reveal how often behaviors occur and which improvements will have the greatest impact.