Use data, not debate in meetings
- Engineers are increasingly framing technical disagreements with artifacts instead of rhetoric: commits, pull requests, design docs, benchmarks, incident logs, and production metrics that executives can inspect after the meeting. - The most durable examples are measurable ones: DORA tracks change lead time, deployment frequency, change failure rate, and recovery speed, while modern teams can now prototype alternatives quickly and compare outputs. - The approach fits a broader shift toward evidence-based software delivery and AI-assisted experimentation, where leaders can judge shipped results instead of confidence alone. (dora.dev)
Software teams are pushing more decisions out of conference-room debate and into code, benchmarks, and production data. (dora.dev) (docs.gitlab.com) A clean commit history shows what changed. A pull request shows who reviewed it, what tests ran, and which objections were resolved before code merged. (docs.gitlab.com) (opsera.ai) A design document does the same job earlier in the process. It turns a vague opinion into a written proposal with scope, tradeoffs, and rollback plans that other engineers can challenge line by line. (dora.dev) (research.google) Benchmarks matter because they replace “I think this is faster” with repeatable numbers. In machine learning and AI work, that can mean latency, cost per request, accuracy, failure rate, or human preference scores across two competing prototypes. (research.google) (dora.dev) Production data is the hardest evidence to wave away. If one version cuts response time, reduces incidents, or lifts conversion after release, the meeting shifts from persuasion to interpretation. (docs.datadoghq.com) (docs.gitlab.com) The standard management shorthand for this is DORA. The framework tracks deployment frequency, change lead time, change failure rate, and failed deployment recovery time to measure how quickly teams ship and how reliably they recover. (dora.dev) (docs.gitlab.com) Those metrics were built for delivery performance, not for winning arguments. But they give engineering leaders a common scoreboard when two teams make competing claims about speed, quality, or operational risk. (dora.dev) (cloud.google.com) The AI boom has made this style of decision-making easier to practice. Teams can now generate a prototype in hours, run a benchmark suite, and compare two implementations before a dispute hardens into politics. (research.google) (dora.dev) That does not eliminate judgment. Leaders still decide which metric matters, whether the benchmark matches real traffic, and whether a local optimization creates a bigger maintenance cost later. (dora.dev) (research.google) The practical rule is simple: bring evidence that survives the meeting. A commit, a benchmark, and a production graph usually travel further than the strongest voice in the room. (dora.dev) (docs.datadoghq.com)
