A feature rollout can look finished from the platform team's perspective while developers are still working around it. The setting is enabled, the announcement has gone out, and training is complete—but the intended task is no easier.
Follow the change into everyday work. Find out whether people can complete the task, where they get stuck, and whether using the new approach improves the result. Low use can point to missing access, an awkward workflow, or a problem the feature doesn't solve. Each calls for a different response.
Define the task, not just the rollout target
Describe what someone should be able to do differently. “Enable security scanning” describes setup. “Help developers find and address exposed weaknesses before release” explains why the setup matters and what else must work.
Choose the projects where that task applies, with their maintainers. A scanner for a language the project doesn't use isn't a missed adoption opportunity. Archived repositories and projects that no longer ship software may not belong in the rollout either. Keep a short explanation of the exclusions so coverage doesn't improve merely because difficult projects disappear from the list.
Agree on who supports the workflow and who can change it. A platform team may own the configuration while application maintainers handle findings and a security team helps assess them. Plan for that work before adding more projects; enabling a feature can create demand elsewhere.
Watch someone complete the task
Start with a project that represents the intended users, rather than the easiest demonstration. Ask a person who normally does the work to try the new approach with their usual access. An administrator completing it on their behalf can hide a permissions problem that everyone else will encounter.
For a scanning rollout, follow a finding from the job result to a decision and, when appropriate, a fix. Can the developer understand the result, find the person who can help, and see whether the change addressed it? A passing scan job answers only the first setup question. Vulnerability management from finding to closure develops the handling process in more detail.
Include a common difficulty: an unclear result, a failed job, or an exception that needs another team's input. People don't need to memorize every recovery step, but they do need a way to find help that works during normal delivery.
Keep the walkthrough focused on the workflow rather than on whether someone passed a test. Ask where they expected to find the next step and what made it hard to continue. Those answers may call for a clearer result, better defaults, or a named support route—not another training session.
Look for use in ordinary work
After the walkthrough, review a few real changes with the team. Check whether people kept using the approach without someone guiding every step. Ask about the cases where they didn't: was the task irrelevant, did the tool fail, or was another route faster or clearer?
GitLab's DevOps Adoption by group report can help you find activity to investigate.1 Its indicators reflect feature use in a project during the last full calendar month. They don't show how many eligible projects use the feature, or whether using it helped. A group's positive indicator isn't a substitute for checking the projects included in your rollout.
You can also inspect activity directly in projects. Compare it with the work you expected to happen during the period: a team with no releases won't demonstrate a new deployment workflow that month. Avoid treating every quiet period as resistance or every recorded use as success.
For a wider rollout, keep track of which teams can use the workflow, which are using it in normal work, and which have seen the intended improvement. Don't combine these into a single score; knowing where a team is stuck is what helps you respond.
Match the follow-up to the problem
Review what you saw with the people responsible for setup, support, and the task itself. Separate a problem with access or configuration from a problem with the workflow's value.
| What you find | What to try next |
|---|---|
| The task works for an administrator but not its intended user | Check permissions and configuration with that user's normal access |
| People can start but can't handle an unclear result | Improve the explanation and the route to someone who can help |
| A guided demonstration works, but normal work takes another route | Investigate the extra effort, missing integration, or exception the demonstration missed |
| Use is regular, but the intended result hasn't improved | Reconsider the approach and the capacity to act on its output |
For example, a scanner might run on every change while findings accumulate without an owner. Promoting more scanner use won't address that queue. Agree who assesses findings and which work should take priority before extending coverage further. Conversely, if findings are handled but the scan delays every small change, improving pipeline feedback may be the next step.
Put the chosen improvement in the team's existing backlog with an owner and a follow-up date. There's no need for a separate adoption reporting process when an issue can hold the examples, proposed change, and result.
Expand when teams can sustain the approach
Before adding more projects, compare the result with the problem you set out to solve. Are weaknesses being addressed sooner? Can developers make the next decision without waiting for the rollout team? Is the extra support and maintenance work manageable? Choosing measures that guide improvement explains how to make that comparison without reducing it to activity counts.
Begin expansion with projects that have similar languages, delivery processes, and support arrangements. Carry forward the setup and explanations people needed in the first group. Check again when you introduce different runners, release methods, or team responsibilities; the earlier result won't explain those differences for you.
Make a deliberate choice when the approach doesn't fit. You might change the configuration, offer an alternative, or leave a project out with a reason. The goal is a better way to do the work, not universal use of a feature.
Tell participants what changed because of their feedback, and make sure someone remains responsible for the workflow after the rollout ends. Revisit it when the product, team, or delivery process changes enough that the old instructions no longer help people complete the task.
Further Reading
Notes
-
DevOps Adoption by group is available in Ultimate. It summarizes feature-use signals; teams on other tiers can review activity in the projects involved in their rollout. ↩
- DevOps Adoption by group explains the included features, reporting period, and processing schedule.
- Vulnerability management from finding to closure helps you put ownership and follow-through behind the scanning example.
- Choosing measures that guide improvement helps you compare the intended benefit with the effort and unwanted effects of a change.