Releases Feel Rushed? Diagnose Them With Start-Stop-Continue

Rajesh Kumar

Rajesh Kumar is a leading expert in DevOps, SRE, DevSecOps, and MLOps, providing comprehensive services through his platform, www.rajeshkumar.xyz. With a proven track record in consulting, training, freelancing, and enterprise support, he empowers organizations to adopt modern operational practices and achieve scalable, secure, and efficient IT infrastructures. Rajesh is renowned for his ability to deliver tailored solutions and hands-on expertise across these critical domains.

Does this sound familiar? The release goes out, everybody exhales, and somebody near the door mutters that it felt rushed again. Nobody can say exactly why. The sprint was planned, the pipeline is automated, and yet the last two days were a scramble. If so, you have a problem worth diagnosing, and one of the gentlest tools for diagnosis is the start-stop-continue retrospective. Let us walk through it together, problem first, then solution.

The problem: a feeling without a cause

“Rushed” is a symptom, not a diagnosis. A doctor does not treat a fever without asking what is behind it, and a team should not treat a rushed release by simply promising to plan better. Plans do not fail because people forget to plan. They fail because specific habits push work into the final hours: late requirement changes, reviews that queue up behind one person, tests that run only at the end, or a handover that happens by chat message at the last minute.

The first job, then, is to turn the feeling into observations. That is exactly what the start-stop-continue format does.

Step one: look at what actually happened

Before the meeting, pull up the timeline of the last release. When were the last changes merged? When did testing finish? When did the release note get written? A rough timeline in front of the group beats everyone’s recollection, because memory tends to compress the stressful part and erase the quiet part.

In many teams this exercise reveals something surprising, such as most of the changes landing on the final day, or the release note being the last thing written because nobody owned it. The diagnosis begins to appear without anyone saying a word.

Step two: sort observations into start-stop-continue lists

Give everyone a few minutes to jot down notes, then sort them. Here is how the rushed-release problem often breaks down.

Things that point to “start”

  • A cut-off day for new changes, a little before the release date, announced at the beginning of the sprint.
  • A short review slot each morning so pull requests do not wait for a single reviewer.
  • A checklist that names an owner for the release note, the rollback plan and the customer message.

Things that point to “stop”

  • Accepting “small” late additions without moving something else out.
  • Running the full test suite only once, on the last day.
  • Announcing release dates that were set before the team estimated the work.

Things that point to “continue”

  • The automated pipeline that makes the actual deployment calm.
  • The short stand-up that surfaced two blockers early.
  • The habit of keeping the rollback script rehearsed and ready.

Look at how the three lists together tell a story. The deployment itself was not the problem, which is the continue list. The trouble sat upstream in how changes arrived and how reviews queued, which is the start and stop lists.

Step three: choose the root habit, not every habit

Most teams find ten candidate problems and try to fix them all. A better approach is to ask: if we changed only one thing, which would relieve the most pressure? In the rushed-release example the answer is almost always the late-change habit. A cut-off day stops work arriving at the worst moment, and it makes every other fix easier, because reviewers and testers get their time back.

Pick one start, one stop and one continue to focus on for the next release. Write them somewhere visible. The rest can wait in a backlog for the next retrospective.

Reading the lists for hidden patterns

Once the three lists are on the board, step back and look for repeats. Several items may be the same problem described from different angles. “Reviews wait too long” on the start list, “late additions accepted” on the stop list and “end-of-sprint crunch” in the general notes might all trace back to the same cause: work enters the sprint at a steady rate, but reviewing capacity does not. Group such items and give the group a plain name. A named pattern is much easier to discuss than five scattered complaints.

Also look for items that appear in more than one list. If a daily check-in is on the continue list from one person and the stop list from another, the team probably uses it for two different purposes. Clarify what the check-in is for, and the argument usually dissolves.

Step four: try it, then check

Run the next release with those three changes in place. Afterwards, return to the same questions. Was the last day calmer? Did anyone slip a change in after the cut-off? Was the continue habit still honoured? If the answer is “mostly,” keep going. If the answer is no, find out why. Perhaps the cut-off was announced too late, or perhaps a manager outside the team keeps asking for exceptions. That is a different problem and deserves its own conversation.

A tiny example, told as a story

This one is made up, but you may recognise it. A small team kept shipping on a Friday afternoon, and each Friday ended in a rush. At the retrospective someone put the timeline on the screen and everyone saw the same thing: the biggest merges arrived after lunch on release day. The start was a Wednesday cut-off. The stop was merging anything into the release branch on the day itself. The continue was the rollback rehearsal. Two releases later the team reported that Fridays felt like ordinary days. Nothing dramatic had changed. They had simply stopped inviting the rush.

How this connects to culture and customers

Rushed releases are rarely only technical. They reflect how a team handles pressure, who feels able to say “not yet,” and how much the people who use the product are in mind when decisions are made. If you want to read around those themes, this article on how modern delivery affects customer satisfaction is a thoughtful companion, and the practical guide to managing culture change in DevOps helps when your diagnosis turns out to be about habits and attitudes rather than tools.

An outside example of sorting things clearly

Other communities have had to organise what they know too, and their layouts can be instructive. BestPilotsNow is an aviation website you browse to explore pilots, flight schools, courses and reference material, and it keeps conversation and writing in separate corners. When your team’s problems are mixed together, borrowing that instinct to separate things is half the cure: keep the cause, the symptom and the fix on different lines.

Frequently asked diagnoses

What if the problem is one person?

It rarely is. A person who holds up reviews is usually sitting at a bottleneck the process created. Describe the bottleneck, not the person, and look for a way to spread the load.

What if the same stop returns every time?

Then you have not found the root cause. Ask what makes the old habit appealing. Often it is meeting a real need, and the stop will only stick when that need is met some other way.

What if nobody speaks up?

Use silent writing. Everyone drafts notes alone for ten minutes, and the facilitator reads them out without names. Quiet teams often have plenty to say once the pressure of speaking first is gone.

Wrapping up

A rushed release is a signal, not a sentence. By looking at what actually happened, sorting observations into start, stop and continue, choosing one root habit and checking afterwards, you move from a vague unease to a concrete change. It takes an hour, a timeline and a willingness to look honestly at how work arrives. After that, release day tends to become the calm, uneventful day it was always meant to be.

Leave a Reply