Hard Requirements vs Wishes: A Release Team Story

Hard Requirements vs Wishes: A Release Team Story

Priya Nair led the platform team at a regional logistics company, and in March she inherited a problem that looked like a technical dispute but was really a ranking dispute. Her team had to replace the aging release process for their customer portal. Within a week she had a document with fifty-three “requirements”, three engineers each defending a different favourite, and a deadline that nobody had moved. This is the story of how the group separated hard requirements from wishes in an afternoon, and what changed afterwards.

The meeting where everything was critical

The first workshop went the way these workshops usually go. Dmitri, a senior engineer, wanted one-click rollback. Hana, who handled the on-call rota, wanted a dashboard with every service’s status. A newer colleague, Joel, wanted the new process to support a programming language the team had never used in production. Each request was reasonable. Each was marked “high”. By the end of the hour the whiteboard looked like a subway map, and nobody had agreed on a single thing.

Priya did something unglamorous. She erased the board and wrote one sentence across the top: “If this is missing on go-live day, we do not go live.” Then she asked the room to rewrite their requests so that any of them could pass or fail that sentence.

Sorting by the go-live sentence

The exercise was uncomfortable for about twenty minutes. Dmitri’s rollback survived at once, because a release that cannot be undone is a risk the business would not accept. Hana’s dashboard did not survive in its original form, though a narrower version did: someone on call must be able to see whether the latest release is healthy. Joel’s new language failed the test completely. It was interesting and possibly valuable, but the portal ran fine without it.

By the end, the “hard” column held seven items. Among them: automated tests must run before any release reaches customers, secrets must never be stored in the repository, and a release must be traceable to a person and a change. The remaining forty-six items went into a second column that Priya titled, with some irony, “good ideas for later.”

Where the requirements came from

The seven were not invented in that room. Priya had asked the team to look at past incidents and at what the portal’s customers actually depended on. She also borrowed structure from outside reading. One article on an engineering blueprint for web project requirements reminded her to separate functional needs from operating conditions such as availability and security. Another piece on questions worth asking before a custom web application is built gave her a list of prompts that the team used to pressure-test each remaining candidate. A requirement that could not answer “who is harmed if this is absent?” was quietly demoted.

The tie-breaker column earns its keep

Two weeks into the build, the team hit a snag. The cloud provider’s quote for a build-runner service came back higher than planned. In the old days that news would have triggered another long meeting. This time Priya opened the second column and asked the group to name the three least valuable items that were also the most expensive. They cut those in under fifteen minutes. The seven hard requirements never came up for discussion, and the schedule held.

Hana noticed something else. Because the tie-breakers were ranked and visible, people stopped lobbying in hallway conversations. Everyone could see where their idea stood and what it would take to move it up. Two items climbed the list after the first release, once real usage showed they would save the on-call group real time.

A detour: an engineer’s sister and a lesson from outside the office

Dmitri mentioned the method at home, and his sister, who was arranging care for a relative overseas, borrowed it. She wrote down the practical musts first: travel dates the family could manage, a language the relative felt comfortable in, and a companion for the trip. She then used an educational site, BestCardiacHospitals, which is written for international patients, to see how it frames the stages of sharing needs and comparing options. She also noted that the site describes its content as educational and says it is no replacement for medical advice or emergency care, so she took all clinical questions to the relative’s own doctors. Dmitri told the team the story as a small proof that the habit is portable, and then moved on to the next agenda item.

How the first release week felt

On the Tuesday of release week, Joel asked whether anyone was nervous. Oddly, nobody was. Each hard requirement had been demonstrated in a rehearsal the previous Friday: the rollback was triggered on purpose and finished in minutes, the test gate blocked a deliberately broken change, and the traceability record named the right person. When a checklist is short, rehearsing every line is realistic. With fifty-three items, the rehearsal would have been a fortnight of theatre and still would have missed the real risks.

What went wrong anyway

It would be tidy to say the process was flawless. It was not. In month two, an engineer argued that a faster build cache was a must-have, since slow builds were wearing people down. Priya listened and realised the original sentence had a gap: it covered go-live risks but not the team’s ability to keep going afterwards. They added an eighth line for “build time stays below a limit the team agrees on”, and wrote the reason beside it. The list changed because the evidence changed, which was precisely the point.

A second hiccup arrived with a manager who wanted a reporting feature for an executive review. It failed the go-live sentence, so it sat in the second column, but the manager objected. Priya’s answer was practical: the feature could move up if someone gave a reason it affected go-live, or if the schedule gained slack. The manager found a reason, and the feature moved up with the cost visible to all.

What the team kept

  • A single sentence that defines “hard”, pinned at the top of every planning document.
  • A short list of hard requirements, each with a reason and a name beside it.
  • A ranked second list, reviewed at each milestone, where any item can move up with a reason.
  • A habit of rewording wishes as testable outcomes before they are ranked.
  • A rule that cuts always begin at the bottom of the second list.

What changed afterwards

The new release process went live on the original date. The first customer-facing release under it was uneventful, and so were the next few. Hana later said the biggest difference was not technical at all: meetings got shorter because the argument about “what is critical” had already been settled on paper. Joel’s favourite language eventually found a home in an internal tool, where it did no harm and some good.

Priya’s closing note in the team wiki was brief. “We did not get everything we wanted,” she wrote. “We got everything we needed on the day we needed it, and we know exactly where the rest sits.”

Takeaways for your own team

  1. Write the pass-or-fail sentence for your project before anyone proposes a solution.
  2. Rewrite every request so it can be tested against that sentence.
  3. Keep the hard list short enough to recite from memory.
  4. Rank the rest, and cut from the bottom when money or time shrinks.
  5. Re-open the hard list only when evidence, not mood, says it should change.