A Step-by-Step Estate Review, Told Through One Quiet Friday

A Step-by-Step Estate Review, Told Through One Quiet Friday

Nadia runs a small platform team inside a software company of moderate size, and for most of the year her calendar looks like everyone else’s: planning meetings, release trains, a backlog that never shrinks. Last autumn she did something unusual. After an embarrassing morning when a customer-facing page failed because an old certificate had quietly expired, she blocked out a Friday and told her team that nobody was allowed to start anything new. The task was to look at everything they already ran. This is how that day went, step by step, in a form you can copy.

Step one: write the list before touching anything

The team started in a shared document with a single heading: what do we run? Each person added what they knew, then they compared it against the cloud console, the billing view and the scheduler. The gaps were instructive. There were two staging environments nobody remembered creating, a storage bucket with a name that suggested a long-finished experiment and a job that sent a weekly export to a mailbox belonging to someone who had left. Nadia did not judge any of it. The rule for step one was only to find things.

Step two: name an owner for every line

Beside each entry they wrote a person or a team. Where nobody volunteered, they wrote “unowned” in capital letters. By late morning the unowned list was longer than anyone liked, and it told Nadia where her attention should go. An unowned item is neither safe nor dangerous; it is unknown, and unknown things do not get patched.

Step three: ask the three questions of each item

Is it alive and behaving? Would anyone miss it? Does it lead where people expect? The last question caught things the others would have missed. A webhook was posting to an address in a system that had been decommissioned. A dashboard embedded in the wiki loaded from a test server. A documentation page told new joiners to fetch a tool from a location that had since been reorganised.

Step four: check from the outside

Nadia asked two team members to behave like newcomers. One followed the onboarding guide literally. The other tried to restore a recent backup into a scratch environment. Both found problems within an hour: an instruction that referred to a deprecated command, and a restore step that depended on a key nobody could locate. Neither would have appeared in any monitor. They surfaced only because someone tried.

While that happened, the team also reread the cloud basics they relied on. For anyone building the same habit, the site’s overview of fundamental cloud security habits for newcomers is a sensible checklist to hold the estate against.

Step five: patch the cheap and obvious

After lunch they handled the quick wins. Operating system updates went onto the staging machines first, then production in small groups. The expired certificate was replaced and an expiry reminder was set well ahead of the next one. A library with a known fix was updated and tested. Nadia insisted on a rollback note for each change, one sentence long, stating how to undo it. The goal was boring, reversible progress.

Step six: stage the removals

For each item marked as no longer needed, the team applied a gentle sequence: tell the likely users, switch it off without deleting it and set a date to remove it if nobody complained. The two mystery staging environments were stopped, not erased. The weekly export to the departed colleague’s mailbox was disabled and a short message went to the finance team in case anyone had been relying on it. The storage bucket was made private and tagged with a review date. Cost savings followed, which echoes the warnings in our article on the hidden costs of idle cloud resources and how beginners can curb them.

Step seven: sweep the links

The last technical task was the unglamorous one: links. Wiki pages, README files, the customer status page and the onboarding guide all contained addresses. A teammate began opening them one by one and gave up quickly. Nadia mentioned a link-management service she had seen, URLSnow, whose nightly automatic health checks flag links that no longer respond. They did not need it that day, but they added a calendar item to look at whether such a tool should hold their public addresses, so the next review would begin with a list of broken links rather than a manual hunt.

Step eight: write it all down

In the final hour the team turned the messy working document into something tidy. Each item received a status: kept and owned, patched, scheduled for removal or removed. Each decision had a one-line reason. The document went into the shared knowledge base with the date at the top. Nadia then added the most important line of the whole exercise: the date of the next review.

Step nine: decide what comes next

Only now did the team look at the backlog of new projects. The surprise was that two of the planned builds depended on systems that the review had marked as fragile. They moved the repair of those systems ahead of the new work. A third project turned out to be partly redundant with an existing service the team had forgotten they ran, which saved weeks.

What stayed with the team

One small habit survived better than the rest. Each Friday review item now ends with a question people had never asked before: if this vanished tonight, who would call us? When the answer is nobody, the item is on its way out.

Nobody called the Friday exciting. But on Monday the unowned list was shorter, the certificates had reminders and the onboarding guide worked. When the next customer-facing hiccup came, it was a minor one, and the team traced it in minutes because they actually knew what they were running.

To borrow the idea for your own team, the whole structure fits on one page: list, own, question, test, patch, stage, sweep, record and decide. Do it in a half-day, do it again on a schedule and let the new projects wait until the old ones have been looked after. Teams that follow this pattern often find that the best new build is the one they no longer need.