Website

Releasing Changes to a Business System Without Stopping the Business: Staged Rollouts Explained

November 11, 2026 · 6 min read · Etzal Group Team

A distribution company updates its order system on a Sunday night, four months of changes in one release. On Monday at eight, forty warehouse staff cannot print picking lists. A change to the address format has broken the label template. Rolling back takes three hours, because the database was altered in a way the old version cannot read. Nine hundred orders ship late. The developers resolve to release less often, which guarantees that the next release will be even larger. The opposite cure works better: a staged rollout of small changes, each one reversible.

Why big releases fail

  • More change means more ways to break.
  • When something breaks, the cause is hidden among hundreds of changes.
  • Everyone is affected at once.
  • Rollback is all or nothing, and often impossible.
  • Users face many differences in one morning, with no time to learn any.

Releasing rarely feels cautious. It concentrates risk.

The stages

  1. Internal test. The team and a few key users try the change on a copy of real data.
  2. Pilot group. One branch, one warehouse shift or five named users work with it for real.
  3. Wider group. A quarter of users, or all users of one function.
  4. Everyone.

Between stages, somebody looks: error logs, speed, and a direct question to the pilot users. The time between stages can be an hour for a small fix or a week for a new process.

Choose pilot users with care.

  • People who use the affected function daily.
  • Willing to report problems quickly.
  • Representative, including the slow internet branch and the old printer.
  • Not only the enthusiasts.

Feature switches

A switch lets new code be installed and remain off.

  • Deploy on Tuesday, switched off. Nothing changes for users.
  • Turn on for the pilot group on Wednesday.
  • Widen on Friday.
  • If something is wrong, turn off. Seconds, not hours.

Switches can be set by user, role, branch or company. They also let the business choose the moment: the feature goes live after the training session, not when the developer happened to finish.

Discipline is needed.

  • Each switch has an owner and a removal date. Old switches left in the code become confusion.
  • A list of all switches and their current state.
  • Both states tested, on and off.

The hard part: the database

Code can be rolled back in a minute. Data cannot.

The safe pattern is to change the database in steps that keep the old version of the code working.

  1. Add, do not alter. A new column or table beside the old one.
  2. Write to both old and new, for a time.
  3. Move existing data in the background.
  4. Read from the new.
  5. Remove the old only when nobody needs it, weeks later.

At every step, the previous version of the application still works, so rollback remains possible. Renaming or deleting something in the same release that depends on the change is how the opening story happened.

For every release, ask: if we return to the previous version in ten minutes, will it run against the database as it is now?

Rollback as a rehearsed action

  • Written steps for returning to the previous version.
  • A named person who can decide to roll back, reachable on release day.
  • A trigger agreed in advance: which symptom means we go back without debate.
  • Tested on a staging environment, not attempted for the first time in a crisis.
  • Backups verified by an actual restore.

Teams that can roll back in minutes release calmly. Teams that cannot, release with dread.

Watching after release

A release is finished when it has been observed, not when it has been installed.

  • Error rate compared with before.
  • Speed of key screens.
  • Business measures: orders created per hour, invoices issued, labels printed. A fall can reveal a problem that produces no error.
  • Background jobs and integrations.
  • User reports, through one known channel.

Decide before release what normal looks like, so that abnormal is recognised.

Timing

  • Not before a weekend or holiday, when nobody is watching.
  • Not at month-end closing, payroll day or peak season.
  • Early in the day and week, with the team present.
  • Announced: users know what changes, when, and whom to contact.

The traditional Sunday-night release hides problems until Monday morning, when the whole company finds them together.

Telling users

  • What is changing, in their words, with a picture.
  • Who is affected.
  • When.
  • What to do if something looks wrong.
  • For process changes: a short training or a two-minute video before the switch is turned on.

A change that works technically and surprises its users is still a failed release.

Small changes, often

Staging works best when releases are small.

  • Each release has a few changes, so a fault is quickly traced.
  • Automated tests run before each release.
  • A staging environment that resembles production.
  • A repeatable, scripted deployment, not a manual sequence someone remembers.

These are investments. They pay back the first time a problem affects five pilot users for ten minutes and not four hundred for a morning.

What a business owner should ask the team

  • How long does it take to roll back? When was it last tested?
  • Who gets a change first?
  • Can a feature be turned off without a new release?
  • How do we know within an hour that a release is healthy?
  • When was the backup last restored as a test?
  • How many changes are in a typical release?

Clear answers mean the system can change without stopping the business.

A release-day checklist

  • Changes listed and tested.
  • Database change compatible with the previous version.
  • Rollback steps written and decision maker named.
  • Backup verified.
  • Pilot group chosen and informed.
  • Monitoring ready, with normal values noted.
  • Users told.
  • Not a risky date.
  • Someone watching for the first hours.

Common mistakes

  • Months of changes in one release.
  • Everyone at once.
  • Releasing on Friday evening or Sunday night.
  • Database changes that break the old version.
  • Rollback never tested.
  • No monitoring of business measures.
  • Pilot users who are only the enthusiasts.
  • Switches never removed.
  • Users not told.
  • Responding to a bad release by releasing less often.

We build and operate business systems with staged releases, feature switches, monitored deployments and rehearsed rollback, as part of Enterprise Software. If each update of your system is an event the company dreads, contact us.

Summary

Release small changes in stages: internal test, pilot group, wider group, everyone, with a look between each. Use feature switches to separate installing from releasing, change the database in steps that keep the previous version working, and rehearse rollback with a named decision maker. Watch errors, speed and business measures after release, choose sensible dates, tell users what is changing, and treat a bad release as a reason for smaller releases, not rarer ones.

Frequently asked questions

What is a staged rollout?

It is releasing a change to a small group of users first, checking that it works under real use, and then widening the group in steps until everyone has it. If a problem appears, only the first group is affected and the change can be withdrawn before it reaches the rest of the company.

What is a feature switch?

It is a setting that turns a new feature on or off without installing new software, often for chosen users, branches or roles. The code is delivered switched off, enabled for a pilot group, and turned off again in seconds if something goes wrong. It separates the technical act of deploying from the business decision to release.

How often should a business system be updated?

Small changes released often are safer than large changes released rarely, because each release contains less that can go wrong and less to investigate when it does. Many teams release weekly or more often. The right frequency depends on how well releases are tested, staged and reversible, not on a fixed number.

Author

Etzal Group Team

Etzal Group Team, helping businesses grow with technology. Contact us.

Related service

Website Development

Fast, SEO-ready websites that turn visitors into customers.