Blog / Technical SEO /

A site migration risk register for a small team

Most migration risk registers get written at 11pm, the night before launch. By then it’s a list of worries, not a plan.

Write it earlier. Small teams may not have a migration manager, but they still need one place to capture what could go wrong, how likely it is, who can spot it and what to do if it happens. Keep the register to real decisions. A theoretical list of every possible failure helps nobody.

Write risks as specific scenarios

I write each risk as a cause, an event and an effect. For example: the redirect map is incomplete, old URLs return errors, and visitors can’t reach equivalent content. Add scope, likelihood and impact using definitions the team understands, and separate a known issue from a possible risk. Don’t fake numeric precision. A plain high, medium or low is enough if you explain it.

Assign prevention and response

For each important risk, name an owner, a preventive check, a signal that would reveal it and a response. Risks come from URL mapping, DNS, certificates, analytics, content, forms, integrations and staff access. Keep a rollback route for anything that can interrupt the live service. Dependencies need owners too. “Waiting on platform” isn’t a plan.

Use it during the migration

Review the register at planning, staging and launch checkpoints. Close an item only when there’s evidence the risk is resolved or accepted by the right person. Add new risks as you find them and remove stale ones.

Keep the register small enough to use

Ten well-described risks beat sixty vague ones. If an entry has no owner and no signal, it’s a worry, not a risk. Put the top few where the whole team sees them, such as the project channel or the launch checklist.

For a small team, the usual suspects are worth checking first: redirects, DNS and certificate timing, analytics and tag changes, forms that stop sending, and anyone who loses access to a system on launch day. Review the list together before each checkpoint and ask one question. What would we do if this happened at 9am on launch day?

Apply the review to one real case

Say what a finished review changes for the person using the site, not just what work the team did. Use the same scope and definitions across examples and note what you excluded, because a lopsided sample can make one failure look like the norm.

Follow one real URL from the old site to its published state on the new one, and check what a visitor can understand or do without an internal explanation. Before acting on detection and rollback steps, confirm which source holds the underlying fact and whether it’s current. A copied export may describe an earlier template or configuration.

Otherwise a sound recommendation sits stuck between queues. Make trade-offs visible too: accuracy against speed, coverage against upkeep, simplicity against a useful exception. And separate pages that share a template from pages that share a user need.

Check the outcome in the environment customers see. An approved draft can still be altered by imports, apps, caching or theme behaviour. Where detection and rollback steps conflict, compare timestamps, definitions and ownership before choosing, because the disagreement may reveal a process gap. If a question can’t be answered with the evidence available, write down the limit and assign a follow-up.

Keep the register’s source and reasoning in an approved system and link to it. Don’t paste personal or confidential material into shared notes. If the information is incomplete, I’d rather run a small reversible test than apply a sitewide rule, especially when a wrong change could hit several customer journeys.

Allow an exception only when it solves a real operational problem, and document its boundary. Trigger the next review by a meaningful change to the platform, market or policy, not an arbitrary reminder. When closing an item, check nearby routes that rely on the same facts, such as feeds, support answers and linked pages.

Share the outcome as a decision, the evidence and the next step. Make sure whoever owns the migration knows where the current guidance lives and how to restore the previous state. Before reusing the register for another site, check the audience and technical conditions really match.

Which migration risk would your smallest team struggle most to detect in time?