Launches fail internally before they fail publicly.

Last-minute changes collide with fixed trafficking deadlines. QA gets compressed. Handoffs are missed.

The difference between a smooth launch and a scramble is usually the process. When the process is incomplete, people compensate with adrenaline. It works for a while, until it doesn’t..

Where I start

Set the gates. Milestones and go/no-go rules. People relax when the constraints are explicit.

Establish the pulse. Short status calls. A decision log that stays current. Less panic and more movement.

Define the standard. Review checklists and sign-off thresholds. “Looks fine” isn’t enough.

Run the rehearsal. Run-throughs and pre-flight checks. Discovery is not for launch week.

Find the root cause. When something breaks, look at the process that allowed it.

Evidence

Examples are anonymised to respect client and agency confidentiality..

Interactive build that passed testing, but failed live
A face-recognition component was unpredictable in real-world conditions. We built fallback paths and made the go/no-go criteria specific. We prepared a Q&A script for the client support team, and launch stopped being a coin toss.

Vendor charging for in-scope fixes
I worked on-site at a government body and briefed and tested fixes in real time. We avoided government legal issues, the email chains stopped and the fixes happened. The vendor issues were resolved after the website was launched.