Every stalled rollout tells the same story in the postmortem. The technology was fine. The technicians were fine. What failed was the assumption layer: sites that were not ready, scopes that did not match the floor, dependencies nobody owned, and a schedule built on hope.
Readiness is a gate, not a vibe
A site is ready when specific, checkable things are true: access is confirmed with a named contact, power and connectivity exist where the plan says they do, materials have arrived and been verified, constraints (trading hours, freeze dates, escort requirements) are documented, and the local scope has been validated against reality. If any of those is an assumption, the site is not ready. It is a future exception with a date on it.
The discipline: no technician is committed to a site that has not passed the readiness gate. A slipped site is an inconvenience. A failed visit is a cost, a morale hit, and a schedule wound that bleeds into the next wave.
The first ten sites are the real plan
Pilot sites are not a formality. They are where the runbook meets buildings, and buildings always win. The purpose of the first wave is to find what the plan missed while it is cheap: the mount that does not fit the older store format, the circuit that was never actually provisioned, the closing procedure that takes forty minutes longer than modeled. A disciplined program treats every early exception as runbook input, so site 41 inherits what site 40 learned.
Sequencing beats speed
Teams under deadline pressure want maximum parallelism immediately. Experienced programs ramp: prove the runbook, then widen the waves. Counterintuitively, this finishes faster, because nothing burns schedule like rework across a hundred sites that all inherited the same early mistake.
What to demand from a partner
Ask to see the readiness checklist before you sign. Ask what percentage of sites they expect to fail the gate on first pass, a confident partner has a number from experience and a process for it, not a promise that it will not happen. Ask how exceptions feed back into the runbook. The answers tell you whether you are buying a wave plan or a wish.