Measurement ends on completeness
The measurement stage determines the outcome and is the one that gets compressed, because it produces nothing visible. The exit criterion has to be completeness rather than elapsed time.
Analogue endpoints are the most frequently omitted category and the most consequential. Lift emergency telephones, fire alarm panels, door entry systems, franking machines, alarm diallers and lone-worker devices are all attached to the estate and appear on no asset register, directory or organisation chart.
Undocumented call flows follow. Where nobody can describe the disposition of a call arriving on a published number at eleven at night, that behaviour is lost at cutover and its absence is discovered under the least favourable circumstances available.
Porting governs the schedule
Number porting is generally the longest item and the one least subject to project control. Lead time belongs to the losing provider and the local process.
Establish it first and build the schedule around it, rather than building the schedule and colliding with the constraint. Then verify that the account details held by the losing provider match exactly what will be submitted — a mismatched registered address is the predominant cause of rejection and is entirely avoidable.
Waves, and where the boundary helps
Simultaneous organisation-wide cutover introduces risk without benefit. Move in waves grouped by risk and dependency, with the incumbent arrangement retained until each wave satisfies its exit criteria.
This is where building in layers pays. Because each layer has a boundary, a wave can move one layer at one site without disturbing the others — which is what makes a phased fitting possible at all.
Choose the pilot for coverage rather than convenience. Include the user with atypical forwarding, the group with complex queue logic and the site with the analogue concentration. Those identify design gaps while remediation is still cheap.
After each wave
The two weeks following each wave generate most adoption issues, predominantly minor: a greeting not carried forward, a transfer sequence that differs, an endpoint needing re-pairing.
Role-specific guidance beats generic training, and a clearly identified reporting channel materially reduces degradation that goes unreported and accumulates as resentment.
Then review queue and answer statistics once representative usage exists. The specification was built from how the organisation described itself; the data describes how it behaves.