A 22-month hyperscale data centre build. ~$2.1B installed cost. Peak headcount of 2,400 trades across seven primes. Schedule tied to a contractual power-on date with liquidated damages running into seven figures per week of slip.

The owner’s earned-value reports tracked SPI at 0.97 through topping-out. That number was technically accurate. It was also useless.

Across the eight months from shell-complete to commissioning start, the project bled approximately 14,800 paid labour hours into white space that never appeared on a cost code.

Same disease as every other capital programme. Different host body.

The reported number and the real number

Earned value was clean because the schedule was clean. Activities were closing. Quantities were tracking. SPI green.

The Crew Productive Hours Ratio (CPHR) –  installed-work hours divided by paid hours,  landed at 48%.

Forty-eight. On a hyperscale build with seven-figure weekly LD exposure, just over half of every paid labour hour was already lost before the question of “are we tracking schedule” was even asked.

EV reported what was budgeted. CPHR exposed what was happening. They were not the same conversation.

Where the white space lived

Three deficit zones, in order of magnitude:

1. Badging and gate queues (≈6,200 hours). The site had two access gates, both routing through a single contractor-staffed badging trailer. At 06:30 shift change, queue depth peaked at 240 workers. Mean queue time per worker per shift: 11 minutes. Across 2,400 headcount and 22 production weeks in this phase, the math is brutal. Nobody owned gate throughput because nobody was measured on it.
2. MEP trade stacking on data hall floors (≈5,100 hours). Electrical bus duct, mechanical CRAH placement, fire protection mains, and overhead cable tray were all chasing the same overhead corridor. The 4D model existed. It hadn’t been updated since IFC. Crews arrived at the floor with conflicting work fronts and either stood by, retreated, or worked around — none of which produced installed work.
3. Vertical transport queueing (≈3,500 hours). Three temporary hoists, twelve floors, four primes booking the same hoists informally. No dispatcher. No logged manifest. Average wait at the call station: 14 minutes. Material crews routinely abandoned the queue and redeployed, leaving the trades upstairs idle.

None of this showed up as variance. All of it showed up as carrying cost funded by the owner, invisible to the owner.

What the framework demands

Three rules, applied in sequence:

Measure installed work, not reported work. SPI tells you whether activities are closing on schedule. It does not tell you whether the people you are paying produced installed work today. Different question. CPHR forces the right one.

Name the white space. Gate queue. Trade stacking. Hoist queueing. Inspection rework loop. Each gets a label, an owner, and a number. Anything you cannot name, you cannot govern.

Govern the deficit, don’t just report it. Weekly utilization deficit review at the same cadence as cost and schedule. An empowered chair. Working session, not steering committee. Decisions out the door inside seven days.

What the intervention recovered

By month four of the eight-month window, three governance changes had landed:

 A second badging trailer stood up at the south gate; queue time at shift change dropped from 11 minutes to under 3.

 The 4D model was re-baselined every two weeks against actual installed work, with a coordinator empowered to sequence the floor.

 A hoist dispatcher was hired and cost-shared across the four primes; manifest moved to a published schedule.

CPHR moved from 48% to 64% across the back half. On a ~$310M direct labour spend in that phase, every CPHR point was worth roughly $3.1M. Sixteen points equals $49.6M of recovered value and a power-on date that hit window.

The takeaway nobody wants to print

A green SPI is not a green project. Earned-value reporting was never built to surface utilization deficit. It was built to track activities against a budget. The two questions are not the same, and the gap between them is where the carrying cost lives.

If you are an owner, a hyperscaler, a sponsor, or a programme director, and you cannot tell me your CPHR within ±5 points, you do not have a number. You have a story your reporting system is paid to keep telling.