COMMISSIONING · L0–L6

Greenfield hyperscale hall: full Cx across 32 MW in 11 weeks

New hall handed over ~3 weeks late by GC. Client couldn't slip the go-live date without breaching customer contracts, so the Cx window had to compress.

Back to all case studies

A 32 MW hyperscale hall was handed over three weeks late by the general contractor. The go-live date was locked to customer contracts. We had 11 weeks to do 14 weeks of L0-L6 commissioning — and every hour of overtime introduces its own risk. Here is how we sequenced it, and where we refused to compress.

01 · SCHEDULE REBUILD

Parallelising what everyone said couldn't be parallelised

The default L2/L3 (component/system-level) test order goes switchboard by switchboard. We split the hall into three electrical zones and ran L2/L3 in parallel across all three — with separate leads owning each zone and a single test-data spreadsheet everyone wrote into.

The prerequisite: rigorous energisation lockout tables. Zone A's L3 running while Zone B's L2 was still in progress meant every crew had to know exactly what was live where. We printed the lockout map on A2 sheets and had a physical whiteboard in the war room.

  • 3 electrical zones commissioned in parallel
  • One shared test-data spreadsheet, updated live
  • Zone leads with veto power on lockout state
  • Daily 07:00 stand-up + 17:00 status wash-up
02 · L4 IST — WHERE WE ADDED PEOPLE

6 extra engineers for the integrated-systems phase

L4 IST — Integrated Systems Testing — is where cross-system interactions surface. It's also where you can't safely compress by parallelising, because interactions are the whole point. So we didn't compress it; we added six more field engineers so we could run longer shifts without individual fatigue.

Every fault an engineer catches at L4 is a fault the customer doesn't catch in production. This is the phase we told the client we would not shortcut, regardless of the schedule pressure.

03 · L5 BLACK-OUTS

Two 24-hour utility failures — planned

The L5 protocol calls for a full 24-hour utility failure. We ran two, back to back, on the same weekend. Generators sync-checked, fuel farm drawn down, ATS timing verified under full IT-emulator load. Every alarm, every response, every log entry was replayed and signed off.

“The GC had cost us three weeks. The Cx team gave them back. That's not something I've seen twice in my career.”
— Head of Data Center Operations, European hyperscaler
TAKEAWAY

Commissioning schedule pressure is real, but it's a game of what you dare NOT compress. Parallelise L2/L3, add heads at L4, hold the line on L5. The result was zero L6 defects at handover — the customer's own PAC (Preliminary Acceptance Certificate) inspector filed the fastest sign-off he'd done that year.

Bring us the site, the schedule, and the problem.

Start the conversation
OTHER CASES
PRIVACY

We keep this site simple.

We use one first-party cookie to remember your language choice and light analytics (pageviews only, no cross-site tracking) to improve the site. No third-party ad networks, ever.

Read the full privacy note →