A data center is the only building I've worked on where the punch list matters less than the commissioning script. On a normal commercial job, you turn the building over, the owner moves in, and if a receptacle is loose you fix it next week. On a critical facility, the whole point of the project is proving that the power and cooling never blink — and you prove it by running the systems to failure on purpose, watching the backups pick up the load, and doing it again. That single fact reorders everything about how you sequence the work. If your look-ahead schedule treats a data center like a big warehouse with a lot of conduit, you are going to miss your energization date, and on these jobs the liquidated damages are not a rounding error.
Here's how I think about running the trades on one of these, and where a short-interval look-ahead actually earns its keep.
Why the sequence is inverted
On most buildings, work flows toward finishes: structure, dry-in, MEP rough, walls, finishes, occupancy. On a data center, the finishes are trivial and the MEP is the building. You have redundant utility feeds, medium-voltage switchgear, paralleling gear, generators, UPS systems, PDUs, RPP/RDPs down to the rack, and on the mechanical side you've got CRAHs or fan walls, chilled water plants, and increasingly liquid-cooling manifolds. Every one of those systems has an N+1 or 2N twin that has to be installed, wired, and tested independently and then together.
The practical consequence: you are not scheduling toward a substantial-completion date. You're scheduling toward a commissioning window, and everything upstream is just clearing a path to that window with enough float that a failed test doesn't blow the whole job. When I build the pull plan for one of these, I start at the integrated systems test date and work backward, because that's the constraint that doesn't move.
Electrical: get ahead of the long-lead gear
The schedule killer on data centers is almost never labor. It's equipment lead times. Switchgear, generators, and large UPS modules have run 40 to 60+ weeks in tight markets, and paralleling switchgear is often the single longest pole in the tent. You can have a perfect crew plan and still sit dead in the water because the gear is on a boat.
So the first job of the look-ahead is to protect the equipment path, not the labor path:
- Sequence around delivery dates, not crew availability. Your foreman can rough in feeders all day, but if the gear lands in week 30 and you needed the electrical room dried-in, sealed, and at temperature by week 28 to accept it, the room is now the critical activity — not the conduit.
- Terminate, then megger, then energize. Don't let anyone tell you the runs are "done" at termination. Insulation-resistance test (megger) every feeder before you close anything up or apply power. Finding a nicked conductor with a meter costs an hour; finding it during first energization costs a day and a very unpleasant conversation.
- Build the temporary-to-permanent power transition into the plan. There's always a moment where you switch from utility temp power to permanent, and it needs its own scheduled outage, not a "we'll figure it out that morning."
A rolling look-ahead is the right tool here because gear dates slip constantly, and when they do you need to re-cascade the affected trades within the hour, not at the next monthly update. The value of a weekly work plan on a critical facility is that it forces that conversation every Friday: what moved, what's now on the path, who needs to know today.
Cooling: the trade that finishes last and can't be rushed
Mechanical is where I've watched the most schedules quietly come apart. The chilled water plant, the air handlers or fan walls, the piping, and the controls all have to be complete and stable before you can put a real heat load on the room during commissioning. And you can't fake a heat load — you're renting load banks and running them for days.
The gotchas that bite teams:
- Pipe cleanliness and flushing take longer than anyone budgets. Chilled water systems have to be flushed and treated, and if a strainer clogs during flush you're doing it again. Put real duration on this, with a buffer — I plan 1 to 2 days of slack after flush before controls point-to-point starts.
- Controls are always the pinch point. The mechanical iron goes in on schedule; the controls contractor shows up late, understaffed, and discovers the BAS points don't match the submittal. Pull the controls contractor into the weekly plan early and hold them to point-to-point checkout milestones, not a vague "controls" bar three weeks wide.
- Coordinate cooling readiness with the load-bank plan. Cx can't run a thermal test until cooling is proven, so cooling completion isn't a mechanical milestone — it's a commissioning gate. Schedule it as one.
Raised floor and the underfloor pile-up
If the design uses a raised access floor, it becomes a coordination chokepoint because everything wants that plenum: cooling supply, power whips, and cabling. The classic failure is trades tripping over each other underfloor, then the floor going in before the pressure and airflow are verified — and now you're pulling tiles back up.
The rule I hold to: underfloor work gets sequenced and closed out zone by zone, with a sign-off that the plenum is clear and airflow is verified before floor tile goes down in that zone. Trying to run all trades under the whole floor at once and then tiling is how you end up demoing your own work. A location-based look-ahead — planning by physical zone rather than by trade in the abstract — makes these underfloor conflicts visible before the crews are standing on top of each other, which is exactly the kind of collision a tool like LookAheadWall is built to surface.
Fire suppression, security, and the trades everyone forgets to sequence
Data centers run clean-agent or pre-action suppression, and both have coordination traps. Pre-action ties into detection and the BAS, so it can't be commissioned in isolation — it lives inside your MEP integration, not off to the side. Clean-agent rooms need a room-integrity (door-fan) test, which means the room envelope has to actually be sealed — every cable penetration firestopped — before the test. I've seen that test fail on a Friday because a low-voltage sub left twenty unsealed penetrations, and the whole suppression turnover slid a week.
Security and low-voltage are the same story: access control, cameras, and the SOC integration are late-touch systems that everyone treats as an afterthought until the owner's security team shows up for their walk and nothing's programmed. Put them on the weekly work plan with real predecessors. If it's not on the wall, it's not going to happen on the day you assumed.
Commissioning is the real project
Here's the part that separates people who've done critical facilities from people who haven't: commissioning is not a phase at the end. It's levels of testing, and you should be scheduling all five levels from day one.
- L1 — Factory witness. You (or the Cx agent) witness gear tested at the factory before it ships. Miss this and you inherit problems on site.
- L2 — Component verification. Equipment arrives, gets checked against submittals, installed correctly.
- L3 — Individual system startup. Each generator, UPS, and cooling system started and functionally tested on its own.
- L4 — Integrated systems test (IST). The big one. You put load banks on, kill utility power, and watch the generators and UPS carry the whole facility through the transfer — then do it again under fault scenarios. This is what the owner is actually paying for.
- L5 — Sustained load / pull-the-plug. Run the full load over a sustained period to prove stability.
Two things blow up commissioning schedules. First, teams treat Cx as a single bar instead of five levels with dependencies, so nobody sees that L3 on the generators has to finish before the L4 IST can even start. Second — and this is the one that hurts — nobody schedules remediation time. Systems fail their first integrated test. That's normal; that's what the test is for. If your look-ahead runs L4 right up against the energization date with zero slack, the first failed transfer puts you over the cliff. Build in the rework buffer. I want at least a few days between last integrated test and required-on date, and more if the sequence is aggressive.
Phased turnover: don't let live rooms contaminate active construction
Big campuses deliver in phases — data hall 1 goes live while data hall 2 is still being built. The moment a hall is energized and under owner control, it's a live critical environment, and construction next door now operates under change-management rules: no unplanned outages, coordinated tie-ins, escorted access. Crews who were moving fast during initial construction suddenly can't shut down a shared bus without a form and a window.
Your look-ahead has to reflect that shift. Activities that touch shared infrastructure between a live phase and an under-construction phase need to be planned around approved maintenance windows, and those windows are scarce. This is a place where the weekly work plan stops being a nicety and becomes the thing that keeps you from tripping a live data hall.
What actually keeps these jobs on track
None of this is exotic. It's the same short-interval discipline you'd use on any complex job, applied to a building where the tolerances are tighter and the failure modes are more expensive. The teams that deliver critical facilities on time do a few things consistently:
- They schedule backward from the integrated systems test, not forward from the notice to proceed.
- They protect the long-lead equipment path above the labor path.
- They plan by physical zone so underfloor and overhead conflicts show up on paper, not on site.
- They treat commissioning as five levels with real remediation buffers, not one bar at the end.
- They re-cascade the plan every single week, because on a job with this many interdependent systems, something always moved.
Whatever tool you use to run that weekly conversation — a wall of sticky notes, a spreadsheet, or purpose-built look-ahead software — the discipline is what matters. The value of a good short-interval scheduling app on a data center is that it keeps the trade-flow dependencies and the commissioning gates visible to everyone at once, so when the switchgear slips two weeks, you can see in five minutes exactly which crews and which tests just moved with it. On a building whose entire reason for existing is that it never goes down, that kind of visibility is the difference between hitting your energization date and explaining to the owner why you didn't.