Every acceptance signature on a data center rests on one question: what evidence sits behind it? Commissioning (Cx) is the discipline that produces that evidence, and most of the industry organizes it into five levels. The levels are a gated progression: each proves something specific, each ends in signed documentation, and nothing advances until the gate is met or its open items are documented, dispositioned and formally accepted.
Field teams make program state visible with colored tags on the equipment itself. One common convention runs red at Level 1, yellow at Level 2, green at Level 3, blue at Level 4, white at Level 5; the colors vary by owner, but the principle does not. Walk a well-run site and you can read where every device stands without opening a binder. Here is what each level proves, and where programs quietly go wrong.
Level 0: the unofficial stage before the tags
One more term worth knowing before the tags start. Many programs use Level 0 for the design-phase work that never gets a tag color: reviewing the owner's project requirements (OPR) and basis of design, writing the commissioning plan, reviewing drawings and submittals for testability, and confirming there is a safe energization pathway and a workable testing sequence before construction locks the answers in. L0 is unofficial; you will not find it in every commissioning plan, and some owners fold the same work into design review. What matters is not the label but the timing, because the cheapest finding on the whole program is the one raised while the design can still change.
Why the gates matter
Construction schedules push everyone to declare progress, and commissioning sits where that pressure peaks. The level structure converts "we are basically done" into a binary state: the exit criteria are met and documented, or the level is not complete. Gates waved through on goodwill are how defects survive to surface later, at a worse time, at a higher price.
Level 1: Factory Acceptance Testing and Site Receipt
Level 1 is factory acceptance testing (FAT): witnessing the performance tests of major equipment at the factory before it ships. For a data center that means uninterruptible power supply (UPS) systems, generators, switchgear, power distribution units (PDUs), chillers and the rest of the cooling plant. The witness confirms the unit on the test floor matches the approved submittal, performs to its rated values, and carries the firmware, options and wiring the project bought.
The point of L1 is ownership. At the factory, deviations still belong to the vendor: a transfer time out of tolerance or a control board at the wrong revision gets fixed on the manufacturer's floor, at the manufacturer's cost. The same finding after delivery becomes a site visit, a return shipment or a concession negotiation. Witness the critical equipment in person, accept certified reports for the rest, and disposition every deviation before shipment release. On many programs Level 1 also covers site receipt: inspecting equipment on arrival, documenting shipping damage while the claim window is still open, and verifying storage conditions until installation.
Level 2: Installation Verification
Level 2 is offline quality control after installation and before energization: the least glamorous level and the one that prevents the most expensive failures. The installing contractor and its independent testing agency verify torque on electrical connections against manufacturer values, run insulation resistance testing on cables and bus, measure contact resistance across breaker and bus joints, and perform secondary injection testing on protection relays to prove they operate at their intended settings; the Cx team witnesses the critical tests and audits every record. Those settings are checked line by line against the coordination study approved by the engineer of record, not against a vendor default file. Labeling is verified against the drawings, because during an incident an operator acts on what the label says.
It all lands on check sheets, and the check sheets are a hard gate: energization proceeds only when the L2 package for that equipment is complete and signed, alongside the inspection release from the authority having jurisdiction (AHJ), utility requirements at the service point, and the contractor's energization plan. Panels are cheap to open at L2. After energization, each of these checks becomes an arc flash boundary, an outage request and a schedule negotiation.
Level 3: Start Up and Pre-Functional
Level 3 is start up and pre-functional testing: the first controlled energization, usually with the manufacturer's technician on site. Motors are bumped to confirm rotation. Control wiring is proven end to end. Sensor feedback is checked for plausibility: a chilled water supply temperature reading 54 degrees on a loop sitting at ambient is a finding, not a rounding error. Breakers are operated electrically, not just racked. Initial point verification begins against the building management system (BMS) and the electrical power monitoring system (EPMS), confirming devices report to the head end and the readings make sense.
The exit state of L3 is deliberately narrow: every component proven stable in isolation. Nothing at L3 claims the systems work together. That comes next.
Level 4: Functional Performance Testing
Level 4 is functional performance testing (FPT): scripted tests proving each system performs to the approved sequence of operations (SOO) under load. UPS systems transfer through their modes, normal to battery to normal, to static bypass and back, to maintenance bypass and back. Generators start on signal, accept load and hold it. Heat load testing with load banks proves the plant stages pumps and chillers as the SOO says it should. Permissive chains and interlocks are exercised deliberately. Alarms are triggered at the device and verified at the head end with the correct priority and text.
L4 is where commissioning schedules live or die, and most of what kills them is decided before the first script runs:
- Point-to-point verification and testing, adjusting and balancing (TAB) belong in the entry criteria, not the test scope. If they are still running during FPT, the window is doing L3 work at L4 cost with a full test team standing by.
- Scripts need agreed pass and fail criteria before anyone mobilizes. "Verify correct operation" is an invitation to argue. Every step needs a number, a state or a time bound, accepted in advance by the owner, the commissioning agent (CxA), the design team and the contractors.
- Missing setpoints or an absent alarm list means parts of the script have no acceptance criteria. If the setpoint schedule or the alarm matrix is still in draft when scripts are written, those steps are placeholders, and the gap surfaces mid-test with fifteen people in the room.
- Decide in advance where values will be forced at the controller and where the physical sensor will be manipulated. Forcing a value proves the logic and takes seconds. Heating the actual sensor proves the whole chain and costs minutes every time. Agreeing that split is where test-day margin is won.
- A fully allocated test window with zero retest allowance only holds if everything passes first time. Nothing passes everything first time. A schedule without retest days is not optimistic, it is fictional.
Level 5: Integrated Systems Testing
Level 5 is integrated systems testing (IST): the facility tested as one machine, at design load from load banks. Utility power is failed at the facility's own mains. UPS carries, generators start and take load, paralleling where the topology calls for it, cooling rides through, and when utility power returns the facility retransfers cleanly, automatically or under supervision, whichever the design intends. Redundant components are failed one at a time to prove the redundancy on the drawings exists in the field. Recovery logic is proven too, because returning to normal is its own sequence and it fails more often than people expect.
The rule of good IST is hands off. Once a scenario starts, hands touch the facility only where the sequence explicitly calls for an operator action, or the test fails. Pull-the-plug scenarios feel theatrical until the utility performs one for you, unannounced, with production load on the floor.
The five levels at a glance
| Level | Tag | What is proven | Gate to pass |
|---|---|---|---|
| L1 Factory Acceptance | Red | Equipment performs to its ratings and matches the approved submittal | Witnessed tests passed, deviations dispositioned before shipment release |
| L2 Installation Verification | Yellow | Installation is correct, protected and safe to energize | Signed check sheets, with AHJ and utility releases in hand, before energization proceeds |
| L3 Start Up | Green | Each component runs and reports correctly in isolation | Start-up reports and initial point verification complete before FPT |
| L4 Functional Performance | Blue | Each system performs to the approved SOO under load | All scripts passed against agreed criteria, issues closed or accepted |
| L5 Integrated Systems | White | The facility behaves as one machine through failure and recovery | IST scenarios passed at design load, accepted by the owner |
The economics: an order of magnitude per phase
The cost to correct the same defect climbs by roughly an order of magnitude per phase. A miswired control circuit found at FAT costs the vendor hundreds of dollars on their own floor. Found at L2, it costs thousands: site labor, access, retest. Found during FPT, it costs tens of thousands, because it burns scripted test time with a team mobilized and the window already tight. Found after go live, it is an outage, and a single significant outage can be worth more than the entire commissioning contract.
Where programs quietly go wrong
Programs rarely fail loudly. They fail through three quiet habits. The first is entering L4 with open L2 and L3 documentation: the test window turns into triage, scripts stall on faults that yellow-tag and green-tag work should have caught, and whatever retest allowance existed evaporates in the first week. The second is compressing the test window. Commissioning sits last in the schedule, so it absorbs every upstream delay, and the pressure lands on the phase whose entire purpose is proof. The third is treating the levels as paperwork: signing gates because the date arrived rather than because the evidence exists. The facility's documented state quietly diverges from its real state until IST, or operations, exposes the gap.
What good looks like when you review a Cx plan
An owner reviewing a commissioning plan does not need to check a thousand things. Check these. Entry and exit criteria for every level, specific enough that a level can fail them. Scripts that reference the approved SOO, the setpoint schedule and the alarm matrix, with pass and fail criteria agreed before mobilization. Point-to-point and TAB listed as L4 entry criteria rather than buried in test scope. A visible retest allowance in the schedule. An issues log with severities and a stated rule for which ones block a gate. IST scenarios traced to the design intent, including utility loss and retransfer at design load. A statement of what is witnessed at the factory versus accepted by report review.
That is the standard we hold our own programs to. Skapa MC's commissioning professionals have taken 284 MW of hyperscale capacity through full commissioning, with another 134 MW in progress today. The levels are how an owner buys certainty, one gate at a time, and the commissioning plan is where you find out whether you are actually getting it.