PUE Under Pressure
You are auditing a 45 MW colocation data centre against ISO 50001. The certified EnPI is power usage effectiveness (PUE), reported monthly as 1.42. Utility bills for the last 12 months imply 1.61 when total facility electricity is divided by the IT load recorded by the tenant billing system. The PUE spreadsheet is maintained by the facilities supervisor and uses three months of BMS trend data plus 'engineering estimates' for the rest of the year. The 30 computer-room air conditioners are sub-metered at the distribution board, but no meter has been calibrated or verified since installation in 2021; the UPS output meters are calibrated annually by the tenant. The energy review identifies cooling and IT load as significant energy uses. A tenant has added 30% more IT load since the baseline was set, and the baseline has not been adjusted. You have two auditors and two days on site.
1. How do you evaluate the reliability of the PUE figures, and what sampling and evidence would you use with only two days?
Read the worked answer
Rank the evidence. A utility invoice is external and independently metered, so it is more reliable than an internal spreadsheet. The 1.42 versus 1.61 gap is about 13%, large enough to matter. Start by reconciling annual facility electricity from invoices with the spreadsheet total, then trace one complete month from utility meter to BMS to formula to reported PUE. Check whether the IT load denominator comes from billing data that captures all tenants and all IT equipment, or only billed load.
Assess the estimation. Nine months of 'engineering estimates' are not measurement. Check who prepared them, on what basis, and whether the energy review or data collection plan (6.6) authorises estimation. Estimates may be acceptable in defined circumstances, but they must not be presented as measured PUE without disclosure.
Sampling with two auditors and two days. Split: one auditor on cooling and PUE, one on IT load, billing and the energy review. Do not attempt all 30 CRAC units; select a risk-based sample of six to eight, prioritising the largest units, the oldest installations and any unit with a BMS value inconsistent with its panel meter. Verify a sample against a portable meter where available, and check UPS calibration certificates. Test the BMS trend export for completeness and tampering.
Record the limitation. State what was sampled, why, and what could not be verified. The conclusion should be that the reported PUE is not supported, not merely that the numbers differ.
2. Draft the nonconformity statement(s) and classify them, including requirement, evidence and consequence.
Read the worked answer
NC 1 - major - 9.1.1 (with 6.6). Statement: 'The organization's reported PUE of 1.42 is not based on verified measurement. Reconciliation of 12 months of utility invoices with reported IT load gives approximately 1.61. Nine months of the reported figure are engineering estimates rather than measured data, and the 30 CRAC sub-meters have no calibration or verification records since installation in 2021. Consequently the organization cannot demonstrate energy performance against its EnPI.' Evidence: invoices, PUE spreadsheet, BMS export, absence of calibration records, interviews.
Classification reasoning. This is major because the EnPI is the primary means of demonstrating energy performance, the failure is systemic in the data collection design, and it undermines the intended result of the EnMS. It is not a single uncalibrated instrument.
NC 2 - minor - 6.5. Statement: 'The energy baseline has not been adjusted following a 30% increase in tenant IT load, although the load is a significant energy use and the change materially affects the relationship between energy use and the EnPI.' Minor if the organization has not claimed performance improvement and the omission is a single missed adjustment; major if the baseline is actively used to claim improvement that is not real. Justify whichever classification you choose using that reasoning.
Report consequence. Without a reliable PUE, objectives, action plans and management review are operating on unverified information. Link the findings rather than listing them in isolation.
3. The facilities manager disputes the finding and offers a live BMS demonstration; the audit is closing in two hours. What do you do?
Read the worked answer
Accept the evidence, not the pressure. Auditors should consider all relevant evidence, so agree to the demonstration, but keep control of the process: ask for a read-only walkthrough, observe the actual screens and values, and note what is shown. Do not allow the demonstration to become a tour of selected points.
Test what it can and cannot prove. A live BMS screen shows current values. It does not prove the 12 months of reported PUE, does not replace calibration or verification of the CRAC meters, and does not explain the nine months of estimates. If the demonstration reveals that the BMS has historical data the auditor had not seen, ask for an export and check it after the meeting; if it materially changes the evidence, revise the finding and say so in the report.
Keep the finding evidence-based. If the demonstration does not resolve the gap, retain the nonconformity and record in the audit notes that the auditee provided additional evidence and why it did not change the conclusion. Do not negotiate the classification at the closing meeting; explain that classification is the auditor's judgement and is subject to review and appeal through the certification body.
Close properly. Present the finding with its objective evidence, allow the auditee to respond on the record, and note any disagreement. The two-hour constraint is a reason to prioritise the major findings, not a reason to drop them.