Baseline Adjustment Dispute in a Data-Centre Energy-Savings Guarantee
A colocation data centre in Tseung Kwan O has a baseline year with 24.0 GWh of IT energy and 40.0 GWh of total electricity, giving a PUE of 1.67. An ESCO contract guarantees 11% annual energy savings from aisle containment and free-cooling, with savings measured by IPMVP Option C against a 12-month baseline. In the first post-retrofit year, IT energy grows to 28.3 GWh and total electricity is 42.0 GWh. The site reports a simple comparison against the baseline and shows a 5% increase, so the client refuses to pay the ESCO. The ESCO argues that the baseline must be adjusted for IT load and for cooling degree days, which were 5% above the baseline year. The contract is silent on the adjustment method, and the next payment milestone is due in 30 days. Both parties want an independent view before a dispute is escalated.
1. Explain why the simple before-and-after comparison is invalid and calculate an adjusted baseline using the information given. State the assumptions you make.
Read the worked answer
Why simple comparison is invalid. The baseline year had 24.0 GWh IT energy; the post year had 28.3 GWh, an 18% increase. A data centre's electricity use is dominated by IT load, so a change in IT output changes total energy independently of efficiency. Cooling degree days were also 5% higher, which increases cooling energy. Comparing 42.0 GWh with 40.0 GWh mixes a load increase, a weather increase and the effect of the retrofit. It cannot isolate savings.
Separate IT and non-IT energy. Baseline total 40.0 GWh; baseline IT 24.0 GWh; therefore baseline non-IT energy = 16.0 GWh. This includes cooling, UPS losses, lighting and auxiliary systems. The non-IT portion is the part affected by weather and by the efficiency measures.
Adjust for IT load. The post-retrofit IT energy is 28.3 GWh. If the baseline non-IT energy scaled with IT load alone, it would be 16.0 x (28.3/24.0) = 18.9 GWh. But not all non-IT load is IT-dependent; some is fixed. The contract is silent, so a defensible simple method is to adjust only the weather-sensitive cooling portion. Without a breakdown, use the common IPMVP Option C approach: regress baseline total energy against IT energy and cooling degree days, then apply post-period drivers.
Illustrative calculation using the data given. Assume all 16.0 GWh non-IT energy is weather-sensitive for simplicity. Weather adjustment: 16.0 x 1.05 = 16.8 GWh. Adjusted baseline = 28.3 + 16.8 = 45.1 GWh. Actual post energy = 42.0 GWh. Savings = 3.1 GWh, or 3.1/45.1 = 6.9% of adjusted baseline. That is below the 11% guarantee (which would require 4.96 GWh). The simple comparison showed no saving, so the adjustment changes the conclusion from no savings to some savings but a shortfall.
Caveat. The 6.9% figure is illustrative. A proper regression would separate fixed and weather-dependent non-IT loads. The key point is that the baseline must be normalised before the guarantee is tested.
2. Assess the ESCO's claim that it has delivered the guaranteed 11% saving. What further information would you request, and how would you decide the payment?
Read the worked answer
Assessment. On the illustrative adjustment, the project has saved 3.1 GWh against an adjusted baseline of 45.1 GWh, or 6.9%, not 11%. The guarantee is not met on the available data. However, the ESCO is right in principle that normalisation is required; the client's simple comparison is wrong. The dispute is therefore about method and data, not about whether savings exist.
Information to request. (1) The full baseline and post-period monthly data: IT energy, total energy, cooling degree days, chilled-water output, outside temperature and humidity. (2) The baseline regression equation, including coefficients, R², and whether IT energy and degree days are separate variables. (3) A breakdown of non-IT energy into cooling, fans, UPS, lighting and plug loads. (4) Evidence of operational changes: set points, server-room temperatures, occupancy and tariff changes. (5) Metering accuracy and any data gaps. (6) The contract's definitions of baseline adjustment, guarantee and exclusions.
Decision framework. If a properly specified regression shows savings below 11%, the ESCO is entitled to payment for measured savings, not the full guarantee, unless the contract has a shortfall remedy. If the ESCO can show that the shortfall is due to client-caused changes, such as IT load growth beyond an agreed cap or altered set points, the baseline should be adjusted further. If the data are insufficient, appoint an independent M&V professional and withhold the disputed portion, not the whole payment.
Recommendation. Do not simply pay or refuse. Calculate the adjusted baseline transparently, share the workbook with both parties, and use a decision gate: pay the undisputed savings now, hold the disputed amount in escrow pending a joint review. This preserves the relationship and avoids escalation.
3. Recommend how the contract and future M&V should be improved to avoid similar disputes.
Read the worked answer
Contract improvements.
- Define the baseline adjustment method in the contract, not in an appendix. Specify the regression form, variables (IT energy, cooling degree days, maybe occupancy), data frequency and minimum R². State how fixed and variable loads are separated.
- Set a cap or dead band on IT load growth. If IT energy moves beyond, say, ±10% from the baseline without a change in the facility, require a joint baseline review.
- Define the guarantee as a percentage of adjusted baseline, with a clear formula and a shortfall remedy (payment reduction, buy-out or carry-forward).
- Allocate responsibility for data quality, meter calibration and missing data. Specify the independent M&V reviewer and the dispute-resolution process before the project starts.
M&V improvements.
- Meter IT energy separately at the UPS output or rack PDUs, not by subtracting estimated loads.
- Trend cooling degree days from an on-site weather station or a recognised airport station, and use the same source for baseline and post periods.
- Use monthly data with at least 12 baseline points and 12 post points; avoid a single annual comparison.
- Normalise for server load, not floor area. Data centres are not office buildings.
- Include a post-installation commissioning test for the containment and free-cooling, so physical performance is verified independently of the savings calculation.
Governance. Hold a joint baseline workshop before signing, with both parties initialling the data sources and calculations. A small upfront cost avoids a large dispute later. The independent reviewer should be appointed at contract signature, not after the dispute.