Data Centres Industry Pack

Data Centre Construction Cost & Schedule Forecasting

Data centre programmes are governed by long-lead equipment and commissioning gates, not by civils. A forecast built on trade progress will look healthy right up until the switchgear slips.

How the engine works

Typical project-controls challenges in data centres

The method is the same everywhere. These are the sector conditions that make a generic approach wrong, and what each one means for how you control the project.

Long-lead equipment sets the date

Generators, chillers, UPS and switchgear carry lead times measured in quarters. Once ordered, no amount of site resource moves the date, and the whole programme keys off delivery.

Site productivity cannot recover a lead-time slip

Cost is measured per MW, not per m²

Buyers, funders and operators all think in £/MW of IT load. A forecast expressed in total cost alone cannot be benchmarked by anyone who matters.

The wrong denominator kills comparability

Commissioning L1–L5 is where programmes die

Integrated systems testing routinely uncovers issues that individual system tests passed. It is the least predictable phase and the one most often given a fixed duration.

L4/L5 needs a distribution, not a bar

Phased handover means partial revenue

Halls are handed over progressively. Forecasting a single completion date ignores the revenue timing that justifies the investment.

Per-hall dates drive the business case

Escalating power density mid-build

Client IT load requirements change during construction. Late density increases ripple through cooling, electrical distribution and floor loading.

A scope change disguised as a spec update

Which capabilities apply

Every tool below is a separate workbook that works on its own. These are the ones data centres teams tend to reach for, and what each one answers.

Analyse

Whether the programme you were sent is built properly, before you rely on it
Where the critical path really runs, and how much float is protecting it

Schedule Quality Analyser

Control

Cost and schedule performance against the baseline, in CPI and SPI
Change and risk exposure on one register, not in email

The core suite

Forecast

Outturn cost and completion date as a range, not a single number
The contingency needed to be 80% confident, not 50%

Forecast Engine

Measure

Progress claimed against rules of credit rather than opinion
An update block you can paste straight back into P6

Progress Measurement System

Contract & claims

Whether a submitted programme meets NEC4 clause 31.2
Productivity loss measured against an unimpacted period

NEC4 Compliance Disruption & Productivity Loss

Sustainability

Capital carbon baseline and forecast against target
Evidence trail for PAS 2080 assurance

Carbon & PAS 2080

How it runs, month to month

Export two tables from P6 or Microsoft Project. An activity list and a relationship list. Nothing in your scheduling tool changes.

Paste them in. The shaded cells are the only inputs. Everything else is formula-driven.

Check the programme is sound. The schedule quality scorecard runs the DCMA checks and tells you what to fix before you forecast from it.

Enter the period progress and cost. Earned value, actual cost and planned value by period.

Read the forecast and issue the report. CPI, SPI, EAC and the P50 to P80 range, with the assumptions recorded alongside them.

What's in the Data Centres edition

The engine is identical across industries. The edition supplies the content that makes it speak your sector's language, so it is useful the first time you open it rather than after a fortnight of setup.

WBS skeleton

Shell & core
Electrical infrastructure
Mechanical & cooling
White space fit-out
Commissioning L1–L5
Handover by hall

Cost codes

£/MW IT load
Generators & fuel systems
UPS & battery
Chillers & CRAH units
Switchgear & busway

Risk register

Long-lead delivery slip
Grid connection delay
IST failures at L4/L5
Client density change
Commissioning agent availability

Gate milestones

Power-on / energisation
L1 factory acceptance
L3 individual systems
L4 integrated systems
L5 load bank & handover

Worked example

24MW colocation build, month 11 of 18

BAC £310M. EV £171M, AC £178M, PV £174M. Switchgear delivery has slipped 7 weeks; halls 3 and 4 handover is at risk.

CPI

0.96

£/MW forecast

£13.4M

P50 handover

Month 19

P80 handover

Month 21

Cost performance looks fine at 0.96, the problem is entirely schedule. Two months of delayed handover on halls 3 and 4 is the real number, because it is revenue, not cost. A cost-only forecast would have reported this programme as healthy.

Illustrative figures, shown to demonstrate the method. Your numbers replace them.

Reports and outputs

Each of these is a sheet in the workbook. They recalculate when you enter the period progress, so there is nothing to assemble at month end.

Cost forecast

Three EAC methods side by side
CPI, SPI, cost and schedule variance
Variance at completion and TCPI

Monte Carlo

1,000 iterations, no macros
P50, P80 and P90 outturn cost
Forecast duration and delay against baseline

Earned Schedule

Time-based schedule performance
Forecast completion date
Stays honest late in the job, unlike SPI

Cash flow and resources

Payment lag and retention modelled
Peak funding requirement
Resource forecast by period

Productivity

Trend against planned rate
Early warning before CPI moves

Assumptions and audit

Every distribution, its source and date
The record that defends the number

Data Centres questions

Can it forecast per hall rather than per project?

Yes. Each hall is a separate forecast unit with its own handover date, which is what the revenue model needs.

Does it include £/MW benchmarks?

The pack ships with placeholder benchmark ranges. Replace them with your own build data, your numbers beat any published average.

How is commissioning modelled?

L4 and L5 carry triangular distributions rather than fixed durations, because integrated testing is where the variance actually lives.

Does it track grid connection?

As an external dependency milestone with its own risk profile, since it is rarely within your control.