Data center schedules don’t forgive much anymore.
Power-on dates get committed early, tenants move in fast, and every trade on site is working against a critical path that can’t afford a bad week. When schedules slip, it’s almost always because the data feeding them is wrong or late, and that is the problem most construction tech still has not solved.
The cost of getting this wrong on a data center build is hard to overstate. Over 60% of data center outages trace back to decisions made during construction, decisions rarely captured in any auditable form. At $500M+ per build, industry rework rates of 5 to 10% translate to $25 to $50 million of exposure on a single project. Nearly half of that rework traces back not to bad workmanship, but to bad data.
That is the data primacy problem: the field-captured record of what actually happened is rarely the record the project runs on. The inspection report gets written up the next day. The binder gets assembled at turnover. The spreadsheet gets reconstructed from memory. By the time anyone needs to act on the information, or prove what was done, the data has been transcribed, summarized, and degraded.
Why Pilots Keep Failing
It is also why the construction tech pilot keeps failing in the same pattern. A tool gets introduced, a superintendent champions it, that person rotates off, and the tool fades with them. The real issue is that most pilots are designed to prove the technology works, not to make the data it generates the authoritative record of the job. Without that, the tool is optional. And optional tools don’t survive a personnel change.
When the Data is the Record
The tools making a real impact on data center projects today are not flashy: clash detection, material tracking, reality capture, field data platforms. None of it is new. What has changed is the cost of getting it wrong, and the expectation that the data these tools produce is the record the project runs on, not a parallel artifact alongside it.
Data primacy means capturing data at the point of work, not on a clipboard to be transcribed later. The torque value, test result, and inspection status are recorded in real time, at the connection, by the person doing the work. That fidelity is what makes the data trustworthy enough to run a job against, and, eventually, the contractually controlling record of what happened.
When the record comes directly from the work itself, issues surface earlier, handoffs carry the right context, and the schedule reflects what is actually happening in the field. If a dispute arises about whether work was done correctly, you don’t argue over who remembers what. You look at the data. If a connection fails years later, you don’t dig through filing cabinets. You pull the digital record that shows exactly what was done, when, by whom, with what tool, at what torque value.
That is what defines a smart jobsite. Not the tools. The data those tools generate, treated as the foundation the whole job runs on. It’s the principle behind our Quality Execution System at Cumulus, and we believe it’s where construction quality is headed across the industry.
