Owners are spending more on jobsite documentation than at any point in this industry’s history, and most of them still cannot tell you with confidence whether a subcontractor is on schedule. In most of the owner project reviews I am involved in, somebody always asks how a scope of work is tracking. Then there is always a short pause. Then two or three people give slightly different answers, none of them dishonest, and the meeting moves to the next item because nobody in the room has a better instrument than the people in the room.

What is strange about that pause is how well documented the project usually is. Weekly 360-degree walks, drone flights a couple of times a month, laser scans at the milestones, and somewhere a folder holding forty thousand photographs. Every square foot of that building has been recorded, and still, when the question gets asked out loud, the honest answer is a range.

The distance between capturing what is happening on a site and understanding it well enough, and fast enough, to do something about it is not a data problem in the way our industry usually means that phrase. There is plenty of data. What is missing is the step between the data and the decision, and that step is still being performed by a person with a spreadsheet on a Sunday night.

The underlying mistake is one a lot of us made in the first wave of reality capture. We treated the record as the outcome. If the site was fully documented, we assumed the visibility problem was handled. It is not. Forty thousand photographs are an archive, and an archive only answers questions you already knew to ask. It will not tell you, for instance, that drywall on three floors has been quietly running four days behind the curve your schedule assumed since the second week of August. That surfaces later, usually in a conversation about liquidated damages.

Autodesk and FMI surveyed more than 3,900 construction professionals for their 2021 report, Harnessing the Data Advantage in Construction, and only 12 percent said they always use project data when making decisions. The same study pinned 14 percent of all rework in 2020 on bad data, which it valued at close to 89 billion dollars. An earlier FMI study put the share of captured data that goes entirely unused at 95.5 percent. Those studies are five and eight years old. Everything about capture has improved since then, and I am not convinced the decision layer has moved much at all.

Capture hardware got cheap enough that it stopped being a decision and became a default, so the volume coming off every site keeps multiplying whether or not anyone has a plan for it. At the other end, the large project management platforms have started buying reality capture companies outright, at prices that make very little sense if the cameras are the point. They are not paying for the cameras. They are paying for the layer that sits above the imagery and turns it into an answer, because that is where the defensible position is. Capture is becoming a commodity input, roughly the way GPS did, and the margin moved up a level while most owners were still procuring capture as a line item.

The owner holds the real exposure on a capital project. You sign the guarantee, you carry the cost of capital, you answer to the lender, and you absorb whatever a late certificate of occupancy does to a lease-up schedule. Then you manage that exposure using the least rigorous instrument anywhere on the job, which is a narrative report, assembled by hand, describing conditions already a week old, written by the party whose performance it describes. I am not suggesting bad faith here. Most of the people writing those reports are doing careful work under real time pressure. But no lender would accept a borrower’s own summary of the collateral and call it diligence. It is worth asking why we accept the equivalent on a project.

Most of this is fixable in the contract rather than the budget. The most immediately actionable changes are specification changes, and they are worth the argument. The first is specifying what capture has to produce rather than just how often it happens. Most owner agreements set a documentation cadence and stop there, but cadence without a latency requirement only fills the archive faster. Agreeing on specific metrics, in a specific format, within a defined window after each capture is the step most contracts skip.

It is also important to determine definitions before anyone actually pours concrete. Things like: what percent complete does a job need to be before it is considered complete” What observable conditions constitute completion for each major milestone? These answers then need to be tied back to a schedule activity ID. Nearly every pay application dispute traces back to two parties who never agreed on a definition and then tried to negotiate one under time pressure with money already in dispute. Putting time-stamped visual evidence on the pay application as a standard exhibit gets read as adversarial, but most draw disputes are not about bad faith.

Owners running multiple active projects should also standardize across the portfolio, because six projects with six reporting formats and six working definitions of complete produce no real ability to compare risk until one job fails loudly enough to get attention. Best practice is also to specify in writing that the capture record belongs to the owner and survives a change of contractor.

If you take one thing from this, change the number you manage to. Stop reporting on how much of the site was captured last month, and start tracking how many days pass between something happening physically on site and someone with authority acting on it. That number might be unflattering the first time you measure it, which is exactly why it is useful.

There is real movement toward agentic systems that do not simply report but flag, forecast, and initiate, and I expect that to arrive faster than most capital programs are prepared for. But an agent pointed at an unstructured pile of photographs will produce confident, well-formatted guesses, and a confident guess inside a draw approval workflow is worse than no automation at all. Whatever autonomy turns out to be useful will need a structured, schedule-linked, consistently defined record underneath it. The owners sorting out their definitions and their cadence this year are doing the unglamorous work that makes the rest of it possible. The next competitive layer in this business is not going to be more cameras. It is going to be how long it takes for what a building knows about itself to reach the people who are financially responsible for it.

The post Owners Are Drowning in Jobsite Data and Still Guessing at the Schedule appeared first on Propmodo.

​