Same data, two uses.

The case for pricing construction risk from site data is now being made seriously, in casualty. On a physical loss and damage policy, the same data has a second use that does not wait for claims to develop.

Shepherd, a US construction insurance MGU, has published a clear and careful argument for using jobsite data in commercial underwriting. It is worth reading in full, and it is linked below.

Its starting point is the US firm Progressive. In personal auto, Progressive built a pricing advantage from data other carriers were not using, starting in the early 1990s with credit history and moving on to telematics. By the time the FTC reported to Congress in 2007, credit-based scores were in use across most of the auto market.

Shepherd's argument is that construction now has the same opportunity. A great deal is knowable about how a contractor actually runs a job: safety records and OSHA history, inspections and near-miss reporting, camera and sensor data, its record on comparable projects. But, as the piece puts it, "Most of it sits in separate software platforms that underwriters never see."

What the pricing case rests on

Shepherd's piece is about construction casualty, and it is plain about the difficulties. Accounts vary, claim frequency on lines like general liability is low, and the tail can run a decade. An underwriter may not find out whether a pricing decision was sound for seven to ten years, and, in the piece's words, a flawed model "can look like a good one for years." It is explicit about the pace: "Anyone claiming a data edge in this line should be modest about how fast it proves out."

What has changed, it argues, is time. Enough projects that ran on these platforms have now finished and moved through the full tail, so Shepherd can match how a job was run to how its claims developed. Ten years ago, it says, there was not enough history to run that study. Shepherd Savings, its programme of premium credits for contractors and owners using partner site platforms, is built on it: a platform only qualifies once it has passed a study against Shepherd's own loss experience.

That is the shape of the pricing use. It looks back across many finished projects, and it gets stronger as more of them close out.

Physical loss and damage is a different shape

Pikt works on the physical loss and damage side of construction insurance: construction and erection all risks (C/EAR) and Builders' Risk. Shepherd writes US Builders' Risk too. Under its programme with Brickeye, builders receive a tiered protection classification based on the contracted scope of monitoring, which can earn premium credits or lower deductibles. It is one of the clearest examples of site monitoring already earning better terms.

Physical damage claims develop far faster than casualty claims. What runs long on project business is the period of cover, because most builds take several years. But much of the data is the same, and so is the problem. Most of it never reaches the people who insure the project.

Two things make it different.

The first is that the insured risk is the physical project itself, and it changes every day. Materials arrive and wait. Plant is installed. The building is closed in, or is not. Temporary power goes live, fire protection catches up with the fit-out, or falls behind it. The values at risk build up week by week, and what is exposed this month is not what was exposed at placement.

The second is counting. In our view, physical damage has its own reason why a model built only on claims learns slowly. The frequent losses are largely retained below the deductible, where few outside the project see them in detail. The severe losses reach the policy, and they are too rare to count.

The use on the live project

Site data does not have to reach a price to be useful. It can reach a risk engineer now.

A risk engineering fee buys a finite amount of an expert's attention, spread across the visits and reviews a project needs. Between those points the site keeps moving, and what reaches the engineer arrives second-hand and after the fact. The judgement is sound. The information it works from was true when it was written.

On a C/EAR or Builders' Risk project, the same cameras, sensors and walk-throughs that a pricing model learns from, together with the project's IFC/BIM models and progress tracking, can answer the risk engineer's questions today. What arrived on site this week, and where it is stored. What has been installed, and in what order. Whether the building is closed in before the weather turns. What a moisture sensor read overnight. Which areas have changed since the last visit, and which have not moved at all.

None of that depends on a loss history. It needs no closed projects and no developed claims. It needs one live project, the record of what it is doing, and an expert to look at it.

The value of that lands during the build, which is when a physical damage policy is exposed. Attention goes to the conditions that matter now, while they can still be acted on, rather than being reconstructed afterwards. It is more risk engineering from the same fee.

Two questions

The two uses ask different questions.

The pricing use asks: across many projects, did sites run like this one have fewer or smaller losses? That can only be answered in aggregate and in hindsight, from projects that have finished and claims that have developed.

The risk engineering use asks: on this project, what is happening now? On a physical damage policy, that is the question that runs through the whole period of cover, and it can be answered every day of the build.

They are not competing. The pricing case compounds, as Shepherd says, slowly at first, and those who build it carefully will have something valuable. But on every live project already running site technology there is also a return available from the same data, through a function the market already pays for, and it does not wait for the project to finish.

What this does not say

It does not say continuous visibility reduces losses as a settled fact. That is a proposition with a precedent in other lines, not a settled finding in construction. And it does not say anything about how any underwriter should price a risk.

What it does say is that site data is not only an input to a price, which needs closed projects and developed claims to validate. On a C/EAR or Builders' Risk policy it is also a record of the insured risk as it changes, and the risk engineer can use that record today.

Pikt connects to the technology already running on site and gives the risk function continuous visibility of the project. What a site shows is recorded, not assessed: what a source said, and when.