Delayed, denied, or withdrawn, and how long, for infrastructure projects facing a local decision body.
We simulate the community around the decision. Scored as of the date you ask.
Delay-in-start-up underwriters, private credit, and project-finance lenders.
Data Center Watch, Q1 2026.
Sample, invented data
Developer: Example Infrastructure LLC
Delay past the jurisdiction's normal approval time by more than 6 months
30–50%
Months past the normal approval time
Invented project and figures. Not a registered forecast.
Drivers behind each number
Projects rarely get rejected. They get delayed.
Nobody prices the months.
Source: Data Center Watch, first quarter 2026. Cited for market context. Not used as a modeling input.
We use modern AI to build the simulation, and we keep it out of the number itself. The probability comes from a method we can explain, check and score.
Opposition is not a fixed quantity you can look up. It spreads. Each resident's view stays fixed. What changes is who acts: one neighbor brings another to the hearing, a local group turns concern into turnout.
So we simulate it. We build a population of the community around a decision, map who influences whom, and run how opposition organizes and spreads through that network. The simulation is added to the forecast. AI helps build the simulation and never produces the number.
Scores registered before the outcome. Every case stays, including the misses.
No entries yet.
Wherever these decisions leave a public record, we can register a forecast before the outcome and score it when it resolves.
We forecast whether a decision making body approves, delays or rejects, not what the price should be.
Something else? Tell us what you are looking at. We will tell you straight whether the public record is there and what it would take, before anyone commits to anything.
One project. One approving body.
Project finance lenders and private credit
Whether schedule risk is priced into the facility
Advisors and consultants
A defensible number for a client file
Owners and operators are routed through the form, not marketed to.
A founding pilot runs a history test on your own past projects, chosen by a rule we agree upfront. Each assessment is made as of its own date and scored against what actually happened. It runs six to eight weeks. Delivery is concierge: a person does the work and signs the output.
Founding slots are limited.
SOC 2
Whether a decision making body approves, delays or rejects your project, and the likely delay. Not what the price should be.
We are preparing our first evaluation on resolved cases. The range shows how much each estimate moves under different settings; it does not measure how reliable the estimate is.
A founding pilot runs six to eight weeks. We reply to every inquiry within two business days.
The project details, and the decision you want forecast.
Founding pilots are priced per engagement. We quote after we agree the scope.
No language model produces the probability. Language models help prepare inputs. A statistical method produces the number; the simulation's results enter it as one more input.
No. Your documents are not used to train AI and are not shared with any AI provider for training. We use AI providers only under zero data retention agreements.
Yes, in transit and at rest.
Only the team, with MFA.
A person delivers each score card directly to you. There is no shared platform.
No. If a developer or operator is our client, we do not score their projects for a lender.
Tell us what you are looking at. We will tell you straight whether the public record is there and what it would take, before anyone commits to anything.
Submissions are reviewed after 90 days and deleted unless there is an active engagement. Consent records are kept 36 months.
Have a project in front of a board, a commission, or a county? Book a 20-minute call. A person will reply within two business days.
