How do you test out GEO satellite data prior to launch? Synthetic data.
Sometimes we’re asked if it’s possible to put our sensor on an airplane or drone to prove impact in advance of launch. The technical answer is, yes we could put it on a drone, but no, it won’t prove impact. The value of our data will be its temporally persistent cadence. Always-on data. By definition, you cannot prove that with a short duration drone or aerial flight. You could prove the sensor, but the sensor itself, while cool, is not the game changer. Persistence is.
So how do we prove the value of persistent data?
We build synthetic data.
Our synthetic data consists of physics-based simulations of infrared scenes, calibrated to Fire Sentry sensor specifications and run through the same pipeline that will process the live feed Same formats. Same cadence. Same plumbing that will need to be built.
Our Synthetic Data Program is organized in three phases, each answering a different question:
Detection validation. How sensitive will the system be across intensity, atmosphere, lighting, and background? Partners receive synthetic infrared snapshots, sensitivity analysis, detections, and temperature maps to test against their own risk appetite.
Time series. What does a fire look like when it can be watched continuously? This phase covers progression modeling and multi-frame processing, and it is where GEO stops being a snapshot and becomes a record.
Mitigation modeling. What is early detection worth? Partners will be able to do outcome comparisons on fire size, spread rate, and loss, with and without Fire Sentry data, calibrated to their geographies.
Beyond the phases, synthetic data is useful for integration work, model architecture and feature engineering ahead of live data, internal business case development, and procurement evaluation on a real artifact rather than a datasheet. (Datasheets are lovely. They are also not data.)
Besides the obvious case of using this data to prove the value of GEO, why else embark on this program?
Fire Sentry has not launched. Pathfinder, our precursor validation mission, is in progress. If you are an insurer, a utility, or a platform that will eventually build on this data, that can sound like a long runway. It is not. It will fly by.
Consider what a team has to do before a new data feed is useful: evaluate the data, build ingestion pipelines, integrate it into models and workflows, and navigate internal governance.
None of that can start against a data product that does not exist. Wait for orbit, and the first operational fire season goes to plumbing instead of results.
Nobody asks a pilot to learn on a first flight with passengers. They use a simulator. We are leveraging the same idea.
Why now
The timing matters as much as the data. The Objective system requirements are not frozen. Partners working with synthetic data today help specify what an insurance-grade and utility-grade wildfire product should look like, while that specification can still change the satellite. After launch, you are a customer. Before launch, you are a collaborator.
This also reflects how Orbital Sentry is being built. Insurers, reinsurers, and utilities fund the platform, and qualifying first responders will receive detection alerts at no cost. Faster suppression means smaller fires, and smaller fires mean lower losses. The commercial case and the public case are the same case.
A drone can prove a sensor. It cannot prove persistence. Synthetic data lets partners test the thing that matters: what changes when a fire is watched continuously instead of photographed occasionally.
To be clear about what this is, synthetic data is not observation. It is not suitable for filed rate support, reserving, or anything that requires observed loss experience. What it will tell you is whether your systems, models, and workflows are ready when real data arrives, and what that data could be worth to your organization.
Interested in learning more about our synthetic data program? Reach out!
