The 411 on Accuracy
Practice makes perfect.
- Problem: Industry requires accuracy to make results actionable.
- Solution: Early adopters help us validate every new algorithm.
- Benefit: Accurate, actionable alerts.
At Satelytics, we strive not only to expand our capabilities but also to perfect them. Our science team doesn’t release a new algorithm unless it is within 10% ensuring our clients receive high-quality, actionable data. And with machine learning, these algorithms become increasingly accurate, maturing like a fine wine.
Our algorithms analyze the infrared wavelengths of imagery from satellites and other data-capture platforms. Once an algorithm is built, our artificial intelligence experts go to work building models that ingest enormous datasets, often terabytes for singular projects. Machine learning and artificial intelligence allow the already excellent accuracy to improve with every pixel of analyzed data. This is how we continuously refine our measurements for constituents like phosphorous and metals to be accurate within a few parts per billion.
Satelytics has had the privilege of working with several early adopters to develop new algorithms, demonstrate The Satelytics Difference, and prove the accuracy of our algorithms. A methane analysis conducted for a large utility customer this past summer correctly identified 98.5% of leaks over the target area, and flow rate accuracy was within 6% of the ground-measured value. Similar results had been gained for liquid pipeline leak detection in the Bakken in 2020: 98.93% accuracy.
Practice makes perfect, and the same goes for our algorithms! Today we have over 40 algorithms, ready to meet a variety of industry needs — all with accuracy.