Odin Ai uses pressure, rate, treatment, and geology context to calculate CFrac—a live measure of how the fracture system accepts fluid during pumping. That signal ties frac execution to production outcomes while the job is still underway.
Production varies with landing depth, rate schedule, fluid and proppant loading, geology, offsets, and execution quality. Teams usually see the full picture only after capital is spent and the well is on production.
That delay is not a people problem. It is a workflow problem: the field generates rich time-series data during pumping, but most organizations do not convert that behavior into a forward-looking production signal at stage level.
Odin Ai is built for a narrower question engineers can act on: what does this stage’s pressure and rate behavior imply about stimulated productivity, and what should change on the next stage or the next well?
Definition. CFrac (formerly referred to as C-Factor) is Odin Ai’s cumulative fracture compliance metric: it characterizes how much fluid the active fracture system accepts for a given fitted pressure response during pumping, integrated through the stage.
Not the same as shut-in compliance. Traditional fracture compliance is often discussed around closure tests or post-ISIP behavior. CFrac is different in use—it is built for live operations, when rate and pressure are moving and decisions still matter.
Interpretation vs. proof. High or low CFrac suggests how effectively fluid is being taken up during the stage; field studies then test how those patterns line up with production, landing depth, rate quality, and design changes. The case study walks through that separation with Halo data.
Visuals from our Montney conference work illustrate how stage-level CFrac is compared across wells and tied back to placement and execution. The full narrative, blind test, and landing-depth examples are on the case study page.
In Halo’s dataset, cumulative CFrac showed a statistically significant relationship with six-month production. A reserve well was held back from model training; the resulting forecast was within a few percent of actual six-month oil rate—evidence the signal is not only curve-fitting history.
At the Saga SÖKKVABEKKR conference in Montréal (2025), Odin Ai presented Montney results using this approach. The same underlying methodology has since been reproduced in the Haynesville, with additional projects in development across U.S. plays including the Wolfcamp, Eagle Ford, Woodford, and Marcellus—different mechanics and completion styles, same rate–pressure discipline.
Rate behavior and landing depth emerged as major drivers of performance divergence in Halo’s analysis—exactly the kind of operational levers completions and reservoir teams already control.
Charts like this support buyer due diligence: does modeled CFrac move with production the way theory says it should, in a field where geology and completion style match your questions?
High-frequency frac data in. Stage-level CFrac and forecasts out.
Inputs. Pressure, rate, and treatment data from standard surface feeds; geology and well context where available.
Process. Physics-informed calculations—trained from large sets of manual rate–pressure work—generalize the relationship between pumping behavior and CFrac so it survives field noise and non-linearities.
Outputs. Stage-level CFrac time series, well aggregation, production forecasting where calibration data exists, and diagnostic views for landing depth, rate schedule, and design comparison.
Relate cumulative CFrac to producing outcomes so type curves and pad forecasts can be informed earlier—then stress-tested with blind wells where policy allows.
See when stage behavior lines up with landing relative to the pay target, and when porpoising or barrier breakthrough changes the response during pumping.
Quantify how rate ramps and interruptions change CFrac development in otherwise similar rock—so execution issues become measurable, not anecdotal.
Roadmap: combine CFrac growth targets with screen-out risk models so crews can push productive stimulation without crossing operational limits.
Use live CFrac growth alongside cost and schedule to judge whether continued pumping on a stage is justified before the treatment is closed.
SkyFrac is built on a founder-developed portfolio of granted patents and pending applications spanning hydraulic fracturing optimization, real-time analytics, and related sensing concepts—not a single slide deck.
The work is grounded in stage-level studies, operator-facing validation, and collaboration with experienced reservoir and completions practitioners.
The first commercial CFrac neural net generalized thousands of hand-checked rate–pressure calculations into a model that runs on live frac data—intended for engineers who care about repeatability, not novelty.
Start with a technical review if you want to validate CFrac on representative wells, understand outputs, and map data requirements.
Choose a commercial discussion if you are ready to talk deployment scope, economics, and integration with existing workflows.