Back to Insights

Completions & Frac Software: Design, Monitoring, and Optimization Platforms

Dr. Mehrdad Shirangi | | Published by Groundwork Analytics LLC
Editorial disclosure

This article reflects the independent analysis and professional opinion of the author, informed by published research and professional experience. No vendor reviewed or influenced this content prior to publication.

Completions engineering in unconventional plays is where the most money gets spent and the least certainty exists. A typical Permian Basin horizontal well costs $8-12 million to drill and complete, and somewhere between 40% and 60% of that cost is in the completions phase -- the pumping, the proppant, the fluid, the perforating, and the equipment on location. The difference between a well that pays out in 18 months and one that takes four years often comes down to completions design decisions made weeks before the frac crew arrives.

The software that supports these decisions has evolved significantly, from simple 2D fracture models in the 1980s to complex 3D planar and fully 3D simulators that attempt to capture the mechanics of hydraulic fracture propagation through layered, naturally fractured rock. But despite the sophistication of the modeling tools, completions engineering remains one of the most empirical disciplines in petroleum engineering. Engineers rely heavily on offset well performance, trial-and-error optimization, and institutional knowledge because the subsurface is simply too uncertain for any model to capture completely.

This article surveys the frac design, monitoring, and optimization software landscape, assesses where AI is making inroads, and identifies where the real opportunities lie.

Frac Design Simulators

Frac design simulators solve the coupled problem of hydraulic fracture propagation, fluid flow, and proppant transport. They predict the geometry (height, length, width) of hydraulic fractures created during stimulation, the distribution of proppant within those fractures, and the resulting fracture conductivity.

Halliburton: FracPro and Gohfer

Halliburton's completions software portfolio includes two major frac design platforms with distinct heritage:

FracPro (originally developed by Resources Engineering Systems, later Pinnacle Technologies, then Halliburton) is a 3D planar fracture simulator that models fracture propagation, fluid leakoff, proppant transport, and production performance. FracPro has been an industry standard for decades, particularly in conventional completions and tight gas applications.

FracPro's strengths include a robust proppant transport model, well-validated leakoff calculations, and a production model that estimates post-frac well performance based on the predicted fracture geometry. The tool is widely used for both design (predicting fracture geometry before pumping) and analysis (calibrating models to treatment data to understand what the fracture actually looked like).

Gohfer (Grid Oriented Hydraulic Fracture Extension Replicator) takes a fundamentally different modeling approach. Rather than assuming a planar fracture with an idealized geometry, Gohfer uses a grid-based finite difference approach that can model complex fracture geometry, including non-planar propagation, fracture turning, and natural fracture interaction.

Gohfer is particularly valued for its ability to model near-wellbore effects (fracture initiation from perforations, tortuosity) and its stress analysis capabilities. The tool can import mechanical earth models and predict fracture behavior based on the in-situ stress field, which is important in environments with complex stress conditions or natural fracture systems.

Strengths (FracPro): Long track record, well-validated models, widely understood output formats, strong production modeling integration.

Strengths (Gohfer): More physically realistic fracture geometry modeling, natural fracture interaction, stress analysis, near-wellbore effects.

Limitations: Both tools model individual treatment stages. Modeling the interaction between multiple stages along a horizontal well -- a critical factor in unconventional completions -- requires additional complexity that neither tool fully captures. The models are deterministic: they produce a single predicted fracture geometry for a given set of inputs, without explicit uncertainty quantification.

NSI Technologies: StimPlan and NETool

NSI Technologies provides StimPlan, a 3D fracture design simulator, and NETool, a discrete fracture network (DFN) simulator for modeling complex hydraulic fracture networks in naturally fractured reservoirs.

StimPlan is a fully 3D fracture simulator that models fracture propagation in layered media with variable mechanical properties and stress contrasts. StimPlan's architecture allows for detailed layer-by-layer modeling of rock properties, which is important in formations with significant vertical heterogeneity (Wolfcamp, for example, where different benches have very different mechanical properties).

NETool models hydraulic fracture network complexity -- the interaction between induced hydraulic fractures and pre-existing natural fractures. This is critical in formations like the Woodford, Barnett, and Niobrara where natural fractures significantly influence stimulation effectiveness and drainage patterns.

Strengths: Detailed layer-by-layer mechanical modeling (StimPlan), natural fracture network modeling (NETool), independent from any service company, focused exclusively on frac modeling.

Limitations: Smaller user base and development team compared to Halliburton or SLB tools, less integration with broader subsurface modeling workflows.

SLB: Kinetix and Mangrove

SLB's completions software suite includes:

Kinetix -- A reservoir-centric fracture simulator that models hydraulic fracture propagation and its interaction with the reservoir. Kinetix is integrated with Petrel, allowing fracture models to be built within the same geological model used for reservoir simulation. This integration is Kinetix's primary advantage -- the fracture model inherits the geological model's mechanical properties, stress field, and natural fracture characterization directly.

Mangrove -- SLB's unconventional completions design platform within Petrel. Mangrove provides a workflow-oriented approach to completions design that includes stage placement optimization, cluster spacing design, and treatment schedule optimization. Mangrove combines simplified fracture models with statistical analysis of offset well performance to guide design decisions.

Strengths: Deep integration with Petrel and the broader SLB subsurface modeling ecosystem, reservoir-centric fracture modeling, workflow-oriented design tools (Mangrove), access to SLB's global completions database for benchmarking.

Limitations: Works best within the Petrel environment -- operators who do not use Petrel for their subsurface modeling will find the integration less compelling. Kinetix is computationally demanding for complex models. Like other SLB software, pricing reflects the integrated platform value.

Real-Time Frac Monitoring

The shift to pad drilling and simultaneous operations has increased the importance of real-time monitoring during hydraulic fracturing. The ability to observe what is happening downhole during a treatment -- and to adjust the design in real time based on those observations -- has become a significant competitive differentiator.

Fiber Optic Monitoring (DAS/DTS)

Distributed acoustic sensing (DAS) and distributed temperature sensing (DTS) using fiber optic cables deployed in the wellbore have become standard monitoring tools for unconventional completions in technically advanced operations. Fiber provides continuous measurements along the entire wellbore, showing fluid distribution across clusters, fracture initiation from individual clusters, and inter-stage communication.

Several companies provide fiber-based frac monitoring services and interpretation software:

  • SLB -- Through their fiber optic sensing solutions, integrated with Kinetix for real-time model calibration
  • Halliburton -- Fiber optic monitoring services with interpretation tools
  • OptaSense (Luna Innovations) -- Stand-alone DAS monitoring and interpretation, widely used by operators who want vendor-neutral fiber interpretation
  • Silixa -- DAS/DTS monitoring with specialized interpretation software

The data volumes from fiber optic monitoring are enormous -- a single frac stage can generate gigabytes of DAS data. This creates both an opportunity (dense data for AI analysis) and a challenge (data management, storage, and processing infrastructure).

Pressure Diagnostics

Real-time pressure analysis during fracturing provides information about fracture geometry, complexity, and interaction with offset wells:

  • DFIT (Diagnostic Fracture Injection Test) analysis -- Pre-frac mini-tests that provide closure stress, pore pressure, and leak-off coefficient estimates
  • Net pressure analysis -- Comparing actual treating pressure to modeled pressure to infer fracture geometry in real time
  • Frac hit monitoring -- Measuring pressure responses in offset wells (parent wells) to detect fracture communication and estimate stimulated rock volume

These analyses are typically performed using the frac design simulators (FracPro, StimPlan, Kinetix) in real-time mode, though some operators use specialized diagnostic tools or custom analysis workflows.

Microseismic Monitoring

Microseismic monitoring -- detecting the small seismic events generated during hydraulic fracturing -- was the dominant frac monitoring technology of the 2010s. Companies like MicroSeismic Inc., ESG Solutions (now Pinnacle/Halliburton), and SLB's microseismic services provided real-time mapping of hydraulic fracture growth.

However, the use of microseismic has declined significantly in recent years, replaced in many applications by fiber optic monitoring. The reasons include cost (surface microseismic arrays are expensive to deploy), interpretation uncertainty (microseismic event locations have inherent positional uncertainty that can be larger than the features being mapped), and the emergence of fiber optic alternatives that provide complementary information at lower cost.

Microseismic remains relevant for specific applications, particularly for monitoring large-scale stimulations in formations with complex fracture systems, and for induced seismicity monitoring related to regulatory requirements.

Completions Analytics and Optimization

Beyond frac design and real-time monitoring, a growing category of software focuses on completions analytics -- using historical data from completed wells to inform future completions design decisions.

The Data-Driven Approach

Completions analytics platforms attempt to answer the question that every completions engineer asks: given the well's geological setting (landing zone, mineralogy, TOC, stress), what completions design (stage length, cluster spacing, proppant loading, fluid volume, proppant type) will maximize well performance?

This is fundamentally a supervised learning problem, and several companies have built products around it:

Novi Labs -- Provides AI-based well performance prediction and completions optimization. Novi's platform uses machine learning models trained on large databases of well completions and production data (primarily from public state databases) to predict how changes in completions design will affect well performance.

SLB (Mangrove/Stimulated Reservoir Volume analytics) -- SLB's completions analytics combine their proprietary databases with physics-based models within the Petrel environment.

Liberty Energy -- Liberty has invested in data analytics capabilities that leverage their extensive frac treatment database (Liberty is one of the largest pressure pumping companies in North America). Their analytics inform both completions design recommendations for operators and Liberty's own operational efficiency.

ProPetro -- Similar to Liberty, ProPetro leverages its treatment database for performance analytics, though with a smaller scale and less publicized analytics capability.

The Problem with Pure Data-Driven Completions Optimization

The data-driven approach to completions optimization faces a fundamental challenge: confounding variables. Well performance depends on geological factors (reservoir quality, stress regime, natural fractures), completions factors (the design variables the engineer controls), and operational factors (execution quality, frac fleet performance, fluid quality). Separating the geological signal from the completions signal in historical data is extremely difficult.

A machine learning model trained on public production data may find that wells with higher proppant loading produce more oil. But is that because more proppant is better, or because operators tend to pump more proppant in better-quality rock? Without careful control for geological variables, data-driven models can confuse correlation with causation, leading to design recommendations that are geological artifacts rather than completions insights.

The best completions analytics combine data-driven pattern recognition with physics-based understanding. A model that incorporates rock mechanical properties, stress gradients, and fracture mechanics can separate geological effects from completions effects more reliably than a pure statistical approach.

The AI Opportunity in Completions

Opportunity 1: Real-Time Treatment Optimization

The highest-impact AI application in completions is real-time treatment adjustment. During pumping, the frac engineer observes treating pressure, pump rate, proppant concentration, and (if available) downhole measurements from fiber or gauges. Based on these observations, the engineer makes decisions: adjust the rate, change proppant concentration, extend or cut short the stage.

These decisions are currently made based on experience and judgment. An AI system that integrates real-time treating data with a calibrated fracture model and offset well performance data could provide data-driven recommendations for treatment adjustments. This is a constrained optimization problem that is well suited to machine learning: maximize expected well productivity subject to operational constraints (maximum treating pressure, available proppant, pump capacity).

The challenge is speed -- the system must process data and provide recommendations in real time, during the treatment, which limits the complexity of the models that can be used.

Opportunity 2: Cluster Efficiency Optimization

Fiber optic data has revealed that cluster efficiency in multi-stage horizontal completions is often poor. In a typical 5-cluster stage, two or three clusters may take the majority of the fluid and proppant, while the remaining clusters are understimulated. This non-uniform distribution reduces the effective stimulated area and leaves recoverable hydrocarbons behind.

AI can address this in two ways: by predicting cluster efficiency based on wellbore conditions and completion design (informing stage design before pumping), and by identifying real-time indicators of poor distribution (informing mid-treatment adjustments like diversion or rate changes).

Opportunity 3: Parent-Child Well Interaction

Frac hits -- communication between a new child well's hydraulic fractures and existing parent wells -- are a major operational and economic issue in developed unconventional plays. Managing the parent-child relationship requires understanding the depletion state of the parent well, the stress field between the wells, and the fracture geometry of both wells.

AI models trained on historical frac hit data (pressure responses in parent wells, production impacts, fiber optic data from monitored treatments) can predict frac hit severity for planned wells and recommend mitigation strategies (pre-loading parent wells, adjusting child well design, modifying spacing). This is an area where the data is accumulating rapidly and the economic impact is significant.

Opportunity 4: Automated DFIT Interpretation

DFIT interpretation -- determining closure stress, pore pressure, and leak-off parameters from a diagnostic fracture injection test -- is a specialized skill that requires understanding of pressure transient analysis, fracture mechanics, and poroelasticity. Different interpretation methods (G-function, square root of time, compliance method) can yield different results, and the "correct" interpretation often depends on the analyst's experience and judgment.

Machine learning models trained on large databases of DFIT data with expert interpretations could automate routine DFIT analysis, providing consistent interpretations that reduce the dependence on specialist expertise. This is a narrow but high-value application where the training data exists and the problem is well-defined.

What the Frac Companies Actually Do With Data

It is worth noting the distinction between frac design software (used by operators to plan completions) and the data analytics capabilities of the pressure pumping companies themselves.

Liberty Energy, ProPetro, Halliburton, SLB, and other frac companies collect detailed operational data from every treatment they pump: treating pressures, rates, proppant concentrations, equipment performance, fluid properties, and operational efficiency metrics. This data is proprietary and represents a significant competitive asset.

The larger frac companies use this data for several purposes:

  • Operational efficiency optimization -- Reducing non-productive time, optimizing equipment maintenance schedules, improving pump-up and shut-down procedures
  • Equipment reliability -- Predicting component failures (pump fluid ends, valves, iron) to reduce unplanned downtime
  • Treatment design recommendations -- Using their treatment database to recommend designs that work well in specific geological settings
  • Pricing and bidding -- Understanding their cost structure at a granular level to bid competitively

For operators, the key question is whether they can access or benefit from the frac company's data and analytics. Some operators negotiate data-sharing provisions in their pumping contracts. Others rely on the frac company's recommendations. The most sophisticated operators build their own analytics capability and use it to evaluate and challenge the frac company's recommendations.

Practical Recommendations

For completions engineering teams:

Do not trust any single model. Frac design simulators are tools for understanding, not oracles. Run multiple models (or multiple scenarios within a single model) and pay attention to the range of outcomes, not just the base case. If different models give you very different answers, the uncertainty is large and the design should be robust to that uncertainty.

Invest in monitoring. You cannot optimize what you cannot measure. Fiber optic monitoring (DAS/DTS) has become cost-effective enough to deploy on a routine basis, at least on a sample of wells. The insights from monitoring data -- actual fluid distribution, cluster efficiency, fracture height growth, inter-well communication -- are essential for calibrating models and improving future designs.

Be cautious with pure data-driven completions optimization. Geological confounding is a real and severe problem. Any AI system recommending completions design changes should be able to demonstrate that it has adequately controlled for geological variation. Ask how the model separates the geological signal from the completions signal.

Build your own completions database. Public data is useful for broad trends but insufficient for well-level optimization. Your own treatment data, combined with your geological models and production data, is the most valuable dataset you have for completions optimization. Structure it carefully -- consistent field naming, comprehensive metadata, linked geological and completions parameters.

At Groundwork Analytics, our approach to completions analytics emphasizes the integration of physics-based models with data-driven methods. Pure data analytics risks confusing geological quality with completions effectiveness. Pure physics models lack the empirical calibration needed for reliable prediction. The most effective approach combines both, and requires deep understanding of subsurface physics alongside computational methods.

Dr. Mehrdad Shirangi is the founder of Groundwork Analytics and holds a PhD from Stanford University in Energy Systems Optimization. He has been building AI solutions for the energy industry since 2018. Connect on X/Twitter and LinkedIn, or reach out at info@petropt.com.

Related Articles

Have questions about this topic? Get in touch.

Back to Insights