From post-mortem to real-time: A scalable approach to fracture diagnostics

Tyler Szilagyi, ShearFRAC 

OPTIMIZING COMPLETIONS THROUGH REAL-TIME FEEDBACK 

Understanding fracture behavior during active well completions is essential to improving the efficiency and effectiveness of hydraulic fracturing operations. Historically, diagnostic technologies, such as microseismic, fiber optics, and chemical tracers, have provided invaluable insights into subsurface responses. Yet, their utility is often hampered by a common limitation: they are typically analyzed, once operations have concluded, limiting their ability to influence active operations. This "post-mortem" approach forces completion design adjustments to be deferred to subsequent wells or pads, rather than the active wellbore itself.  

The dynamic nature of fracture propagation from stage to stage highlights a need within industry for real-time measurement solutions capable of providing immediate feedback during operations, enabling optimization within a single treatment. Operators can address this challenge by implementing a surface-based fracture measurement technology that analyzes high-resolution pressure fluctuations captured at the wellhead during stimulation.  

Through derivative pressure analysis, the method quantifies the intensity and timing of discrete fracture propagation events, offering stage-specific insight into subsurface responsiveness in real time. This article demonstrates a case study on how this workflow enabled a transition from overstimulation to data-driven optimization, resulting in measurable improvements in completion efficiency and overall well economics without sacrificing productivity.  

THE SHIFT TOWARD SCALABLE DIAGNOSTICS 

Unconventional reservoir development relies heavily on effectively stimulating low-permeability rock to create a productive fracture network. This process involves injecting high-pressure fluids and proppants into hydrocarbon-bearing rock to create fractures that serve as pathways for resource extraction. However, even when applying identical fluid volumes, rates, and proppant loadings to adjacent stages in a single lateral wellbore, resultant fracture geometry, cluster efficiency, and stimulated rock volume (SRV) can vary significantly. 

Findings by Grasselli et al. (2023) highlight that in highly laminated formations, fracture propagation is strongly influenced by rock fabric. Furthermore, subtle variations in in-situ stress layering and the evolving effects of stress shadowing are primary drivers of stage-to-stage variability.  

While traditional tools have improved the industry's understanding of stimulation behavior, their cost and invasive deployment often restrict widespread use. Recent work by Cipolla et al. (2024) reinforces the value of real-time fracture optimization and highlights the need for scalable, cost-effective diagnostics to guide stage-level decisions.  

Surface pressure analysis addresses this gap directly. Because the measurement is taken at the wellhead using existing pumping infrastructure, it can be deployed on every stage of every well without the incremental cost or operational footprint of downhole instrumentation. Just as importantly, the signal is available during the treatment itself, rather than after the fact, so the same diagnostic that makes stage-level coverage economical also makes in-stage decisions possible. That combination of scalability and real-time visibility is what enables stage-level diagnostics to function as a standard part of completion execution, rather than a one-off science exercise, and it is the foundation for the case study that follows. 

APPLIED SURFACE PRESSURE ANALYSIS  

The technical foundation of this approach lies in the continuous interaction between surface pumps, the wellbore, and the growing subsurface fracture network. During hydraulic fracturing, subtle fluctuations in the wellhead pressure signal correspond to discrete events, such as fracture initiation and propagation. By capturing these variations and analyzing them in the derivative domain, a real-time interpretation of fracture dynamics can be obtained. This approach yields stage-specific insights by treating pressure as a dynamic signal, rather than a static value.  

As a conceptual simplification, we can imagine the wellbore and surrounding low-permeability formation as a "vessel." As fluid is pumped into a vessel with a fixed volume, pressure rises exponentially. It is only when additional area is introduced, through expansion or failure of the vessel wall, that pressure will drop. The magnitude of that pressure drop can be used to infer how much additional volume or surface area has been created, while the rate of occurrence provides information on the frequency of these expansion or failure events. Conversely, if no meaningful breaks occur, it can be inferred that fluid has found previously created pathways, natural fractures, and/or faults within the formation.  

From this analysis, two primary quantitative measurements are derived to characterize the formation's mechanical response: 

• Frequency: The rate of occurrence of pressure-derived stress-transfer events over time. It reflects how actively the rock is responding to energy input, serving as a direct indicator for the rate and relative quantity of fracture propagation events.  
• Effectiveness: This quantifies the cumulative magnitude of these events. Effectiveness can be normalized by input parameters to provide a direct measure of how efficiently energy is converted into new fracture surface area.  

Digital filtering is applied to 1-Hz sampling data, suppressing high-frequency artifacts from surface equipment, pipe friction, and turbulent flow. This noise-filtered signal provides a high-confidence measurement of fracture system evolution in real time, presenting a significant advancement over conventional static pressure interpretation.  

FIELD APPLICATION: CENTRAL BASIN PLATFORM CASE STUDY (SPE-230631) 

Surface pressure analysis has now been deployed on more than 1,000 wells across a range of basins, operators, and completion designs, providing a broad base of evidence for the approach in routine field operations. Within that body of work, the Central Basin Platform program detailed below offers a particularly instructive case. That’s because the same wells were instrumented with multiple independent diagnostics, allowing the pressure-based measurements to be cross-validated and the resulting design changes to be tracked from science pad through full-field development.

Fig. 1. Total and normalized pressure derived effectiveness by stage design.

The study targets the ~1,000-ft-thick waterflooded carbonate interval of the Clear Fork Formation, characterized by variable depletion and distinct mechanical layering. Prior horizontal wells suggested that large, high-intensity completions did not necessarily yield improved production; instead, over-fracturing risked connecting with high water-oil ratio (WOR) intervals previously depleted through decades of waterflooding. Conversely, containing the stimulation within a single sub-interval of the Clear Fork could preserve other intervals for future development. 

Sabinal Energy executed a multi-year program to validate this technology, beginning with Pad A, which served as a dedicated science pad targeting the Middle Clear Fork. The two-well pad was fully instrumented with high-fidelity surface pressure diagnostics, microseismic monitoring, tracer analysis for fluid placement verification, and geological log integration.  

The program tested seven design combinations across the laterals, incorporating three stage spacings and five proppant loadings. Surface pressure analysis provided a continuous record of fracture activity, identifying that while designs with higher proppant and fluid loading exhibited the highest total effectiveness, they did not necessarily translate to improved efficiency. 

Fig. 2. Independent fracture counts from microseismic analysis and pressure signal analysis.

When effectiveness was normalized by clean fluid volume, smaller treatment designs demonstrated comparable and even superior efficiency, relative to higher-load stages, achieving similar stimulation intensity with significantly lower injected volumes. For instance, Design 5 provided the most favorable ratio of fracture effectiveness to clean volume injected, Fig. 1.  

The correlation with independent diagnostics reinforced the validity of the pressure signal. Microseismic data, which capture acoustic emissions from subsurface fractures created during stimulation, showed a strong correlation with the pressure-derived metrics. The total event count for each stage across the entire lateral from these two independent analyses is shown in Fig. 2.  

Plotting the per-stage event count from each method against the other (Fig. 3) yielded a coefficient of determination (R²) of 0.56 across all stages, improving to 0.78, once stages with reduced microseismic resolution near the toe of the laterals were excluded. That level of agreement is notable, because the two diagnostics measure fundamentally different physical phenomena. This correlation provides high confidence that the pressure signal is responding to real subsurface behavior, rather than surface or pumping artifacts. 

Furthermore, the normalized shear-stress magnitude derived from pressure analysis corresponded closely with the microseismic-derived Stress Index (SI). Stages with intense pressure activity aligned with those exhibiting higher SI values, both reflecting zones of elevated strain-energy release. 

Fig. 3. Cross-plot of per-stage event counts from microseismic and pressure signal analysis (R² = 0.78).

Tracer analysis provided a third, independent line of validation. Stages with higher tracer recovery consistently aligned with zones of elevated pressure-based effectiveness, indicating that the pressure signal tracks not only the mechanical fracture response but also the intervals contributing to hydraulic productivity.  

LINKING ROCK MECHANICS TO REAL-TIME FRACTURE ACTIVITY 

Pressure signal analysis revealed a clear relationship between the measured elastic properties of the formation and fracture activity. Intervals characterized by higher dynamic Young's modulus (YMd) exhibited both greater total event counts and higher normalized event frequency per clean barrel of fluid injected. This trend aligns with geomechanics literature linking elevated YMd and lower Poisson's ratio (ν) to higher brittleness and increased potential for complex fracture propagation, as such facies concentrate stress at the fracture tip and favor brittle shear failure over ductile deformation.  

Dipole sonic logs were acquired along both laterals to quantify these dynamic elastic moduli. The logs showed a strong spatial correlation between YMd and the surface pressure-derived fracture measurements, leading to the definition of a preliminary brittleness threshold. Facies exceeding this threshold exhibited a higher density of stress-transfer events, while lower-modulus zones dissipated more strain energy through ductile deformation, resulting in reduced fracture connectivity. 

The practical value of this correlation is that a routine log measurement becomes a forward-looking design input. Once subsurface data have been calibrated against pressure-derived activity, then perforation placement, stage spacing, and proppant loading can be tuned to the rock encountered along the lateral, rather than applied uniformly. As the dataset grows across additional wells and basins, the same workflow extends to pre-screening new development targets before a single stage is pumped.  

PHASE 2 IMPLEMENTATION AND ADAPTIVE EXECUTION 

Building on these insights, the implementation of learnings from Pad A were deployed on Pad B, featuring an additional two laterals. Based on the diminishing returns observed at higher loadings on Pad A, total fluid and proppant volumes were reduced by approximately 30%. Stage spacing was standardized to the intermediate configuration that had balanced hydraulic response with minimal stress-shadow interference on the science pad.  

Because surface-pressure diagnostics provide temporal visibility into stress-transfer behavior during pumping, adjustments to the application of rate and proppant concentration were implemented into the schedule. This continuous feedback facilitated adjustments to maintain an optimized hydraulic response along the lateral, representing a meaningful evolution from static job design toward adaptive completion execution.  

QUANTIFYING RESULTS AND ECONOMICS 

The most compelling result of the optimization workflow was the maintenance of production performance despite the significant reduction in treatment intensity. Normalized to lateral length, cumulative production from Pad B tracks closely with Pad A, confirming that the reduced-volume design did not compromise recovery. Similarities in normalized oil production reinforce that reservoir contact was preserved despite lower treatment intensity.  

The transition from overstimulation to optimization delivered the following quantitative impacts: 

• Cost reduction: The revised completion design reduced costs 30% when compared with Phase 2 wells.  
• Water management: First-year water production decreased 19%, indicating improved fracture containment and reduced communication with legacy waterflood intervals.  
• Production stability: Daily oil production from Pad B stabilized quickly after flowback and has maintained consistent output trends over time. 

THE FUTURE OF REAL-TIME DIAGNOSTICS 

The application of surface-based pressure signal analysis represents a shift from retrospective diagnostics to proactive, real-time optimization. This approach enables stage-level decision-making "on the fly," allowing engineers to benchmark performance and adjust treatment parameters, such as pump rate, slurry concentration, or stage volumes, to enhance stimulation effectiveness while reducing operational cost and risk.  

Fig. 4. Multi-variate machine learning application.

This adaptive capability is currently being advanced through the integration of machine learning and AI-driven analysis. Time-series models trained on historical completions data can recognize pressure signatures that indicate inefficient fracture propagation or suboptimal stimulation behavior. When these patterns are detected, the system can recommend real-time operational adjustments to influence optimal fracture activity. This feedback loop transforms the design from a static plan into a responsive, data-informed process that learns and improves with each stage.  

Historically, each stage contributed a single summarized data point for post-job evaluation, which was sufficient for evaluation but limited for predictive analysis. With pressure signal analysis, engineers are now capable of analyzing second-level measurements. An average 2-hr stage yields 7,200 data points, or over 360,000 data points per 50-stage well, transforming the dataset into a continuous time series suitable for machine-learning applications. Multivariate analysis can be applied to this data to quantify how design parameters influence the evolving pressure response as shown in Fig. 4. 

These correlations enable predictive modeling to test "what-if" scenarios and guide completion design selection before pumping. As datasets expand, model accuracy will improve, allowing pre-job simulation and in-job adaptive control to converge within a single, physics-informed workflow.  

CONCLUSIONS 

The field application study yielded several conclusions, as follows: 

1. Surface pressure analysis provides a real-time framework for evaluating fracture behavior. This approach overcomes the inherent delays and costs associated with traditional post-treatment diagnostics.  
2. Higher effectiveness does not always yield better efficiency. Normalized effectiveness metrics can be used to identify diminishing returns, guiding optimized stage designs in real-time.  
3. Pressure data correlates strongly with independent diagnostics. Validation against microseismic, tracer, and geological data reinforces the physical basis and reliability of pressure signal measurements.  
4. Optimized designs can significantly improve well economics. It’s possible to reduce fluid and proppant volumes strategically while maintaining recovery performance and reducing completion costs. 
   
5. Adaptive completion execution represents the future of the industry. Integrating pressure-based insights with machine learning and artificial intelligence transforms hydraulic fracturing from a reactive process into a continuous learning system that maximizes efficiency.  

TYLER SZILAGYI is senior technical account manager at ShearFRAC, with an extensive background in well completions and hydraulic fracturing. Throughout his career, he has held roles spanning field operations to technical sales, supporting projects in more than 20 countries. At ShearFRAC, Mr. Szilagyi focuses on real-time pressure diagnostic solutions that standardize completion workflows, reduce operational risk, and improve well completion outcomes.