Sie sind auf Seite 1von 13

SPE 166468

Cross-Play Shale Gas Well Performance Analysis

Narayan Nair, Linn Energy; Mark A. Miller, Promethean Technologies Group

Copyright 2013, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 30 September2 October 2013.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by t he author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract mus t contain conspicuous acknowledgment of SPE copyright.

Shale resource plays often present formidable reservoir management challenges, particularly with regard to capital utilization
and allocation. In spite of significant efforts to measure and analyze key reservoir and completion data, uncertainty typically
remains in the physical characteristics of the stimulated reservoir volume (SRV) accessed by hydraulic fracturing, namely:
shale permeability, fracture spacing, SRV spatial dimensions, and gas-in-place.
In this study, well performance histories of several hundred wells spanning the Haynesville, Woodford, Barnett, Horn
River, Marcellus, Fayetteville and Montney shale plays were investigated with a common and consistent analytical framework
that determined: a) a well productivity measure during infinite-acting linear flow, b) completion pressure losses (between
sandface and bottomhole), and c) apparent original-gas-in-place in the SRV. Parameters determined from the analyses are key
indicators of the combined result of reservoir quality and hydraulic fracture performance. Results of this multi-well cross-play
study provide information about both inter- and intra-play variability and commonality.
A large proportion of the wells (85%) showed prolonged periods of linear transient flow, indicative of low matrix
permeabilities. Wells of higher productivities tended to go into depletion flow earlier, which could be consistent with either
high matrix permeability or large fracture surface areas that are close together (complexity). Completion pressure drops were
not observed as the dominant productivity-loss mechanism. Productivity of wells during infinite-acting flow normalized by the
total mass of proppant used in each well, an SRV creation efficiency, and the original-gas-in-place in the SRV for all wells in
this study showed log-normal distributions. The 100-day flow efficiency of the completions in these wells showed a truncated
normal distribution.
Recognizing the variability in the reservoir characteristics of the studied shale plays, and SRVs that were outcomes
of diverse completion practices, the observed predictability in key well performance characteristics is remarkable. The
distributions presented in this paper can be used to improve well performance predictability, help identify sweet spots for
drilling, establish relatively narrow ranges of a priori well performance characteristics (therefore reducing uncertainty at
undeveloped well locations), improve fracturing practices, and provide support for decisions of capital allocation.
A major objective of this paper was to improve well performance predictive capabilities in order to improve capital utilization.
The technical challenges in achieving this objective require diagnostic evaluations of reservoir, completion, and petrophysical
variables. Indirect inference of feasible ranges of these important variables from models of well production remain an
invaluable tool.
There has been much research into different methods for forecasting shale gas well production performance. They can
be broadly categorized into three categories: empirical methods (e.g., decline curves), analytical methods (e.g., rate transient
analysis), and reservoir simulation.
Reservoir simulation provides the most rigorous treatment of reservoir dynamics. Studies such as those conducted by
Ehlig-Economides et al. (2012) and Jayakumar and Rai (2012) can provide the most robust treatment of the combined reservoir
and completion complexities associated with production from low-permeability multi-fractured horizontal wells. And though
simulation can provide the most complete analysis of well performance, simulation studies are time-consuming and often
require special training to perform.
At the other end of the usability spectrum are empirical decline curve methods (e.g., Ambrose et al., 2011, Joshi
and Lee, 2013). Though decline curve analysis provides rapid and often useful results, applying decline curves to shale wells
has a number of serious drawbacks. First, because decline curves are empirically based, they rely on having a large body of

SPE 166468

historical data to provide validity to the results. But perhaps most importantly, standard decline curve analysis techniques rely
on wells being in quasisteady-state depletion, assuming constant bottomhole pressure and constant operating conditions. For
shale wells, especially those relatively early in their producing lives, decline curve forecasting has considerable uncertainty,
both because of the lack of long-term historical precedents, but also because the decline behavior is significantly different from
conventional wells in quasisteady-state flow.
Analytical and semi-analytical methods (e.g., Bello and Wattenbarger, 2010, Miller et al., 2010) provide the means to
more rapidly analyze and forecast well performance while retaining grounding in reservoir physics not possible with decline
curve analysis. Traditional rate-transient methods rely on solutions to the diffusivity equation, approximately accounting for
non-linearities in the solution (using pseudopressure and pseudotime) as well as changes in operating conditions. Similarly to
reservoir simulation, analytical methods can provide insight into the possible ranges of key reservoir and completion data via
calibration (i.e., history matching) of the models against actual well performance.
The nature of resource plays suggest that statistical methods should be employed to do forecasts on a play-wide basis
(Hall et al., 2011). This paper uses a semi-analytical methodology to develop statistical intra- and inter-play descriptions of
key reservoir and completion parameters across eight North American shale plays.
The well dataset consisted of daily production and flowing pressure data from 650 multi-fractured horizontal wells in dry gas
shale plays across North America. Half the wells are from the Marcellus formation in the Northeast United States. There are
100 wells from the Barnett, Fayetteville, and Horn River plays. The Haynesville and Woodford shales contributed the
remaining 225 wells in this study. Note that the wells were not picked randomly. Rather, the selected wells were those with
sufficiently good data quality to justify high-level analysis. More importantly no wells were excluded to avoid bias or to
smooth the results. There are almost no exceptions to slickwater fluid designs in the hydraulic fracturing in the wells in this
dataset; however there was significant diversity in placement and diversion techniques (Fig. 1, Table 1).
The theoretical framework used in this paper is described in Miller et al. (2010). Additional real world applications
were presented by Sahai et al. (2012), Jayakumar and Rai (2012) and Boulis et al. (2012, 2013). The volumetric assumptions
in the drainage area of each horizontal well were determined using rigorous characterization (Shebl et al., 2013; Ambrose et
al., 2010). Pressure-dependent reservoir permeability was not used in these wells for this comparison study. The development
of the analytical workflow relied heavily on comparing and learning from reservoir simulation test suites spanning the range of
expected reservoir and completion parameters in the different plays. Under similar assumptions, the learnings using the
analytical workflow is fairly consistent with reservoir simulation. The challenge is dealing with non-uniqueness in reservoir
permeability and dimensions of SRV, and especially the number of clusters that initiated producing fracs. In this paper, the use
of unique lumped parameters, recognizing non-unique reservoir and completion parameters, is consistent with Miller et al.
Depending on the flow regimes seen in a well, only certain lumped parameters can be uniquely determined (Table 2).
As described in Okouma et al. (2012), individual well analysis involved two basic steps: 1) identifying the flow periods
observed in well performance and 2) obtaining initial estimates of J lt - a well productivity measure during transient linear
Rcomp - completion pressure losses (between sandface and bottomhole), and STFtransition - the superposition time
function at the onset of intra-fracture interference (internal SRV depletion). Examining Eq. 2 in Miller et al. (2010) it is clear
J lt Af k
During infinite-acting linear flow no assumption of fracture geometry is required to characterize J lt and Rcomp . The
SRV linear flow parameters J lt and

Rcomp provide a means of comparing well performance. For wells in the same play,

looking at Eq. 1, the reasons for a higher J lt while comparing wells can be due to a higher Af caused by better stimulation
practices or higher permeability. Higher Af can be due to better fraccability or perhaps the presence of natural fractures that
are opened as a result of hydraulic fracturing. The metric can be used to compare wells from different plays as differences
caused by reservoir pressure fluid and rock properties is also accounted in the formulation. Rcomp is a way of evaluating
fracture-to-wellbore connectivity effectiveness.
For consistent comparisons, a flow efficiency of wells at 100 days can be calculated as
PI t

PI t Rcomp
J lt 1
In wells that have seen the transition from infinite-acting flow to SRV depletion, a bi-wing linear frac assumption
was used to calculate the original-gas-in-place in the SRV:
GSR J lt STFtransition

SPE 166468

In wells that are still in infinite-acting flow, a minimum gas-in-place in the SRV GSR ,min can be calculated. In order
to qualify the onset of depletion with the SRV parameters the following relationship should be noted for bi-wing linear fracs.
STFtransition s
The transition happens when the pressure response between the fracs begin to interfere at a boundary halfway
between fractures. Note that for the same fracture spacing, higher permeability causes earlier transition. Also, for wells with
identical permeabilities, smaller fracture spacings (e.g., due to frac complexity or cluster design) will result in earlier
In our experience the amount of proppant placed in a well generally correlates to higher well performance (Fig. 2).
Similarly to Okouma et al. (2012), a Normalized Linear Transient Productivity, J lt M p was thus used for comparison
The initial production rates in a well are mostly sensitive to J lt and the initial pressure of the reservoir; while EUR is
correlated with the size of the SRV. In a much broader sense, for a given lateral length there are two basic tools to enhance
value: acceleration and extension. Acceleration can be achieved by increasing Af . Extension is achieved by increasing the
effective SRV dimension available for drainage. In order to isolate the efficiency of gain in extension and to compare
fraccability, an SRV creation efficiency is defined as
ESRV c SR ,min
Ll M p
The volume inside the SRV is readily calculated as
VSR,min Bgi GSR,min


ESRV is defined as the ratio of two parameters: 1) VSR,min Ll which is indicative how much SRV-extension (increase in
frac half-length and/or frac height with units of ft2) in SRV dimensions in achieved, and 2) M p nc which represents the
amount of proppant allocated per cluster to achieve extension.
A large proportion of wells (85%) in this dataset were interpreted to still be in infinite-acting linear flow. Such prolonged
durations of infinite-acting flow could be indicative of low permeability rock (Eq. 4). This will obviously change with longer
well histories. Thus the estimated GSR was only calculated in wells that had seen transition. The results from all the wells are
in this study are shown using probability plots in Fig. 3, Fig. 4, Fig. 5, and Fig. 6. The following observations can me made:

J lt and GSR appear to be log-normally distributed.

The P10/P90 ratio in J lt and GSR across all shale plays shows very little variance; they are less than 10.

The flow efficiency of wells shows a truncated normal distribution. The P50 value indicates that after 100 days of
production, reservoir pressure drops dominate the flow behavior. In the median well, completion pressure losses
contribute only 16% of pressure drops. Wells above 0.8 standard deviations from the mean showed no completion
pressure drops. This could mean either most created fracs are of infinite acting conductivity, or only fracs with
sufficient proppant coverage, and thereby of high conductivity, remain effectively connected to the wellbore.
The wells of comparatively higher J lt are associated with sweet spots in a play. These wells tended to see an early
transition from linear to depletion flow in the SRV. However, it may not be possible to distinguish between complex
fracs (smaller Ls ) and higher permeability as reasons for the early transition seen in high- J lt wells (Fig. 6).

The distributions of J lt M p and ESRV in all wells are shown in Fig. 7 and Fig 8, respectively. Cross-play comparison
probability plots of J lt M p , ESRV , and 100-day flow efficiency are shown in Fig. 9, Fig. 10, and Fig. 11, respectively. Crossplay comparisons of GSR are not presented due to lack of a sufficient number of wells that had observed transitions. However,
the estimated GSR,min in the wells still in linear flow is shown in Fig. 12. The key observations are:

Both J lt M p and ESRV show log-normal distributions. The P10/P90 ratio (~10) seen in ESRV and J lt M p suggests
that the expected range of overall rock quality has a very narrow range of outcomes. This is a remarkable observation
considering the variability in the various shale plays, and the completion practices used in the wells.
Wells from the Woodford Shale showed the highest J lt M p , followed by the Fayetteville, and are outliers. It is likely
the ranges of productivity seen in these plays may look similar, but due to different reasons. Inspecting the ranges of
GSR ,min and ESRV it is clear that the resulting dimensions of the SRV are much larger in the Woodford compared to
the Fayetteville. One might infer that the fraccability of the Woodford Shale is higher, leading to better Af , while

SPE 166468

permeability in the Fayetteville is higher. ESRV in the Woodford is an outlier compared to the other plays, confirming
the fraccability hypothesis. However, it is interesting to note that none of the Woodford Shale wells in this dataset
showed a transition, in spite of long production histories.
The distribution of wells in the Marcellus was much smoother showing that the sampling was less biased in the larger
dataset. However, there is a definite break delineating the higher productivity wells in the Montney and Haynesville
The largest SRV dimensions were noted in the Horn River Shale, while J lt M p and ESRV in this play are relatively
lower compared to the others. This is probably due to the single perforation cluster per stage design in long laterals,
combined with relative high proppant placement per cluster, resulting in SRV-extension (also see Fig. 5 in Okouma et
al., 2012). This affirms the need for normalizing results by completion size.
100-day flow efficiencies in the Haynesville Shale are outliers compared to the other plays, while most other plays
were grouped together. The median well showed between 10 and 30% pressure drops in the completion.

The reasons for the narrow ranges and predictable distributions seen for the variables in this paper are very thought provoking.
There are several questions that arise from the results:
Why do J lt M p , ESRV , and 100-day flow efficiency show predictable distributions? One could speculate J lt M p is

the product of two variables permeability and fracability to justify the log-normal distribution. A fracture stimulation
expert could guess the independent distributions that result in GSR and ESRV (Fig. 9 and Fig. 10). Also, completion
pressure losses are the sum of independent variables, such as fracture conductivity, convergence pressure drops (from
the fracs into the wellbore), etc. to explain normal distribution observed in the data (Fig. 11).
As an example, in the Woodford, could the contribution of natural fractures play an important role in enhancing
fraccability? And, why do we not see an earlier transition due to lower values of Ls (Eq. 4) associated with frac
complexity? If natural fracs are the primary driver for ESRV , it does serve as caution to separate the existence of
natural fracs from frac complexity. In our experience, interpreting natural fracs and frac complexity by modeling
production data are independent assumptions.
How can we improve J lt M p , ESRV , and 100-day flow efficiency? Another way to phrase the question would be: is
the variance and range seen an artifact of nature? Will improvements in technology change these entirely, or just
narrow the ranges? Of these three parameters, it seems flow efficiency is a parameter over which operators can have a
fair degree of control.
Why is the 100-day flow efficiency in the Haynesville Shale an outlier? Perhaps a combination of much higher initial
reservoir pressures in very thermally mature and dehydrated shale, lead to reduced frac fluid flowback, and thereby
reduced pressure drops in the fracture system. It is also possible that the overpressured nature and high closure
pressures that result in only fracture asperities that have high proppant coverage can remain effectively connected to
the wellbore.
Since pressure drops in the completion are not generally significant, is fracture conductivity and proppant quality a
non-issue? Though a majority of wells in this study showed little completion pressure losses, proppant quality may
still need consideration. In our experience, it is preferable to compare J lt M p and ESRV in addition to the inferred
effective fracture conductivity in wells completed with different proppant types.

Recent work suggests that the presented framework is very applicable to liquid-rich systems as well as dry gases. We
encourage the industry to use the results shown in this paper for qualitative comparisons against their own wells. With an
existing dataset of wells, some potential application for the observations in this paper could be:
Improving predictive ability Well performance predictability can be a challenge in an asset with geological
heterogeneity and constantly evolving completion practices. Without extensive data it can be unclear if the resulting
well production performance is driven by permeability and/or completion effectiveness. J lt M p , ESRV , and 100-day

flow efficiency, determined by analysis of existing well production data could be compared to Fig. 9, Fig. 10, and
Fig. 11 to improve understanding of the primary drivers of well performance.
Well selection Understanding variability in reservoir permeability and fraccability are crucial for high-grading
drilling locations. Spatial maps of J lt M p in existing wells can be a valuable aid for mapping reservoir variability,
and locating sweet-spots for prioritized development.
Prospect evaluation In the absence of sufficient control points, very wide uncertainties can result in well
performance at an undeveloped location. If non-deterministic simulations consider the individual uncertainty in
reservoir permeability and fraccability, the resulting range in EUR can show too wide of a variance. The results

SPE 166468

justify that the variance in J lt M p and ESRV tend to be lower, hence the individual uncertainty in reservoir

permeability and fraccability can be constrained within some realistic limits. Our initial experience has showed
significant reduction in the predicted EUR range using this approach. In the absence of control data, the ranges
presented in this paper could help support engineering judgment.
Design fracture placement Not all Af is created equal! Stimulation energy is a finite resource that requires efficient
allocation towards the optimum combination of acceleration using multiple fracs vs. extension with larger SRVs.
While the crux of this paper focused on J lt , a key initial performance indicator and required building block for well
performance models, EUR and thereby optimizing PV10 and ROR in a well needs to consider operational variables
such as but not limited to: the number of stages, the number of clusters per stage, amount of proppant designed per
stage, and well spacing. The intended geometry of Af may thus need to be adjusted to the anticipated reservoir

permeability. As an example in rock of higher reservoir permeability, one could reduce the number of stages and
clusters, and direct frac energy to create larger SRV dimensions while leveraging permeability for drainage within the
Portfolio analysis Elaborating the prospect evaluation utility, existing well data can be used to predictively model,
and aggregate asset-level well performance in order to manage capital allocation among varying assets, design
surface facilities, and schedule drilling and completion activities.

The results in this paper were compiled after considerable research activity into understanding shale gas well performance.
This paper shows that relatively straightforward and simple analysis can lead to very meaningful results. Well/formation
productivity (e.g., J lt ) , fracability indicator ( ESRV ), and completion skin (e.g., Rcomp ) are demonstrably rational measures
of well productivity. In a large dataset of wells, the normalized productivity index during linear flow, original-gas-in-place in
the SRV, and an SRV creation efficiency, tend to show log-normal distributions. The 100-day flow efficiency of the
completion showed a truncated normal distribution. These distributions suggest it is possible to establish relatively narrow
ranges of a priori well performance characteristics, and therefore reduce uncertainty at undeveloped well locations, identify
sweet-spots for drilling, improve fraccing practices, and provide support for decisions of capital allocation. Finally, an
unbiased basis for qualitative comparisons of reservoir quality permeability and fraccability is established.
Ambrose, R.J., R.C. Hartman,, M.D. Campos, I.Y. Akkutlu, and C. Sondergeld: New Pore-Scale Considerations for Shale Gas In-Place
Calculations, paper SPE 131772 presented at the 2010 Society of Petroleum Engineers Unconventional Gas Conference, Pittsburgh,
USA, 23-25 Feb.
Ambrose, R.J., C.R. Clarkson, J. Younblood, R. Adams, P. Nguyen, M. Nobakht, and B. Biseda: Life-Cycle Decline Curve Estimation for
Tight/Shale Gas Resevoirs, paper SPE 140519 presented at the 2011 SPE Hydraulic Fracturing Technology Conference and
Exhibition, The Woodlands, TX, 24-26 Jan.
Bello, R.O., and Wattenbarger, R.A. 2010. Multi-stage Hydraulically Fractured Shale Gas Rate Transient Analysis. Paper SPE 126754
presented at the SPE North Africa Technical Conference and Exhibition, Cairo, Egypt, 14-17 February.
Boulis, A., R. Jayakumar, C. Nyaaba, R. Rai, and V. Sahai: Challenges Evaluating Shale Gas Well Performance: How Do We Account For
What We Dont Know? Paper 16396 presented at the 2013 6th International Petroleum Technology Conference, Beijing, China, 26-28
Boulis, A., R. Jayakumar, and A. Araque-Martinez: An Innovative Approach to Understanding Shale Gas Well Behavior Based on a
Performance Catalogue Fingerprint. Paper 160271 presented at the 2012 Society of Petroleum Engineers Asia Pacific Oil and Gas
Conference and Exhibition, Beijing, China, 22-24 Oct.
El-Banbi, A.H. and Wattenbarger, R.A. 1998. Analysis of Linear Flow in Gas Flow Production. Paper SPE 39972 presented at the SPE Gas
Technology Symposium, Calgary, AB, Canada, 15-18 March.
Ehlig-Economides, C.A., I. Ahmed, S. Apiwathanasorn, J. Lightner, B. Song, F. Vera, H. Xue, and Y. Zhang: Stimulated Shale Volume
Characterization: Multiwell Case Study from the Horn River Shale: II. Flow Perspective, paper SPE 159546 presented at the 2012
SPE Annual Technical Conference and Exhibition, San Antonio, TX, 8-10 Oct.
Jayakumar, R., and R. Rai: Impact of Uncertainty in Estimation of Shale Gas Reservoir and Completion Properties on EUR Forecast and
Optimal Development Planning: A Marcellus Case Study. Paper 162821 presented at the 2012 Society of Petroleum Engineers
Hydrocarbon Economics and Evaluation Symposium, Calgary, Canada, 24-25 Sept.
Hall, R., R. Bertman, G. Gonzenbach, J. Gouveia, B. Hale, P. Lupardus, P. McDonald, N. Meehan, B. Vail, and M. Watson: Guidelines for
the Evaluation of Resource Plays, 2011, Houston, Texas: Monograph Series 3, SPEE.
Jayakumar, R., R. Rai, and V. Okouma: Impact of Uncertainty in Estimation of Shale Gas Reservoir and Completion Properties on EUR
Forecast and Optimal Development Planning: A Marcellus Case Study, paper 154933 presented at the 2012 Americas
Unconventional Resources Conference, Pittsburgy, PA, 5-7 June.
Joshi, K. and J. Lee: Comparison of Various Deterministic Forecasting Techniques in Shale Reservoirs, paper SPE 163870 presented at
the 2013 SPE Hydraulic Fracturing Technology Conference, The Woodlands, TX, 4-6 Feb.
Miller, M.A., C. Jenkins, and R. Rai: Applying Innovative Production Modeling Techniques to Quantify Fracture Characteristics, Reservoir
Properties, and Well Performance in Shale Gas Reservoirs. Paper 139097 presented at the 2010 Society of Petroleum Engineers
Eastern Regional Meeting, Morgantown, USA, 12-14 Oct.

SPE 166468

Okouma, V., S. Vitthal, N. Nair, and M. A. Miller: Playwide Well Performance Analysis in Montney Siltstone. Paper 162843 presented at
the 2012 Society of Petroleum Engineers Canadian Unconventional Resources Conference, Calgary, Canada. 30 Oct-1 Nov.
Sahai, V., J. Jackson, and R. Rai: Optimal Well Spacing Configurations for Unconventional Gas Reservoirs. Paper 155751 presented at the
2012 Society of Petroleum Engineers Americas Unconventional Resources Conference, Pittsburgh, USA, 5-7 June.


hydraulic fracture area open to flow, ft2


initial formation volume factor of gas, ft3/Mscf


original-gas-in-place contacted inside the SRV, Bscf

J lt

SRV- linear transient productivity index, Mscf/psi D


reservoir permeability, md
perforated lateral length, ft


hydraulic fracture spacing, ft

total mass of proppant, MMlbm


number of perforation clusters

volume of SRV, ft3


STFtransition =

transition linear flow superposition time function,










Number of frac stages



Number of perforation clusters per stage

Proppant per lateral length, lbm/ft




Proppant per cluster, Mlbm




Perforated lateral length, ft

Total mass of proppant placed, MMlbm


Flow Regime Types

Known Parameters

SRV Properties

SRV Dimensions

Well is in Linear (Transient) Flow




Well performance in Linear Flow and

Transitions to Intra-Frac Interference




Range of

Range of Jlt


Flow regime seen predominantly in Depletion

SPE 166468


Cumulative Frequency

Cumulative Frequency




Number of Frac Stages


Cumulative Frequency

Cumulative Frequency

Days on Production (days)



Number of Clusters per Stage



500 1,000 1,500 2,000 2,500 3,000
Proppant Mass Per Lateral Length (lbm/ft)

Total Proppant Mass (MMlbm)



Cumulative Frequency

Cumulative Frequency

Perforated Lateral Length (ft)

Cumulative Frequency

Cumulative Frequency


Proppant Mass Per Cluster (Mlbm)


Cumulative Gas (Bscf)


Figure 1: Variability in the completion characteristics of the wells in this paper are shown. A median well has 4000 ft of lateral length,
10 frac stages, 4 clusters per stage, and 1100 lbm/ft of proppant per ft. of lateral. Most of the wells studied in this paper had several
months of daily production data, and the median well had already produced 1 Bscf of gas when this study was conducted.

SPE 166468


Gas Rate (Mscf/D)

Gas Rate (Mscf/D)


Cumulative Gas Produced (Bscf)

Cumulative Gas Produced (Bscf)



Total Proppant Mass (Mp), MMlbm

Figure 2: (a) The two offset wells shown in this plot were completed with different amount of proppant placed per cluster with all
other completion variables held constant. Well A had 180 Mlbm placed per cluster, while only 100 Mlbm per cluster in well B. The
analysis shows

J lt 65 in well A and J lt 38 in well B. The 100-day flow efficiencies were 65% and 45% respectively. In an ideal

scenario the larger volumes placed in well A is the reason for the higher productivity, due to a proportional increase in

Af . (b) In

this offset well pair, well A had a 2100 ft lateral stimulated using 4 stages, 8 clusters per stage, and 220 Mlbm per cluster. While, well
B had 3900 ft of lateral stimulated using 12 stages, 8 clusters per stage, and 120 Mlbm per cluster. Clearly well B is a much better
performer with the analysis showing J lt 48 , compared to J lt 13 in well A. (c) Similar to the well examples in Fig. 2a and 2b, wells
completed in similar rock quality, show proportional increases in

J lt and thereby Af to the total mass of proppant used in the

completion M p . Each data point is a horizontal well subject to a systematic completion variability study conducted in the Marcellus
Shale. The wells are colored by region showing the impact of rock quality on the slope of

J lt vs. M p .

SPE 166468

Figure 3: Probability plot showing distribution of J lt in all wells in this study. The P10/P90 ratio is approximately 7, suggesting a
narrow range of outcomes considering the different shale plays and completion practices.

Figure 4: Probability plot of 100-day flow efficiency for all wells in this study. The median perentage of pressure losses in the
completion after 100 days of production is 84%, while several wells showed pressure drops only attributed to the reservoir.


SPE 166468

Figure 5: 15% of wells in this study showed transition from linear flow to SRV-depletion. Probability plot of

GSR for this subset of

wells is shown.

Figure 6: The average J lt of wells that showed transition in the Haynesville and Marcellus play are compared against other wells in the
sample play that remain in infinite-acting linear flow. The data suggests wells that have seen transition have a higher
wells still in linear flow.

J lt compared to

SPE 166468


Figure 7: Probability plot showing distribution of J lt


in all wells in this study. Similar to the distribution of J lt in Fig. 3, the plot

suggests a relatively narrow range of outcomes, and remarkably all plays exhibit the same variance.

Figure 8: Probability plot of

ESRV for all wells in this showed the same distribution and a P10/P90 ratio of 12.


SPE 166468

Figure 9: Crossplay comparison of

J lt M p

is shown. This plot can be used as a reference for qualitiative comparisons of rock

quality combined effect of reservoir permeability and fraccability.

Figure 10: Crossplay comparison of

ESRV is shown. This plot can be used as qualitatively to compare the contribution of fraccability
to well performance.

SPE 166468


Figure 11: Crossplay comparison of 100-day flow efficiency is shown.

Figure 12: Crossplay comparison of

GSR,min is shown. Note that the range is very analagous to the range of predicted EURs in shale