How Arthur Berman Could Be Very Wrong…


Allen Gilmer, Ramona Hovey, and Jason Simmons,Drilling Info, Inc. Energy Strategy Partners

Download the PDF Version of this post

It was with great interest that we read Arthur Berman’s commentary “Lessons from the Barnett Shale suggest caution in other shale plays” in World Oil’s August 2009 edition.  From examining production data, Mr. Berman makes a variety of declaratory statements about the Barnett Shale, and thus Gas Shales in general, which are superficially real, yet disintegrate as predictive factors upon further examination.  Whereas we typically do not respond to the claims or methodologies of E&P analysts and pundits, and reserve our expertise for our clients benefit, Mr. Berman’s conclusions need to be examined and responded to where necessary. The cost of superficial analysis has very real consequences in terms of energy policy today like never before.

The thrust of Mr. Berman’s article is that the Barnett is an uneconomic play in general, that various technologies don’t provide a return on investment, and that the play itself is unpredictable, thus all other shale plays are unpredictable. Mr. Berman’s major claims are that 1) core areas do not have appreciably higher average EUR’s than the play, overall,  2) there is little correlation between IP rates and EUR’s,  3) horizontal wells do not produce substantively more than vertical wells 4) the volumes of commercially recoverable reserves are greatly over-estimated,  5) average well performance has decreased consistently since 2003 for horizontal wells, and  6) hyperbolic predictions overestimate reserves.  Left entirely unaddressed in his analysis is the entire concept of a marginal producer, the statistical effect poor producers have on raw data, and whether it is proper to analyze un-normalized data, since many factors can be quantified  to alter production tremendously, such as corrections  for lateral length, number of frac stages, and the results of simulfracs, among others.

CORE AREAS DO NOT HAVE APPRECIABLY HIGHER AVERAGE EUR’S THAN THE PLAY OVERALL– In making this statement, Mr. Berman analyzed horizontal wells in Tarrant and Johnson Counties.  Analysts used the very broad brush of “Core”, “Tier 1”, “Tier 2”, and “Non-Core” early on when describing the potential productivity of the Barnett Shale.  This broad brush approach is not an appropriate discriminator for potential production because it isn’t related to anything other than a non-robust early estimate of Barnett thickness combined with a very poorly sampled grid representing potential thermal maturity.  A more robust method to define potential productivity is necessary; one that actually predicts potential on a well by well basis. For instance, we use Net Feet of Barnett Shale combined with a “Cooking Index”, which is derived from GOR that we call Graded Potential of Acreage (GPA), and that is updated periodically from new results.  From our analysis, Tarrant County itself, which is considered “Core” using the broad brush approach actually contains 8 of 10 grades of potentially productive shale by our more localized approach.  Choosing Tarrant County by itself thus does not represent a natural high grading of acreage, rather it is a random grab bag of various quality acreage.  Comparing the production quality of wells from several non-discriminated grades of acreage is a bad start.  Should Mr. Berman choose to look at production responses from a more meaningful subset of acreage, such as our methodology, he would see substantive differentiation in raw productivity between acreage grades.

Figure 1
Figure 1.  Note the strong correlation of production responses to GPA (Graded Potential of Acreage) which uses Net Barnett Thickness and a cooking index derived from the GOR of Barnett producers. Sample size is all Barnett Horizontal Wells in the FWB.

Figure 2

Figure 2.  Note the correlation of average cumulative production to GPA.  These cumulative are raw, actual numbers.  In order to derive PREDICTIVE POTENTIAL productivity, production would need to be normalized for operator learning effects, horizontal lateral length, and Best Practices operations.  

THERE IS LITTLE CORRELATION BETWEEN IP RATES AND EURs – Mr. Berman is correct in this statement if he means test AOF IP data, the least predictive of IP measurments.  However, the correlation between “IP” data and EUR begins to tighten up significantly if one were to use 2nd Month actual production as “IP”, and even more (90% correlation coefficient) using Maximum Monthly production rate as “IP”. The reason for this effect is because it often takes several months for a well to “clean up” after initial stimulation.  Also, using this methodology, maximum monthly rate correlates closely to  6, 12, 24, and 60 month cumulatives production amounts, and is thus an excellent proxy for EUR.

Figure 1

Figure 1.  Although Reported AOF IP doesn’t correlate strongly to EUR (cc=46%), Maximum Monthly Production Rate does (cc=86%).  This is due to the fact that it often takes months for the production stream to clean up after stimulation (sample size is all Barnett wells in the FWB since 2002).

Figure 2

Figure 2.  This relationship is demonstrably true throughout every producing grade of acreage in the Barnett, with correlation coefficient exceeding 90% in several cases (sample size is all Barnett Wells in FWB since 2002).

HORIZONTAL WELLS DO NOT HAVE APPRECIABLY HIGHER EUR’S THAN VERTICAL WELLS- This statement left us a bit puzzled. The charts below illustrate both the raw and normalized maximum monthly production rates that we can definitely tie to EUR.  The only thing we can conjecture is that Mr. Berman is looking at the exceptional response of multiple frac stages over time in vertical wells and comparing these to horizontals.  By the way, the frac response post initial frac can be statistically modeled as well, and evidence is emerging that horizontal wells can exhibit the same behavior.

Figure 3

Figure 3.  As we previously saw, there is a strong correlation between maximum monthly production rates and EUR.  Here we see average raw maximum monthly rates for horizontal wells exceeding vertical wells by a factor of 2 or more depending upon acreage grade.  All Barnett wells drilled since 2002. 

Figure 4

Figure  4.  The maximum monthly production rates after being corrected for operator learning curve.  Note that this is a BIG correction.  As shown in Figure 6, it typically takes operators a year or more to maximize their production.  Wells drilled at the top of the curve typically produce more than twice those drilled at the bottom.  The average boost across all grades of horizontals is a 94% improvement in maximum monthly production rates when corrected for learning curve and 77% improvement in vertical wellbores.

Figure 5

Figure  5.  Same figure as above but after normalizing for operator learning effect AND a 3000’ lateral length.  This corrects an average of 4% across all grades, but can be as high 10% depending on acreage grade.

Figure 6

Figure 6.  Basin wide Learning Curve.  To normalize for predictive analysis, we adjust wells drilled early in the learning stage to perform as well as wells late in the learning stage.  If production declines later on, this is taken as an indication of drainage and not that the operator is getting worse.  This is calculated for each of the top 20 operators and the basin-wide average applied to all other operators.

MARGINAL PRODUCER– Unlike conventional reservoirs, the economics of shale plays highly favor operators that invest in engineering and ongoing experimentation in optimizing their drilling, completion, and stimulation practices.  The saying “One Man’s Gold is another Man’s Garbage” is especially true for these plays.  For an equivalent grade of acreage, the best operators statistically produce 40% or more than the average operator, and up to 4 or 5 times more than a inefficient operator.  Learning curves and operator comparisons are real and quantifiable throughout.  As acreage expires, and the “land rush” acreages are released, the optimal operators will successfully step into “another man’s garbage”.

Figure 7

Figure 7.  In this actual example of the GPA type curves from two different operators, both would appear to be essentially equal.  Company A is considered, and is indeed, a good operator. Company B, with substantively lower GPA acreage, has extracted the same amount of production from much lower grade acreage.

Figure 8

Figure 8.  Now let’s compare the same production type curves from company A to company B for similar quality acreage (same GPA).  Company B is achieving nearly 50% more gas from the same quality acreage as Company A.

HORIZONTAL WELLS ARE NOT ALL THE SAME– There is a strong correlation between lateral lengths and production rates, especially at higher acreage grades.  These need to be normalized before meaningful analysis of horizontal well performance can be made.

Figure 9

Figure 9.  Note the strong correlation between lateral length and productivity.  This needs to be normalized in order to make meaningful performance comparisons for horizontal wells. 

THE VOLUMES OF COMMERCIALLY RECOVERABLE RESERVES ARE GREATLY OVERESTIMATED – This is a very definitive statement for which the jury is still out.  Mr. Berman is absolutely correct at current drilling and completion costs combined with $2.50 MCFG wellhead price.  However, resource turns into undeveloped reserves fairly quickly as wellhead product price goes up or costs come down.  Reducing the cost of drilling and completing a well by 50% is the economic equivalent of doubling the lifetime wellhead price.  It helps to think of breakthroughs in reducing drill, complete, and stimulation costs as “permanent hedges”.  One cannot address any play on wellhead cost alone.  We know that there is a 80 to 100 x differential in play reserves between static D,C,&S costs and $7/MCFG at the wellhead and $2.50/MCFG at the wellhead.

AVERAGE WELL PERFORMANCE HAS DECREASED CONSISTENTLY SINCE 2003 FOR HORIZONTAL WELLS – Has it?  On the surface, Mr. Berman is correct.  However, the quality of acreage being drilled horizontally has also been steadily decreasing as a proportion of acreage drilled.  Once wells are properly allocated to their proper GPA, normalized for lateral length, normalized for operator learning curve, that effect disappears.

Figure 10

Figure 10.  Note the steady decrease in acreage quality drilled by horizontal wells since mid 2003, bottoming out in mid 2007, and increasing slightly since then.

HYPERBOLIC PREDICTIONS OVERESTIMATE RESERVES – Mr. Berman is absolutely correct here.  This is also usually the case for conventional reserves.  Hyperbolic predictions typically provide a ceiling estimate.

CONCLUSIONS – As we discovered when we began analyzing the Barnett, analyzing shale plays is like peeling onions.  Once a layer is exposed, another presents itself.   Many factors need to be normalized to really address the economics and behaviors of these plays in a predictive sense.  Good acreage grade alone is not a recipe for success.  Decent acreage grade needs to be combined with strong, holistic drilling, completion, and stimulation practices that are constantly tested for optimization to create a real repeatable recipe for success and large economic reserve accumulation.  Great operations equate to better recoverable resources/reserves for these plays, not more operators or even more wells.

Raw data combines good operations with bad, early parts of the learning curve results with later, and good acreage with bad.  Making forward-looking predictions of behavior from this soup assumes that every bad habit is propagated forward, and that no operator is better than another. Expanding that conclusion to other plays based on this assumption misses the big picture. What the Barnett is capable of producing is different from what it is producing now.  How it can be optimized can only be determined by identifying who is optimizing it and by how much.  It is ironic that we are concerned about profitability and deliverability since it was only a few years ago when we were told that Hubbert’s Peak was close at hand for US natural gas.  Yes, there will be a lot of dead producers when all is said and done, but for those that invest in operational excellence, the Barnett Field and other unconventional plays will go on making tens or hundreds or thousands of people very wealthy, and providing much needed domestic hydrocarbons for years to come.

For more information, check out  https://www.diold.lo/demos/flashDemos/fwbasin/index.htm

Download the PDF Version of this post

The following two tabs change content below.

Allen Gilmer

Allen Gilmer is the Co-founder, Chairman and CEO of Drillinginfo. Allen is active in all aspects of Drillinginfo’s new product development and is widely recognized for his industry leadership and vision. He holds several patents in the field of multi-component seismology. He received his Bachelor of Arts in Geology from Rice University and his Master of Science in Geology from The University of Texas at El Paso. Follow him on Twitter @allengilmer.