StatisticalAnalysisWolfberry_CEKEngineeringLLC

Statistical Analysis of the
Wolfberry Using “R”
Infill Drilling Study (80 ac. to 20 ac. Spacing)
Chad E. Kronkosky, P.E
President, CEK Engineering LLC
Pursuing Ph.D. “Part Time”, TTU Petroleum Engineering
Anticipated Candidacy Fall 2014
SPEE Midland Chapter
September 10, 2014

Copyright © 2014 CEK Engineering LLC
Presenter Biography
Introduction To CEK Engineering LLC and its President
Chad E. Kronkosky, P.E.
 Education
B.S. in Petroleum Engineering (TTU 2006)
M.S. in Petroleum Engineering (TTU 2009)
Ph.D. in Petroleum Engineering (TTU ~ Anticipated Candidacy Fall 2014)
 Industry Experience
CEK Engineering LLC (2012 to Present)
President of CEK Engineering LLC (CEK). CEK was formed to provide Professional Engineering Services to
the Oil and Gas Industry. Our clients range from Large Private Equity Management Teams to Small Independent
Oil and Gas Operators. We have project experience in Texas, Louisiana, New Mexico, Kansas, Colorado, Montana,
and North Dakota, but our primary focus is the Permian Basin of West Texas.
For more information on CEK Engineering LLC, please visit www.cekengineering.com
Bold Operating LLC (2010 – 2012)
Reservoir Engineer for small Private Equity Management Team (EnCap). Solely focused in the Permian Basin;
Grow By The Bit Company. Gained valuable financial experience; worked with lending institutions, private
equity analysts/managers, etc.
A.C.T. Operating Company (2006 – 2010)
Graduate Petroleum Engineer working under Marshall Watson, Ph.D., P.E. Current Chair of the Petroleum
Engineering Department At TTU. Experience encompassed: Secondary Recovery Projects, CBM, Corporate
Management, Prospect Development, etc. (almost anything you can imagine, especially for a two man company).
The man in red above is a lot of the reason I am giving this presentation today; Thank You Marshall for all you
have done for me.

Statistical Analysis of the Wolfberry Using “R”
This Presentation Is Not About
Statistical Theory!
“I’m Sorry To Disappoint The Academics In The Audience”
This Presentation Is Going To Discuss Observations Of
Improper Use/Analysis Of Data!
We Will Also Discuss “Tools” (Namely “R”) That Can Be
Used To Help Us Understand Our Data.

Presentation Outline
“Conceptually”
Statistical Data Analysis & Visualization
 What’s The Problem With Our Current Methodology? Is There A
Problem?
 Is There A Solution To This Supposed Problem?
 Brief Discussion On My Research And How “CEK” Looks At
Resource Plays.
Wolfberry Infill Drilling Example
 Aggregation Of Production Data.
 Why Is My Unconventional Reservoir Behaving Conventionally?
• We need to understand well performance and completion data to answer this!
(Especially For The Small Operators!)
(Why It’s Important To Get Right!)
(Just Frac The C^@P Out Of It… Right?)

What’s The Problem With Our Current Methodology?
Is There A Problem?
“Damn Excel! How the ‘Most Important
Software Of All Time’ Is Ruining the World!”1

Is There A Problem? – “Damn Excel!”
Current Methodology/Practices
 Gather Data/Generate Data
• We take historical records and store them!
– In Spreadsheets!
– In Word Processing Documents
– In Databases
 Perform (for the most part ad hoc) Analyses
• We take disparate data stores and generate masters data files.
In most instances, this master data file “starts/ends” as an Excel file.
• We Seldom QA/QC our Data!
• We perform analyses (write in-line formulas, etc.) on these master
data files to generate some form of visual communication (plots,
tables, etc.).
• We typically solve most of our problems with a single spreadsheet.
(if need be we can export this data to a professional software
package).
 We Then Repeat The Process At A Later Date!
• This means updating our ad hoc data files (typically manually)
(+80% of the time)
(This is where the headache starts!)
(companies actually populate databases?)
(extremely important but who has the time)

So What’s The Big Problem?
 Who Here Has Actually Ever Back Check
A Spreadsheet?
• Is this even possible once we start building linked
formulas spanning multiple sheets/workbooks?
• Is there a method to our madness that can be followed
by someone a lot less technically inclined ?
• Better yet, could you defend yourself in a Court of
Law with your analysis?
(I’m sure there is a spreadsheet Guru In here, We Know Who You Are!)

London Whale
 2012 – JP Morgan Chase 6.2 Billion Dollar Trading
Loss!
“In the wake of last year’s $6.2 billion JPMorgan Chase JPM trading loss, traders have been
fired, top executives have been hauled in front of Congress, and the FBI, among other
regulators, is investigating. But you know who really needs to be questioned? Bill Gates.
According to an internal report on the trading loss released in February, the model that was
supposed to monitor and limit the amount of risk the bank’s London traders were taking was
“operated through a series of Excel spreadsheets, which had to be completed manually,
by a process of copying and pasting data from one spreadsheet to another.” One key
measure was added when it should have been averaged. The result: Risk officers at
JPMorgan believed the credit derivatives bets were half as risky as they actually were.
So, I guess, CEO Jamie Dimon can pass $3.1 billion off on Excel. The rest is still on him.

MF Global
 October 2011: MF Global Transfers Client
Account Funds To Its Own Account!
“About a year before MF Global went bust, consultants hired by the firm determined it
needed to improve the “end user computer tools such as Excel spreadsheets” that the
commodities broker used to monitor risk and how much money it had in its customers
accounts, and to make sure that some of that money didn’t end up in the account being
used by CEO Jon Corzine to bet on whether or not Europe was about to implode. Those
upgrades were never made.”1

Barclays – Lehman Bros. Purchase
 September 2008 Collapse – Purchase Offer
“When Barclays sent over its offer to buy up Lehman Brothers in the immediate wake of the
firm’s September 2008 collapse, it did so with an Excel spreadsheet. The makers of the
spreadsheet, which detailed Lehman’s assets and what Barclays was willing to buy, hid,
rather than deleted, nearly 200 cells. But when a junior law associate at Cleary Gottlieb
Steen & Hamilton converted the Excel file to a PDF and e-mailed it over to the
bankruptcy court, the hidden parts of the spreadsheet reappeared. The result: Along
with the parts of Lehman Barclays wanted, the British bank was also forced to swallow
losses on an additional 179 toxic deals it never intended to buy.”1

Important Takeaway
 Humans Are Prone To Errors!
 Our World Has Gotten More And More Complex!
• Our industry is adapting to Probabilistic Methods and workflows;
Oracle “Crystal Ball” and Palisade “@Risk”. Five years ago only
a small group of people were using these tools.
• The current activity in the industry (predominantly Unconventional
Reservoirs; 1925 Total Rigs ~ 70% Horizontal as of 9/5/2014) is
stressing our ability to make “Good Decisions” in my opinion.
Tremendous Amounts of Capital are being wasted in part due to our
workflows being behind the curve.
 We Live In A Button Click Society!
• Engineering Judgment is slowly being eroded/replaced by the use of
software. All you do is put straight lines through data points, right!
• Young engineers (and non-engineers) are exposed to these “Black
Box” solutions and immediately start “Button Clicking” their way
to the “answer”.
 As we now know “Button Clicks” do have
consequences !

Is There A Solution To This Supposed Problem?
“We Live In A Complex World That Requires Simplified Solutions!”

Solutions Are Only As Limited As We Limit Ourselves
There’s Lots Of Solutions!
 Microsoft Office Solutions
• Visual Basic for Applications
– User Defined Function (UDF’s) instead of complex in-line
functions.
• MS Access is our friend!
– Once datasets grow beyond a few spreadsheets we really need
to start thinking about generating a database. MS Access is
about the easiest one to learn and use. P.S. A lot of industry
software use MS Access databases as their data store.
 Third Party Applications
• SLB’s Oil Field Manager, Merrick’s RIO, etc.
– These are great off the shelf data stores & management tools.
• Oracle’s “Crystal Ball Suite” and Palisade’s “Decision Suite”
– Monte Carlo simulation, Decision Trees, Sensitivity Analysis,
Statistical Analysis, Neural Networks, Optimization.
Integration with VBA for automation.
 Open Source Software – “R”

Solutions Are Only As Limited As We Limit “R”selves
There’s Lots Of Solutions!
 Microsoft Office Solutions
• Visual Basic for Applications
– User Defined Function (UDF’s) instead of complex in-line
functions.
• MS Access is our friend!
– Once datasets grow beyond a few spreadsheets we really need
to start thinking about generating a database. MS Access is
about the easiest one to learn and use. P.S. A lot of industry
software use MS Access databases as their data store.
 Third Party Applications
• SLB’s Oil Field Manager, Merrick’s RIO, etc.
– These are great off the shelf data stores & management tools.
• Oracle’s “Crystal Ball Suite” and Palisade’s “Decision Suite”
– Monte Carlo simulation, Decision Trees, Sensitivity Analysis,
Statistical Analysis, Neural Networks, Optimization.
Integration with VBA for automation.
 Open Source Software – “R”
During the last decade, the
momentum coming from
both academia and industry
has lifted the R
programming language to
become the single most
important tool for
computational statistics,
visualization and data
science.2

Data Analysis Software
When Excel Just Doesn’t Cut It!
 Excel As A Data Analysis Software
• Excel is a poor choice for statistical analysis beyond textbook
examples, the simplest descriptive statistics, or a few columns of
data. Was never designed to fulfill the Data Analysis role, add-on
packages are required (Crystal Ball, Decision Suite)!
 Commercial Data Analysis Software
• SAS, SPSS, Stata, Minitab, Tibco Spotfire, Tableau etc.
 Open Source “R”
• Pros
– Arguably one of the most popular Data Analysis Software out
there…Free! Also, most commercial packages support “R”
– Open Source! Huge user base! And 1000’s of Libraries of
source code available to download.
• Cons
– Command Line Interface (Sorry Button Clickers). RStudio is
an elegant IDE for programing.
– Support is limited to user blogs, stackoverflow.com

Data Analysis Software – “R”
What Does “R” Look Like?

There Are Lots Of Solutions…We Need To Choose The Right Ones
Our Industry Is Generating Tremendous
Amounts Of Data…
The Problem Now Is Not Having Data…
But How Do I Interpret This Data !
Efficiently!!!
We Now Entering
The I-Field !

Brief Discussion On My Research And
How “CEK” Looks At Resource Plays
“The Goal Is To Work Smarter…Not Harder!”

Brief Discussion On My Research
Data Analysis Tools For Engineers By Engineers
We Need Better Tools!
 Excel Is A Great Software, But Is Not The Right
Tool For Everything!
• Excel is a wonderful data organizer / cleanup tool.
• In-line formulas are difficult to debug to say the least.
• VBA can help organize our thought process.
• Third Party Applications can help extend Excel’s capabilities.
 Data Analysis Software Is Great, But For The Most
Part Was Developed For Data Scientist
• Data Analysis Software can solve problems in a few lines of code
that would take 100’s of lines of code in Excel VBA with Third Party
Applications.
• We need a better way to document our work. “knitR” package can
allows the integration of “R Code” with LaTex, HTML, Markdown;
this allows us to have reproducible research through means of
literate programing.
• We need Data Analysis Software Build for Engineers by Engineers
“rShiny” package can allow us to build web apps for interactivity.

We Do A Poor Job Aggregating Our Datasets
Aggregating Correctly Is Tough!
 We Aggregate To Much & Don’t Explore Our Data
• We need to subset our data into subpopulations before we blindly
start making sample distributions to simulate!
– This is difficult to do. In Excel this is typically a tedious
ad hoc process; even with Third Party Applications.
– To do this correctly we perform:
“Independent Two-Sample T-Test” for normal distributions,
ANOVA, and “Likelihood-Based” or “Bootstrap” methods
for log-normal distributions.
• Combining Sample Subpopulations smears the mean of our data,
invariably in the wrong direction! (i.e. we increase EUR’s)
 Our Data Is Typically Non-Normal
• This presents a major challenge as most of the statistical techniques
engineers are taught assume normality!
– We use linear models (i.e. least squares) on data that is
non-normal!
 Lies, damned lies, and statistics!
Do we do this?
We should use generalized linear models or
non-linear models. Is the engineering community informed
with these techniques?

Brief Discussion On My Research &
I Intend To Generate Software For “Engineers”
Which Will Allow Them To Explore Their Data
Efficiently
My Firm Currently Uses This Software
In Workflow For Our Clients
In Workflows “Engineers” Are Accustom To!

WolfBerry Infill Drilling Example
Aggregation Of Production Data
“Or Improper Aggregation Of Production Data”

Normalized Production Type Curves
Aggregated By Start Date Year (Rate Time Semi-Log Plot)

Aggregated By Start Date Year (Rate Time Log-Log Plot)
half-slope – linear flow period
unit-slope – boundary dominated period?
Note: Horizontal axis is exaggerated, thereby deceiving actual slopes

Aggregated By Start Date Year (Rate Time Log-Log Plot)
Note: Horizontal axis is exaggerated, thereby deceiving actual slopes
quarter-slope – bilinear flow period
half-slope – linear flow period

Aggregated By Start Date Year (Rate Cumulative Oil Production Cartesian Plot)
Decreasing w/ Aggregated
Start Date Time

Aggregated By Start Date Year (Rate Cumulative Oil Production Semi-Log Plot)
Straight Lines On This Plot = Harmonic Decline (b = 1)
Note: Sudden Change In Decline Rate
(2010 – 2011). This coincides
to the infill drilling campaign!
106 MSTB
78 MSTB
~ 25% Decrease In EUR Due To
Decline Rate Increase
(Boundary Dominated Flow?)

2008
78 MSTB
2009
71 MSTB
2010
60 MSTB
2011
57 MSTB
2012
50 MSTB
2008 – 20 Wellbores
Cumulative Wellbore Counts
~ As Infill Drilling Continued On
This Project; There Was An
Precipitous Decline In EUR’s
EUR’s Decreasing

Average
74 MSTB
2008
78 MSTB
2009
71 MSTB
2010
60 MSTB
2011
57 MSTB
2012
50 MSTB
Note:
Average EUR Decline Curve
Approaches The 2008
Aggregate Projection

Implications Of Improper Aggregation
Start Date
Year
# of Wells EUR / Well
MSTB / Well
EUR Total
MSTB
2008 20 78 1,560
2009 14 71 994
2010 53 60 3,180
2011 103 57 5,871
2012 85 50 4,250
Total 275 15,885
Average Oil EUR / Well = 15,885 MSTB / 275 Wells ~ 58 MSTB / Well
Aggregate Wells By Start Date Year And Summarize
Why Not Use The
Average Normalized Decline Curve?

74 MSTB
If We Include Tail End Data…
We Might Be Tempted To Use A
Hyperbolic Exponent b > 1 (i.e. concave up)
120 MSTBTail End Effects Are Dominated
By Early Time Well Completions
Note The Jaggedness Of Curve
Caused By Relatively Few Wells.
Note Increase In Serration
Caused By Decreasing
Aggregation Counts

Start Date
Year
MSTB / Well
EUR Total
MSTB
2008 20 78 1,560
2009 14 71 994
2010 53 60 3,180
2011 103 57 5,871
2012 85 50 4,250
Total 275 15,885
Average Normalized “Type Curve” = 74 MSTB / Well
Reserves / Resources Overstated By 25% Using Average Normalized “Type Curve”
20,350 MSTB instead of 16,110 MSTB
Or Worse… If We Were Tempted To Used A Hyperbolic Exponent > 1 Based On Our
Average Normalized Decline Curve…
We Could Be Overstating Reserves / Resources By 110%
(120 MSTB vs. 58 MSTB).

Start Date
Year
MSTB / Well
EUR Total
MSTB
2008 20 78 1,560
2009 14 71 994
2010 53 60 3,180
2011 103 57 5,871
2012 85 50 4,250
Total 275 15,885
Average Normalized “Type Curve” = 74 MSTB / Well
Undeveloped Reserves / Resources Will Also Be Overstated Using The
Average Normalized “Type Curve” Versus Most Current EUR Projection (2012)
Overstated By ~ 50% (74 MSTB vs. 50 MSTB)
For This Project Area, This Equates To
960 MSTB or 2,800 MSTB (40 - Undeveloped 40 ac. Locations x 21 or 70 MSTB) Of Reserves / Resources Overstated.
or Worse, Overstated By ~ 140% (120MSTB vs. 50MSTB).

Aggregation Of Production Data
Aggregation Errors Such As These
Have Serious Economic Implications!

WolfBerry Infill Drilling Example
Why Is My Unconventional Reservoir Behaving
Conventionally?
“A Tale Of Two Reservoirs” – Literally!

12 Month Cumulative Oil Production
Aggregated By Start Date Qtr. & Lease
Note:
Box Plots Represent (min, max values – end point of
black line tails; base of box = 25th percentile; black line in
box = mean; top of box = 75th percentile). Black Dots are
potential outlier data points.
Colored By Lease.
Note:
In all instances, initial wells outperform
subsequent wells. Statistically this is
impossible; also this violates our
definition of a resource play at least
initially.
Note:
Late time wells exhibit narrower
distributions; and more consistent results.
At this point our “resource play” rock is
dominating production.

12 Month Proppant Normalized Cumulative Oil Production
Note:
Box Plots Represent (min, max values – end point of black line tails; base of
box = 25th percentile; black line in box = mean; top of box = 75th percentile).
Black Dots are potential outlier data points.
Colored By Lease.
Note:
initially.
Note:

24 Month Cumulative Oil Production
Note:
Colored By Lease.
Note:
initially.
Note:

24 Month Proppant Normalized Cumulative Oil Production
Aggregated By Start Date Quarter and Lease
Note:
Colored By Lease.
Note:
initially.
Note:

Total Proppant Amount
Aggregated By Frac Style, Maximum Monthly Rate, and Lease
Note:
Average Completion used ~ 1000 Mlb of proppant
with a maximum variance of ± 200 Mlb of proppant
from 2008 to 2012.
Note:
Average Completion in 2013 used ~ 1400 Mlb of proppant
~ +40% increase in historic use amount.

24 Month
Cumulative Offset Oil Production (640 ac.) - Prior To Start Date
Aggregated By Lease or Section
Note:
Color Band Represent
95% Confidence Interval
Note:
Inverse relationship of 24 Month
Cumulative Production vs. 640 ac.
Total Offset Production At Start Date.

24 Month Proppant Normalized
Cumulative Offset Oil Production (640 ac.) - Prior To Start Date
Aggregated By Lease or Section
Note:
Color Band Represent
95% Confidence Interval
Note:
Inverse relationship of 24 Month
Cumulative Production vs. 640 ac.
Total Offset Production At Start Date.

24 Month Cumulative Water Production
Size Of Attribute Relative To Amount Of Produced Water
Note:
Larger data points clustered in
the eastern portion of this plot
indicate that this area produces
Significantly more water then the
western portion of this plot.
Only one well is in the
10th percentile in this area; whereas
the western portion of this plot has
Zero wells in the 90th percentile.
Transition Zone?
Unconventional Reservoirs (PRMS)
Exist in petroleum accumulations that
are pervasive throughout a large area
and that are not significantly affected
by hydrodynamic influences (also
Called “continuous –type deposits”)
(Emphasis Mine)
Conventional Reservoirs (PRMS)
Exist in discrete petroleum accumulations
related to a localized geological structural
feature and/or stratigraphic condition,
typically with each accumulation
bounded by a downdip contact with an
aquifer, and which is significantly
affected by hydrodynamic influences
such as buoyancy of petroleum in water.
(Emphasis Mine)
It Appears That
Some…If Not All
Of The Completed
Reservoirs Are Behaving Conventionally?

Why Is My Unconventional Reservoir Behaving Conventionally
You Might Not Know This Unless You Explore Your Dataset!
Not All Resource Plays Are
Unconventional Reservoirs!
 I’m Sorry, I Mean…Not All Areas/Zones Of A
Resource Play Are Unconventional Reservoirs!!
 When Analyzing These Plays, We Must Always
Remain Cognizant Of This Fact!
• Tight basin center Shales eventually grade to High Permeability
Platform Carbonates!
• Just because an Operator filed a Completion Reports testifying to the
fact that the wellbore’s completed reservoir is in a “Resource Play /
Unconventional Reservoir”… doesn’t necessary make it so.
 The Only Way We Can Perform Accurate
Estimates Is To Explore Our Datasets As Deeply As
Possible! Hopefully With “R”
Operators
Are known to make errors from time to time; not always unintentionally.
Don't Blindly Start Aggregating Wells Based Solely On Field Names!
Outlier Analysis can generally find these wells.

Questions?
Oh…Please Have At Least One!
Questions?
Thank you for the opportunity
to present this to you

References
1.) Damn Excel! How the ‘most important software of all time’ is ruining the world.
http://fortune.com/2013/04/17/damn-excel-how-the-most-important-software-application-of-all-time-is-ruining-the-world/
2.) What is R
http://www.revolutionanalytics.com/what-r
Images were gathered from various internet sources. Our intention is to not infringe upon any copyrighted work. If we have done
so, I deeply regret this error! Please except out apology and we will be happy to credit, or remove your work.
Respectfully,
Chad E. Kronkosky, P.E.
President
CEK Engineering LLC
5139 69th Street
Lubbock, TX 79424
(806) 702-8954
www.cekengineering.com

StatisticalAnalysisWolfberry_CEKEngineeringLLC

Recommended

Recommended

More Related Content

Viewers also liked

Viewers also liked (20)

Similar to StatisticalAnalysisWolfberry_CEKEngineeringLLC

Similar to StatisticalAnalysisWolfberry_CEKEngineeringLLC (20)

StatisticalAnalysisWolfberry_CEKEngineeringLLC