1. The SPE Foundation through member donations
Primary funding is provided by
The SPE Foundation through member donations
and a contribution from Offshore Europe
The Society is grateful to those companies that allow their professionals to serve as lecturers
Additional support provided by AIME
Society of Petroleum Engineers Distinguished Lecturer Program

2. Two Vital Secrets for Building Reliable Type Wells
Randy Freeborn
My name is Randy
Today I plan to share 2 secrets that I have learned from 5 years of researching type wells.
When you to return to work, I hope you will only use type wells built with history and prediction.
Society of Petroleum Engineers Distinguished Lecturer Program

3. AGENDA
TYPE WELL What is a type well The challenge 1st SECRET All type wells 3 Inherent Errors Case Study 2nd SECRET Probability type wells Time slice method Aggregation method Comparison WRAP UP
I will begin by introducing you to type wells,
then the talk will be split into two equal parts, each part dealing with one of the secrets.

4. TYPE WELL What is a type well?
Rate-time production profile
Shift representative wells to a common start date
Average them to represent new wells
Common method comprised of two parts
History average rate until too few wells
Prediction projection of best fit of history
Rate-time profile
Shift representative wells to a common start date
Average them to represent new wells that we plan to drill
Part 1 – history
Part 2 - prediction

5. TYPE WELL The Challenge
Dr. Lee, 2015 Reserve Summit
SEC’s experience (circa 2008) type wells exceed results by about 25%.
2013 Proprietary Research Report
Drilling results did not meet the objectives set out in 40 of 100 published play specific type wells.
Only 14 of 40 companies consistently met targets.
Personal Experience
EUR more likely to be over estimated, as much as 40%.
The challenge of obtaining accurate type wells comes from 3 factors that are progressive
1. First, type wells are pervasive, used in all aspects and disciplines of our business
2. Second, the decisions we make based on type wells are capital intensive and therefore important
3. Finally, the most popular methods used to create type wells have inherent errors, than may cause us to make bad business decisions
Dr. Lee is best known for his role in modernizing the SEC regulations
At a recent conference he advised that the SEC is experiencing type wells that exceed results by about 25%.
In a 2013 research report
40 drilling programs failed to resultDrillin40 of 100 companies
My personal experience is that type wells are commonly created using only historical data. Hind casting studies show those type wells are more likely to overstate EUR by an amount as high as 40%.
Pervasive … Capital Intensive … Errors

6. 1st SECRET Applies to All Type Well Methods FORECAST EACH WELL THEN AVERAGE HISTORY & PREDICTION
The 1st secret applies to all type wells and
is related to 3 errors that are inherent to constructing time wells using only historical data.
You can avoid all of the errors by forecasting each well, then
building the type well by averaging history and prediction.

7. 3 ERRORS Forecast groups Survivor bias
Never forecast groups, always group forecasts
Survivor bias
Concentrating on things that survive
The treatment of depleted wells
Depleted wells produce with rate = 0
No production (declining well count)
Recent wells that have no rate to average
Also a form of survivor bias
Use best available forecast
There are three errors inherent creating type wells using only historical data
With only history, we average first and then forecast from the average – I call this forecasting groups
We should always forecast first and then average – In my words group forecasts
Survivor Bias - the logical error of concentrating on the wells that "survived“
We fail to concentrate on depleted wells by dropping them from the calculation post depletion
We shift wells to a common start date, sometimes termed normalizing
Newer wells will not have produced long enough to have data for averaging
We also fail to concentrate on these wells.
You will see the nature of these 3 errors,
and learn they will all be avoided
when you forecast 1st and then average history and forecast to construct type wells
SPE &

8. 3 ERRORS #1 Forecast Groups
I will use a hind casting example to demonstrate the concept of grouping forecasts
9 wells producing for 27 continuous years – there is no survivor bias in the data
We truncate the data to 8 years to build a type well
Use the remaining 19 years to confirm results
This is the Common Method
The production data has been averaged
and then best fit to complete the type well
Often hard to fit because of multiple or unclear trends
In this case this data is good, leading to a high quality best fit
In a moment you will see that the business decision would have been bad
When we Group Forecasts
We see trends in the data that are often masked in a group
Inaccuracy in the forecasts cancel, resulting in higher quality
Average history and prediction
Observe the resulting type well is near perfect
More work to forecast, but reliable auto-forecasting tools are available
Large difference between grouping forecasts and forecasting groups is not uncommon
Forecast Groups
Usually no clear trend
High quality best fit Bad business decision
Grouping masked a trend
Group Forecasts
New trends are visible
Forecast errors cancel
Accuracy improves
Type well is accurate

9. 3 ERRORS #2 Survivor bias
We have 3 producing wells
After 3 years one well is depleted, so we average only 2 wells
There is an unintended consequence when the well count is reduced
the type well rate behaves as though the depleted well continues to produce at the average rate
Every well must have a rate
Those that do not have a rate are assumed to continue producing at the average rate
The correct treatment is to assume the depleted well continues to produce at a rate of 0
Common method Depleted rate = type well rate Creates false rate and reserve
Correct treatment Each well must have a rate
SPE

10. 3 ERRORS #2 Survivor bias Common method Compounding effect SPE 162630
We have 3 producing wells
After 3 years one well is depleted, so we average only 2 wells
There is an unintended consequence when the well count is reduced
the type well rate behaves as though the depleted well continues to produce at the average rate
Every well must have a rate
Those that do not have a rate are assumed to continue producing at the average rate
The correct treatment is to assume the depleted well continues to produce at a rate of 0
Common method Compounding effect
SPE

11. AVOID ALL 3 ERRORS Forecast, then average history & prediction
3 ERRORS #3 No production
Common method Well rate = average rate Best wells drilled first
Correct treatment Include every well Use best available forecast
We have 3 producing wells, then one well runs out of production
It is still producing, but hasn’t produced long enough to have a rate for averaging
As with the depleted well, the adverse consequence is that missing production gets the average rate
That average rate may be OK, but remember we usually hi-grade by drilling our best wells first
The new “worst” well gets the average rate from the best wells
Replace the intrinsic forecast with your own
Forecast wells, the average both history and forecast
AVOID ALL 3 ERRORS Forecast, then average history & prediction

12. 3 ERRORS Numerical example
We have 3 producing wells
After 3 years one well is depleted, so we average only 2 wells
There is an unintended consequence when the well count is reduced
the type well rate behaves as though the depleted well continues to produce at the average rate
Every well must have a rate
Those that do not have a rate are assumed to continue producing at the average rate
The correct treatment is to assume the depleted well continues to produce at a rate of 0

13. CASE STUDY 88 Hugoton Kansas wells
Hind casting evaluation of shallow gas wells
5 years of drilling, plus another 5 years of production
5 to ten years of production for average or forecast
Common method
75% cut off – trade off, more data, potential error
best wells first – wells with no production cause higher rate.
All panels look like reliable type wells.
Data truncated 5 years drilling + 5 years producing
Cut off Stop when too few wells
Type well Looks reliable
SPE

14. CASE STUDY 88 Hugoton Kansas wells
Common method yellow line from the previous slide, error band
EUR accuracy about 15%, but rate-time profile terrible
Forecast the wells, then average the history and prediction
Visually compare the type well in white to the measured data in blue
We obtain an accurate rate-time profile with an EUR accurate within 2% after 32 years
1st SECRET average history & prediction
SPE

15. What is uncertain? (EUR, Present Value, Cash Flow, …)
2nd SECRET Applies to Probability Based Type Wells STOP USING THE TIME SLICE METHOD USE THE AGGREGATION METHOD
You must specify 3 things
The Probability of what – EUR, NPV, Other?
Level of certainty
P90 uncertain parameter with meet or exceed the specified value 90% of the time
How many wells will be evaluated with this type well
Certainty (P10, P50, P90)
What is uncertain? (EUR, Present Value, Cash Flow, …)
How many wells?

16. TIME SLICE METHOD Uses only history Normally P10, P50 or P90
For Each Month
Sort by rate
Get the P90 or P50/P10 rate
Decline to complete
Time Slice method
Today’s most popular method, used in most computer software.
Used to find P10/P50/P90 without regard for the number of wells being drilled.
SPE &

17. TIME SLICE METHOD Probability What is uncertain? Unknown
No Aggregation (1 well)
Rates from the full distribution
Ignores EUR distribution
Time Slice method
Today’s most popular method, used in most computer software.
Used to find P10/P50/P90 without regard for the number of wells being drilled.
SPE &

18. TIME SLICE METHOD 9 well example There is a P10 & P90 well
This slide intended to show you why the Time Slice method is wrong
EUR distribution
9 wells: there is a P90 and P100 well
Rate Time chart
Yellow lines are the P10 and P90 rate-time profiles
Orange lines are the rate-time profiles use min or max production
When sorting the P10 = max rate and P90 = min
The blue area shows where the P10 type well rate is replaced with the maximum or
where the P90 type well rate is replaced with the minimum rate
This is a one way exchange, the P10 type well rate and EUR will always get bigger and the
P90 type well rate and EUR will always get smaller.
Illustrated in blue on the left chart
The P10 to P90 ratio will be too high, inferring greater risk than supported by the EUR distribution
9 well example There is a P10 & P90 well
Crossing rate/time Creates additional error
SPE

19. TIME SLICE METHOD Shaded area Rate < P90 or Rate > P10
This slide intended to show you why the Time Slice method is wrong
EUR distribution
9 wells: there is a P90 and P100 well
Rate Time chart
Yellow lines are the P10 and P90 rate-time profiles
Orange lines are the rate-time profiles use min or max production
When sorting the P10 = max rate and P90 = min
The blue area shows where the P10 type well rate is replaced with the maximum or
where the P90 type well rate is replaced with the minimum rate
This is a one way exchange, the P10 type well rate and EUR will always get bigger and the
P90 type well rate and EUR will always get smaller.
Illustrated in blue on the left chart
The P10 to P90 ratio will be too high, inferring greater risk than supported by the EUR distribution
Shaded area Rate < P90 or Rate > P10
P90 low, P10 high Where is the EUR right?
SPE

20. TIME SLICE METHOD Probability of what? Prone to error Disadvantages
Cannot choose at value , e.g. EUR, NPV
Type well does not match the EUR
Prone to error
Errors from using only history
Crossed rate-time profiles
Rates selected from all wells and probabilities
Doesn’t represent a defined group of wells P90 rates from 19 of 25 wells, P4 to P96
Disadvantages
Inset is 88 well Hugoton using 23 years of data.
Read Bullets

21. AGGREGATION METHOD Resolves 4 type well questions The Approach
Step 1 Get Target EUR (237)
Step 2 Weighting Factor
Continue 5 well trials
When mean ~ target Tally the selected wells
Tally more than 100 trials
Calculate weighting factor as a % of the total tally
Step 3 Build type well
Multiply history and forecasts by the weighting factor and sum
Resolves 4 type well questions
Which wells to use?
Should wells have equal weighting?
How does one account for drill program size?
What is the right way to handle probabilities?
The Approach
Find appropriate weighting factors
Step 2 & 3
Follow directions on the slide
SPE

22. AGGREGATION 101 Aggregated Distribution Pick 5 random probabilities
Get values for each
Average the values

23. AGGREGATION 101 Aggregated Distribution Aggregated Results
Pick 5 random probabilities
Get values for each
Average the values
Repeat 100,000 times
Plot distribution of means
Aggregated Results
P90 & P50 values increase
Certainty improves P10/P90
P90 economic with 5 wells

24. AGGREGATION METHOD Step 1 Get Target EUR (237) Step 2 Weighting Factor
Continue 5 well trials
When mean ~ target Tally the selected wells
Tally more than 1000 trials
Calculate weighting factor as a % of the total tally
Step 3 Build type well
Multiply history and prediction by the weighting factor and sum
Step 2 & 3
Follow directions on the slide

25. AGGREGATION METHOD Step 1 Get Target EUR (237) Step 2 Weighting Factor
Continue 5 well trials
When mean ~ target Tally the selected wells
Tally more than 1000 trials
Calculate weighting factor as a % of the total tally
Step 3 Build type well
Multiply history and prediction by the weighting factor and sum
Step 2 & 3
Follow directions on the slide

26. AGGREGATION METHOD Advantages Designed for new drilling
Based on probability of drilling each well
Properly uses aggregated probabilities
Will use any uncertain parameter
Proper ratios for secondary products
Calculated with the correct weighting
Aggregation
Increases P90 & P50 reserves
Adds certainty
Advantages
Type well calculated using aggregation method is shown and
Compared to the one using the Selected Well method.
Selected well is my second choice for building type wells.
The type well is build from a subset of the analogs that have an
uncertainty parameter (NPV) similar to the target (2.2 million)
The answers are statistical, I do not know which is better.
I would choose Aggregation because the wells are weighted in proportion to
the probability of their being drilled in a 35 well program.
Advantages Read list

27. COMPARISON P90 type wells
Method is critical
I choose the aggregation method

28. COMPARISON P90 type wells

29. TWO VITAL SECRETS As a Type Well Builder As a Consumer of Type Wells
Average both history and prediction
Use Aggregation method for new drilling
As a Consumer of Type Wells
Avoid type wells that use only historical data
Type wells should represent the number and quality of wells you plan to drill
Read the list

30. Your Feedback is Important
Enter your section in the DL Evaluation Contest by completing the evaluation form for this presentation
Visit SPE.org/dl
Society of Petroleum Engineers Distinguished Lecturer Program 30