The Management of Oil Industry Exploration & Production Data

Who are we? Purchase Options - B&W USA/World - B&W UK - B&W Germany - B&W France - Color USA/World - Color UK - Color Germany - Color France
1 Introduction 12 Physical data
2 Value of data 13 Documents
3 Subsurface data 14 Auditing
4 Current practice 15 Quality
5 DMBoK 16 Other elements
6 Governance 17 Assessing
7 Architecture 18 Glossary
8 Development 19 Figures
9 Operations 20 Bibliography
10 Security 21 Index
11 Corporate data 22 Further info
Upcoming events New articles Extra material Links
Sample chapter Figures Bibliography Extra material Historical Papers

by Steve Hawtin
20 Apr 2014

Decisions and decision trees

The processes involved with running an effective oil company are intricate and complex. Perhaps rather than attempt to understand the whole Byzantine mess of activities required to efficiently discover and exploit sub-surface resources we could do better by focusing on a single business decision and tracking all the consequent costs and benefits? In the past I've called this type of approach "decision based". Often this would take the form of a "decision tree" (which usually works forward from the decision to the consequences).

Decision Tree

I don't know how familiar most people are with decision trees, so I'll include an example to make the later discussion easier to follow (I apologise to those who already know this stuff). Suppose we are thinking of drilling a new well, at the moment our estimate is that there is a 10% chance of generating $20M in production, but should we do a seismic survey costing $1M? The answer depends on how doing the survey will affect our assessment. Suppose that there are three possible outcomes, a good prognosis increases the chance of a strike to 30%, a bad one makes it 2% and otherwise it stays the same. If we assume the chance of a good prognosis is 20% and of a bad one 50% then working through the tree shown here shows that we're better off saving our survey costs and drilling without one.

The reason for doing this type of analysis is to uncover the implications that are inherent in our current understanding. This has two obvious limitations, first the "unknown unknowns" will almost always conspire to render analysis moot. The second issue was covered in a previous discussion on the TV game show "Deal or No Deal", which explained why simple arithmetic does not work with probability, having a 1% chance to win $100,000 is not the same as actually having $1,000 in your pocket. If your goal is to demonstrate, for example, the impact that a new measurement will have on the risk profile of a particular business decision and therefore justify the cost of gathering and analysing data then this technique fits.

If you're trying to work out the total consequence that emerges from a sequence of interrelated decisions and assumptions this is exactly the tool for you. In reality things never happen like that, each new data item has the capacity to reveal additional opportunities that were unimagined at the start. For the "value of information" process these are usually positive benefits, so the results provide a good "base case" and hence can justify a particular action. But at a strategic level we often want not just to limit our exposure to the bad times but also to have a good idea of the overall potential. If, instead of myopically focusing on a limited domain you want to appreciate the influence that this decision will have on the bigger picture, then this approach will deliver results whose applicability is limited.

prev icon
Article 40
paper icon
rss icon
RSS Feed
news icon
home icon
toc icon
Book Contents
figure icon
All Figures
biblio icon
download icon
website icon
buy icon
contact icon
Contact Us
next icon
Article 42

Comment on the contents of the 'Decisions and decision trees' page
Subject: Email to Reply To (optional):