Key Concern - Project Outcomes May Cluster Below P50

Introduction

In Petroleum Engineering Subsurface Development, it is common to evaluate a project with a range of uncertainty. Typical projects may be planning an infill oil producer, planning water injectors, evaluating a waterflood conversion, or designing a full field development.

Rather than deterministic Low-Mid-High ranges, often probabilistic evaluation using Monte Carlo based methods or more complex and time-consuming full 3D static/dynamic modeling methods (Integrated Reservoir Modeling – IRM) is deployed.

After extensive modeling and analysis, a P50 outcome is derived plus a P90-P10 outcome range. The project is then economically evaluated against the P50 and P90 outcomes and sanctioned if deemed economically robust against these proposed outcomes.

Thinking of a portfolio of sanctioned projects, it is expected that (a) about 80% of the outcomes fall inside the range, (b) most projects cluster around the P50, with (c) few projects approaching both the P90 and P10 outcomes.

Figure: Project outcomes vs. P90, P50 and P10 expectations.

Reality, however, may be quite different. It can be observed that (a) projects tend to cluster somewhat below the P50 outcome, with (b) a further skew towards the lower quartile, and (c) only a very small number of projects being significantly better than the P50. Finally, of the projects that fall outside the range, the vast majority are below P90. Extreme outcomes are possible on either side.

This has puzzled leaders for some time, because staff have generally been doing great technical work and been following the technical standards. Usually, the main reasons for this are that either (a) hydrocarbon volumes in place (HIIP) are below expectation, (b) recoverable volumes are much lower than expected or (c) recovery of the recoverable volumes requires many more wells and time due to much greater complexity. While the industry got better at predicting static volumes (HIIP), total project performance may remain below P50. It should be noted that in reservoir development circumstances where many wells can be drilled cheaply and quickly, project outcomes tend to approach a ‘more expected world’ result.

Why is that?

In my opinion, the main reasons for this are:

Reservoirs are much more complex than given credit for. A proper understanding of the subsurface may not be developed due to time pressure, inefficient integration processes and early anchoring on a “best technical estimate” case (BTE).

Solutions: build a better/deeper understanding of the reservoirs including considering alternative scenarios and screen them vs. project parameters.
Once entering more complex modeling stages, deploy decision-based IRM methods.

Success bias: organizations get paid to ‘make projects go’, and under time pressure often deprioritize looking at alternate outcomes in sufficient detail, may rationalize negative information or do not quantify a viable backup plan.

Solution: build thorough subsurface understanding for entire team, incl. use of analogs and benchmarking.
Clearly define, document and communicate risks through better assurance questioning and specific documents focusing on such aspects.
Develop and ‘stress test’ a backup plan for the ‘HH’ risk scenario (…which is not the Low Case!).

Project studies often take too long, because teams focus on complex modeling too early and are then not prepared to significantly change tack when problems arise and may rationalize or dismiss negative information.

Solution: use truly integrated workshops to build a team-wide deep understanding of the reservoir quickly, use analytical or simple modeling tools to evaluate and screen models and scenarios.
Go to detailed static/dynamic modeling (IRM) only when necessary and deploy decision-based IRM and multi-scenario modeling.
Note: this amounts to a change in the way teams work in subsurface!

Additional Thoughts on “Anchoring too Quickly on BTE”

In my opinion, the reason is that teams tend to anchor quickly on a “most likely” subsurface scenario, then build it out in detail using sophisticated and detailed static and dynamic modeling tools (“Integrated Reservoir Modeling – IRM”). Then, some key parameters are varied to obtain a probabilistic range. For example, one may use a more optimistic/pessimistic top reservoir surface (perhaps even out of a probabilistic range), shift the porosity-permeability distributions up or down some and use some stochastic modeling for distribution. The result is a probabilistic outcome that revolves around the BTE (“best technical estimate”) which has effectively become the P50.

This procedure represents evaluating the ‘random error’ of the data. Even deploying the more sophisticated and complex Experimental Design Approach is not solving this problem, it only evaluates the ‘random error’ better, by using a smarter way to sample the random uncertainty error space.

If, however, the reservoir were to be significantly different from the BTE that was anchored on earlier, e.g., if a reservoir is more channelized instead of a more lobate deposit, or is more proximally deposited than expected, a strong well bias exists or if effective permeability is much lower than rock permeability, then shifting key parameters will not capture any possible alternative scenario. Such alternative and different scenarios would represent a ‘systematic error’, and such alternative scenarios would have to be built separately and evaluated separately. Building separate models for alternative scenarios is very time consuming, for the seismic interpreter, the static geo-modeler and certainly for the reservoir simulation engineer and, therefore, are often not done in the interest of predetermined tight timelines.

Two elements to significantly address this dilemma are proposed

First, develop a thorough and multi-disciplinary understanding of the reservoir. Using advanced data analysis, analytical simple modeling tools and ‘thought sharing’ in a truly integrated workshop setting will allow revisiting all data and previous interpretations, integrate all different disciplines (data & people) effectively, allow evaluation and testing for alternative scenarios and their screening for possible impact. One of the biggest wins of such workshops often is that the entire team walks away with a shared understanding and view of the reservoir and path forward.

Second, effective multi-scenario thinking (and modeling). In multi-scenario modeling, several alternative scenarios are evaluated in a very comprehensive way, using analytical and simple modeling methods to understand their impact and high-level ranges, before embarking on comprehensive 3D static/dynamic modeling path. Once 3D static/dynamic modeling is starting, better information exists to quantify the relevant subsurface scenarios, tailor the modeling process to the decisions at hand, identify risks and develop and stress test a contingency plan for a realized high-risk outcome.

It should be noted that an outcome much below the P50 forecast may not be a pessimistic outcome, but a true P50 outcome of an alternative scenario with its independent distribution curve.