Like what we have to say? SIGN-UP for our NEW FEATURE POSTS & never miss any of our tips, tricks, trends & best practices!


Date Posted: 4/27/16
By Kevin Carpenter, contact:
Scenario planning has historically been an exercise in predicting possible futures and then drafting plans to react to the most likely future, with contingent plans for the other possible outcomes. 
The success of this approach has been somewhat overestimated by the human nature of our selective memory.  In effect, once the event happens we erroneously remember this outcome as the outcome we predicted.
This happened to me with a client once when we had optimized their chemical R&D portfolio to limit their focus to a subset of catalysts, feedstocks, and processes to make a new plastic.  Our optimization reduced their 5-year R&D budget from $80MM to $20MM while only decreasing their chance of success from 98% to 96%.  After the six-week study and analysis, the manager told me - "oh yeah, we were going change the program like that anyway".  This Hindsight Bias as described by Baruch Fischhoff has been studied and documented since the 1970s.
This same bias can feed upon itself to the point that someone believes they are an "expert at making predictions" and that their forecasted scenarios can believe to be the way things are going to be with certainty. 
I believe a better method can avoid this bias danger and lead to not only creative future scenarios, but scenarios that can be predicted with some measure of probability with the associated signposts indicating to which possible future we are headed.
If we begin with the Objective to be assessed, we can map the large uncertainty drivers to the Objective.  However, rather than stopping here and constructing scenarios from this map (the traditional approach), we take each of the key uncertainties and break it into its smaller parts. This disaggregation allows us to examine uncertainties at the appropriate level of granularity to enable experts 
to forecast their ranges without thought as to how it impacts the high-level Objective. This removes not only the bias of the expert seeking a preferred outcome, it removes ownership of the final scenarios so that hindsight bias becomes impossible.
Once we have the uncertainties modeled for interrelationships and impact to the Objective, the model itself can tell us the degree to which the Objective varies (the scenarios), the probability of these outcomes, and the state of the uncertainties driving each of the scenarios.  Furthermore, because we know the state of the uncertainties, we can construct a signpost picture to watch the early 
states of the uncertainties predict which future we're pointed to long before it gets here.
Let's walk through a quick example. 
Say we want to model different scenarios of traffic demand in a large city 10 years out.  The number of cars on the road in 10 years becomes our graphObjective.  High-level factors might include the number of people moving to the region, the location of businesses being developed, the changes in the workforce, and the degree of local amenities.  In the traditional approach, city planners get in a room and discuss what will happen if lots of people move there and have to commute 20 miles to work and shopping (Scenario A with HIGH traffic demand).  Or growth is slow, unemployment increases, and local shopping pops us (Scenario B with LOW traffic demand).  The planners don't like to commute or know someone who's recently lost their job and unconsciously think Scenario B is more likely.  Or the mayor is a big champion of growth in the region and Scenario A becomes more likely.  The planners have a bias in their assessments and ownership in the scenarios.  If by chance they are right, they're picked again for the next planning cycle because they are so good at predicting the future.
With our alternative approach, the same Objective (number of cars on the road in 10 years) and the same high-level factors are considered.  However, now we take those factors and others and begin to break them down into the uncertainties that drive their values.  The "number of people moving to the region" is dependent on the economy, housing prices, changing commodity prices, etc.  Also impacting this is the availability of local amenities, listed as another high-level factor.  Forecasting these uncertainties fall to experts other than city planners and are supported by documented information as well as expert opinion.  These detailed uncertainties are also far easier to reliably estimate a range of outcomes than big picture events. While the uncertainties are tied to the Objective in the model, the assessments of the uncertainties are removed from the outcome and bias is avoided.  Hindsight bias is eliminated as well as the experts are not assessing a scenario outcome, but rather the outcome of discreet uncertainties.  The model can now not only show us the extents of outcomes, from HIGH demand to LOW demand and several points in between, but it can also assign probabilities to each of these outcomes without the bias of city planners.  We can now also predict whether we are moving to a future of high demand or low demand by monitoring uncertainties and their forecasts like housing price trends, global and local unemployment trends, and the price of gasoline.  We can also control the outcome of which scenario occurs to some extent, but that's a topic for the next blog.



Sign up to receive thought leadership summaries from KCA and get access to the FREE LIVE AUDIO
from the September 16th Houston Energy Breakfast including updates from, WoodMackenzie, Houston Technology Center, Burnett Oil and the Society of Petroleum Engineers (SPE).

We respect your privacy and will never share your information.

Spacer 2101 City West Blvd | Houston, TX 77042


Spacer Call Us: 832-303-3308