Guest Post: Connecting Decisions to Data: A Case Study, Part 5: Simulating Actions to Outcomes: The Opportunity Envelope

In my Connecting Decisions to Data post I introduced a Decision Intelligence case study where a manager needs to make three decisions about a new product’s first production run. Then I identified three decision levers the manager can use to achieve their outcome: maximize profit. 

  1. Sale price: How much should I charge for each unit of the product? 
  1. Manufacturing run size: How many units should I order for my first run? 
  1. Marketing spend: how big an investment should I make in marketing the product? 

In my last Decision Intelligence and Business Intelligence post, I showed you the “Product Sales and Manufacturing Planning” dashboard for this case study. It was designed to provide sufficient “insights from data” to solidly justify these decisions. I pointed out that, in practice, it would be difficult for a decision maker to derive actions-to-outcomes insights directly from this dashboard. This is worth reiterating: our goal is business outcomes and the best actions to achieve them, yet typical dashboards don’t tell us this.

In this post I will begin to explore how simulation can reorganize the dashboard data for decision makers. 

The root cause behind the difficulties we encountered in applying the dashboard to our decision is that, even though data is available to supply the information that is missing in each of the model’s Unknowns, no data set exists that directly correlates profit to the range of values we are considering for the three decision levers. Without such a data set, none of the business intelligence/data science tools—regression, trend analysis, charting, drill down, deep learning, etc.—can address the decision maker’s fundamental question:

“If I take this course of action, what will the outcome be?” 

The solution to this problem (identified as Problem 2 in several prior posts) lies in another benefit of using the Decision Model. Once the data for the Unknowns has been supplied (as sets of tables, machine learning models, regression analyses, or even graphs sketched by a subject matter expert) the Decision Model can determine the outcomes corresponding to any particular choice of Levers and Externals. Using a simulator, a very large number of such choices and the outcomes they produce can be generated. That is, the information contained in the Decision Model, when run through a simulator, can be represented in exactly the form the decision maker needs: as a relationship between the actions the decision maker can take, and the outcomes corresponding to each of these. 

The Solution to Problem 2 is, therefore, to use the decision model to simulate a large sample of Lever combinations and calculate the outcomes produced by each. (*Note there are smart ways to do this kind of simulation that don’t take a huge amount of time)

This is the systematic method for transforming the data contained in the Decision Model into the form decision makers need, that is, a representation of “decision outcomes” vs. “decision actions” 

Running the Decision Model for our product launch manufacturing and sales price decision scenario produces the raw data shown in the graph at the top of this post. As we shall see shortly, the data behind this single graph contains a wealth of information for the decision maker that is completely aligned with the questions the decision maker needs answered. 

The Y axis of this graph is the Outcome—in this case, Profit (or loss if the value is negative). Each point on the X axis represents a unique combination of the Levers. These are, specifically, Marketing Spend as a percentage of profit, unit retail sale price, and size of production run. These are indicated individually in the three sub-axes below the main X axis. The simulator selects many thousands of points along the X axis and calculates the corresponding profit and the results of this calculation are represented by the blue line on the chart. (Note that the externals can be varied also, such as the market size or demand curve, if there is uncertainty about the true values of these factors.) 

Opportunity vs. Risk 

The graph above provides the decision maker with the value of the profit associated with each chosen combination of the three decision Levers. But it also provides them with much more if the right tools are applied. The red and green “envelope” curves show the bounds on profit and loss across the range of values the decision maker may choose. These are called the “Opportunity Envelope” and “Risk Envelope”, respectively. The highest peak of the profit curve corresponds to the decision that will yield the highest profit, and the lowest point on the loss curve shows the maximum loss that may be incurred. These are shown in the table next to the graph as the “Maximum Profit Decision”. In our product launch scenario, in the absence of any other consideration, this decision would be your recommendation to the executive team. 

But there are further considerations, most notably, risk. As well as determining the maximum upside outcome available, and identifying the actions that lead to it, a decision maker must also consider the potential negative outcomes which may result from their chosen actions, both in terms of the negative outcomes’ severity and their likelihood. As we shall see in the next post, this is something about which the simulation graph also provides very useful information. 

Mark Zangari
Co-Founder and CEO at Quantellia, LLC | Website

In addition to his duties as CEO, Zangari leads Quantellia LLC's Scalable Solutions division, where he is responsible for financial technology, telecom, and Covid-19 solutions. Zangari is also the architect of the company's World Modeler solution suite.  Before joining Quantellia, Zangari spent 15 years as CTO for a spatial GIS company, where he specialized in providing solutions to utilities and telecoms.

A physicist by training, Zangari's papers on cosmology are still referenced today.

You may also like...