Sequential Decision Analytics Warren B. Powell · Princeton University

How to teach decision framing

This webpage contains some ideas on how to teach decision framing, with a focus on bringing these ideas into a classroom. Our primary focus is teaching students how to break down problems by answering the three decision framing questions:

  1. What are the performance metrics?
  2. What types of decisions are being made (which might include discussing who makes the decisions)?
  3. What are the uncertainties that affect performance?

Note that these questions have to be addressed from the perspective of a decision-maker, which can be an individual, a small group, or a computer system. Identifying who (or what) makes the decision is a step we call scoping.

This webpage offers three ideas to help with teaching (more to come):

Expanding the scope of an existing lecture

Imagine you have a lecture prepared to address a specific problem, whether it is a business problem or any application in engineering, health, energy, or transportation. Your lecture may be designed to illustrate a specific tool or solution approach, which means you have already chosen what decision is being made, and the performance metric.

  1. Metrics: Take a few minutes to step back to a larger problem setting where you want to minimize costs, maximize profits, or increase productivity. As you increase the scope of the problem, you typically are going to encounter multiple metrics. Organize these into a pyramid as we do here.

  2. Decisions: Next switch to a discussion of what steps might be taken to improve one or more of your metrics. These may be discrete actions (change the product design, revise the manufacturing process) or continuous parameters. List anything that may have an impact on any of your metrics. Remember that there are different types of decisions (illustrated here). There are even many different words for “decision” as illustrated here.

  3. Uncertainties: Finally, summarize the uncertainties that might change the impact of each decision on the performance metrics. Again, there are many types of uncertainty (as illustrated here), but you want to identify the uncertainties that have the greatest impact on your performance metrics.

It is possible to manage this discussion so that you summarize the core elements in as little as 5 minutes, or as much as 30 minutes, depending on the level of student involvement. What is important is the constant repetition of thinking about metrics, decisions and uncertainties. Be sure to always summarize these in distinct lists. I then recommend creating an interaction matrix such as those given in the framing interaction matrix.

The goal is to repeat this exercise as often as possible, in a wide range of settings, so that thinking about metrics, decisions, and uncertainties becomes second nature.

Framing a talk

Now imagine that you have invited someone from industry to talk about a problem. This speaker will not know anything about our framing process, and will discuss a problem setting in a general way using the vocabulary of their community.

Give each student a sheet with three headings (metrics, decisions, uncertainties) such as the sheet provided here.

As the speaker progresses through their presentation, the students have to identify different metrics that seem relevant for evaluating performance, what decisions have an impact on these metrics, and the uncertainties that affect performance.

It is possible that the speaker may overlook something, such as decisions. Students should realize that the talk is incomplete, and ask the speaker to fill in missing elements. They have to learn how to do this in a way that is understandable to the speaker. For example, asking “What decisions do you make” may not be obvious to someone not trained in this vocabulary.

After completing the worksheet, the next step would be to follow the framing process.

Case studies on framing

This section contains a number of “case studies” based on the problem settings in Chapter 2 of Framing the Problem. At the end of each case is a series of the usual questions designed to test the reader’s absorption of the facts of the case. Most of the cases are based on the applications in chapter 2 of the monograph “Framing the Problem.”

Then, there is a series of questions which are the same for each case, which are designed to teach students how to analyze the case by identifying performance metrics, types of decisions and sources of uncertainty. These “decision framing questions” are reproduced in the first section below. You should be able to append these to any case, under the assumption that every case has the ultimate goal of improving some process or product.

Feel free to download any of the cases for use in classes or simply to obtain a general understanding of the framing process.

Decision framing questions

Each of the cases below contains the same set of decision framing questions, so students can practice the framing process by identifying the decision maker’s scope, performance metrics, types of decisions, and sources of uncertainty. Use the decision framing tool to help answer them.

  1. Identify and describe the responsibilities of the decision maker. This should set the scope of the problem.

  2. Use the metrics pyramid tool to identify and prioritize the different performance metrics. Click on the oval holding the metric to identify whether it is a metric to be maximized (green), minimized (red), a target that is a floor (light green) or a ceiling (light red), or finally a target (purple).

  3. Use the decision prioritization tool to first list the different types of decisions (use the types of decisions as a guide). These will then populate the rows of the matrix to the right, where the columns are the performance metrics sorted left to right by the priority in the pyramid above. Click on each cell to indicate if the impact of the decision on each metric is high (red), medium (orange), low (yellow) or none (white).

  4. You can grab the handle (three bars) farthest to the left on each row to move it up or down. Sort the rows so that the decisions with the greatest impact on the most important metrics are toward the top. The cells in the upper left-hand corner should have the greatest concentration of red and orange cells. At this point identify the decisions that should receive the most attention in your study. If you do not have any decisions with a medium or high impact on the most important metrics, revisit your list of decisions.

  5. Repeat the process to list and prioritize the uncertainties using the uncertainty prioritization tool. Use the categories of uncertainty to help you identify these.

  6. It is important to differentiate uncertainties that can be evaluated on average, or as risk events:
    • a) Which sources of uncertainty would be captured using average performance over time?
    • b) Which sources of uncertainty represent forms of risk that are not properly captured using averages?
  7. Choose the single decision that seems to have the highest impact on the most important metric.
    • a) What approach do you think you would use to make this decision?
    • b) What data appears to be necessary to make this decision, including the calculation of any performance metrics where the decision would have a high or medium impact.
  8. Using the same decision you identified in (7), identify the people, departments, groups or organizations that you would need to work with to implement the decision.

List of decision framing cases

Asterion Therapeutics — The Enrollment Clock. A biotech’s chief development officer must decide whether to accelerate, restructure, or pause a Phase II trial of a promising therapy as enrollment lags, safety signals emerge, and the patent clock ticks.

Aurora Motors — The Flow Coordination Dilemma. An automaker must synchronize uncertain global parts supply, production, and customer deliveries when tariffs, federal incentives, gasoline prices, and buyer preferences change faster than its plants can respond.

Bellwether Department of Health — The 36,000-Kit Decision. A state public-health office receives fewer naloxone kits than its partners requested and must allocate the limited overdose-reversal supply, decide whom to train, and whether to reserve kits for future waves.

Blue River System Operator — The 6:00 p.m. Reliability Decision. A regional grid operator faces a record evening peak, thunderstorms moving toward the solar corridor, and a warning from a critical generator, and must choose how to cover the reliability gap in real time.

BrightNest Home Services — The Two-Sided Launch Dilemma. A home-cleaning startup with strong household demand but thin cleaner supply must decide how to balance geographic expansion, pricing, and platform investment across three pilot markets.

Crestline Biotherapies — The Week 28 Decision. A biotech must decide whether to expand, redesign, pause, or terminate a Phase II clinical trial when time, patients, capital, and remaining patent life are all limited.

Dispatch Dynamics — The Recommendation Gap. A decision-automation company finds that only 43% of its truckload dispatch recommendations are acted on across twelve customer implementations, and must decide how to close the last-mile-of-implementation gap.

Harbor County — Measles Response. A county health officer must decide how aggressively to respond when measles risk is concentrated, vaccination confidence is falling, and the costs of prevention are immediate — but the benefits are largely invisible.

Harborview Hotel — The Eight-Week Pricing Decision. A hotel general manager must set pricing and inventory strategy for a citywide technology conference when peak nights are nearly booked but surrounding weekends remain soft.

Hearthline Home — The Demand Balancing Problem. A regional furniture retailer must decide how aggressively to stimulate holiday-season demand without creating stockouts in popular styles or deep markdowns on products customers no longer want.

Lakeside Medical Group — The Next Ninety Days. A primary-care team must redesign treatment for a Type 2 diabetes patient when clinical response, daily behavior, technology, and affordability are difficult to disentangle.

The Meridian Campaign — The Ninety-Three-Vote Map. A presidential campaign manager must allocate money, candidate time, information, and field capacity across a changing electoral landscape with eleven weeks to Election Day.

Northbridge Industrial Systems — The Ninety-Day Payment Decision. A manufacturer with enough cash for its own near-term obligations but not enough to absorb tariff-driven inventory purchases must decide how to protect liquidity without destabilizing critical suppliers.

Northstar Living — The Inventory Bet. A home-goods retailer must commit to holiday-season purchases of its fastest-growing products before it knows how strong demand will be or whether ocean shipments will arrive on time.

PrairieLine Freight — The Friday Afternoon Dispatch. A truckload fleet VP must balance freight selection, driver assignment, home time, and network balance when 1,184 tractors are in the wrong places heading into the weekend.

Redwood Department of Recovery Services — Redwood’s Funding Needle. A state addiction agency commissioner must balance naloxone distribution, treatment, and recovery services against an $18.6 million funding gap that threatens 126 positions and a heavily restricted revenue mix.

Ridgeway Industrial Systems — The Energy Crossroads. A manufacturing network COO must respond to a utility’s request to study its service connection as rising electricity costs and data-center-driven infrastructure investment reshape the region’s power system.

Riverton County — Public Health Response. A county health commissioner must decide how to respond when local overdose deaths remain high even as the national crisis appears to be easing.

Summit Ridge Asset Management — The Monday Morning Cash Decision. A mutual fund CIO must decide how to manage investor flows, liquidity, portfolio performance, and trading pressure after a weekend of bad news.