Decision Loops - Your First Loop

This week, I want to equip you with your first decision-making tool, which I’ve named the First Loop. This method allows you to attack ambiguity, the main challenge to making decisions in a business context, and give you some basic examination tools to explore alternatives.

To build a decision-making toolset that can scale with the decision in front of you, you will go through an iterative approach to making a decision, called Decision Loops. Each loop relies on a structured decision process that emphasizes different parts of the process. They guide you as you gain more information about the decision context and seek more confidence in your examination.

The First Loop is meant to take no more than 30 minutes and focuses on clarifying the decision that needs to be made. After this first iteration, you will either decide that further refinement loops are necessary or you will already have gained enough clarity to go ahead and implement the chosen alternative.

Before diving into the details of the First Loop, here’s a succinct introduction to the structured decision process.

Structured decision process

The goal of a structured decision process is to externalize how intuition is used to make decisions and give you a method to mix intuition and data-driven insights in a way that is more explicit, auditable by stakeholders, and reproducible. The result is a compelling decision analysis that will help you get the solution accepted and implemented more efficiently than if the details were left in someone’s brain.

Admittedly, this is a heavier process than simply picking an alternative that feels right. So, the very first decision you need to make is whether to use a structured approach.

• Are the problem and the decision that need to be made clear to you?
• Are you confident in having a dominant alternative figured out?
• Are you empowered to decide and allocate the resources to implement your chosen alternative?

If the answer to all those questions is “yes”, you may not need the added clarity of a structured decision process. If any of those answers is “no”, it is a good sign that you may benefit from at least the First Loop of the following structured decision process.

The full structured decision process is broken down into five steps and comprises those activities:

1. Elucidate Clarify the problem, identify the decision-maker, identify and prioritize the objectives pursued by this decision, define the measure of value used to assess decision alternatives, and assess risk tolerance.
2. Explore Understand the business context, model the problem, generate alternatives, build the value (regret) function.
3. Examine Assess uncertainties influencing the outcome, analyze the sensitivity of decision factors, identify useful data, create risk profiles for each alternative, estimate the value of getting more information.
4. Enact Assert whether to make a decision now, pick an alternative, allocate required resources, plan and implement the chosen course of action.
5. Evaluate Iterative decision quality assessment, in hindsight review of decisions, adjust decision-making systems within the organization.

This represents the high-level topics addressed during a full structured decision process. Each loop will emphasize different activities and may skip others entirely. Unless you know already that you are facing an amalgamation of ambiguity, complexity, or uncertainty, you should approach your decision iteratively, starting with the First Loop.

First Loop

This is where you set the foundations for the decision to be made. In this loop, you’ll focus on the Elucidate and Examine part of the structured decision process. You have two objectives: (1) clarify the decision, and (2) evaluate whether you need to spend more time on it than the 30 minutes set aside for the First Loop.

Elucidate

What is the problem or opportunity that requires a decision? It may sound trite, but clearly articulating this will help you communicate with others who help you make the decision and prevent you from solving the wrong problem, (which happens a surprisingly often). Write it down. Clarify imprecise terms. Make sure to focus on the problem, not the symptoms. In the first loop, spending a few minutes here is all it takes. However, you may want to deepen your understanding of the problem on subsequent loops.

What measure of value will be used to assess the different alternatives? This step helps you crystallize what a good solution looks like. In a business context, it is usually possible (even if it sometimes requires some assumptions) to focus solely on the financial aspects. In those cases, return-on-investment, net present value, or expected monetary value are good starting points. However, in this first loop, it may be worth considering proxy metrics that are closer to the problem. If you choose so, you will need two measures:

• A measure of the impact, such as revenue, market share, churn, customer satisfaction, defects, downtime, work-life balance, employee attrition, etc.
• A measure of the cost of implementing the alternative, such as capital cost, calendar time, employee-months, compute time, expanded code vulnerability surface, etc.

(The observant reader may have noticed the absence of the notion of regret. While it is the superior metric, for a quick assessment it can bog you down in arithmetic gymnastics. It will be back in subsequent loops.)

Whichever you choose, make sure it is consistently measured for each alternative. With those two measures, you can then use the ratio of impact/costs to compare alternatives.

Who is the decision-maker? Let’s start with a definition:

?Decision-maker

The person (or group) with the power to allocate the required resources towards the chosen alternative.

Important: Without the power to allocate resources, any “decision” is merely a recommendation to the real decision-maker.

Your decision process is much more likely to succeed if the decision-maker is on board from the beginning. They can ensure you have a shared understanding of the problem, communicate their preferences relative to risks, or provide a prioritization over competing objectives. If you are not the decision-maker yourself, then at this point you may want to confirm with them that you are working on the right problem.

Explore

This step is fairly light in the First Loop. You only need to list the alternatives under consideration. Those could be multiple choices or a simple go/no-go decision. Note the defining factors of each alternative, what makes them different from one another. Those factors will be useful when modelling the decision.

Examine

The goal of this step is to measure your uncertainty in the outcomes associated with each alternative. This is likely the step that will feel the most uncomfortable at first, but it is also the most valuable of the structured decision process. People uncomfortable with manipulating probabilities are likely to come up with an estimate of the measure of value for each alternative and jump to the next step. This is a mistake, even for a first iteration, because it removes most of the insight harvestable from your intuition about the risks associated with an alternative. As such, for each alternative, come up with three estimates of the measure of value:

• A base estimate This is the estimate corresponds to your best guess of the resulting measure of value given your personal knowledge. Don’t fret too much about the number’s accuracy, just pick something that feels right.
• An estimate of the upper-bound Think about the best-case scenario. If everything turned out just right, what measure of value could you hope to reach? Note the factors involved in getting everything in perfect harmony to guarantee the best outcome and what that outcome would be.
• An estimate of the lower-bound Imagine that everything went wrong. What’s the worst that could happen? Jot down the elements of that worst-case scenario, along with your estimate of the lower-bound.

The upper and lower-bound estimates are meant to assess the boundaries of the possible outcomes. In statistical terms, it is an approximation of the distribution of probability of the measure of value.

In actuality, the boundaries may not exist or be absurdly high. For example, there’s no real upper-bound to people’s height, but values beyond three metres are clearly absurd. In those cases, aim for the 95th-99th percentile. In other words, what’s a value that you would be very surprised to see exceeded (<5% probability). A rough approximation is sufficient at this stage. A word of warning: people are naturally biased towards underestimating the width of the interval of likely values.

With these three numbers, you can get a sense of the risk of different alternatives, along with which way the risk is skewed. If the base value is closer to the upper-bound, that would indicate a higher likelihood of a good outcome. If it is closer to the lower-bound, that alternative is likely to disappoint.

A note for the more mathematically minded readers: keep an open mind about the above method. It is imprecise on purpose. Having estimates of the credibility of one’s base estimate, however flimsy, still beats not having any uncertainty estimates. In fact, this is the only way to be able to have a meaningful discussion about the risk profile of the different alternatives.

First, you may have a dominant alternative. An alternative dominates another one when all three estimates (low, base, high) are greater than the estimates of that other alternative. If you have an alternative that dominates all the others, it is a dominant (and optimal) alternative. In that case, you have an easy path forward for your decision and are ready to go to the next step.

If you don’t have a dominant alternative, then you’re left with alternatives that offer different risk profiles. For instance, one may offer a chance for a higher value, but also be a risky bet with a worse lower-bound. The question becomes whether that risk is significant, and the attitude of the organization towards risks. Here, there are no right answers, it all depends on whether your organization can endure the possible lower values in order to have a chance at the higher values.

At this point, you need to make an auxiliary decision: should you go for another loop to deepen the decision analysis? If the numbers in front of you are enough for the decision-maker to make a confident decision, further refinement is likely a waste. Don’t fall into the trap of assuming that more analysis is better. Digging into the details would be warranted when:

• you have too much uncertainty in your estimates to be confident in the decision outcomes;
• you have multiple alternatives with different risk profiles that requires further comparisons to highlight the right one;
• you want to explore potential risks mitigation for certain alternatives; or
• the cost of analysis is negligible compared to the impact of a wrong decision.

In the First Loop, there is no need for extensive estimates. Is it worthwhile for you to spend about a day deepening this analysis? In that case, now is a good time to share your findings with the decision-maker and go to the Second Loop.

If you are satisfied with your chosen alternative, proceed to the next section.

Enact

In the case of a First Loop, this step is fairly straightforward.

If you’re not the decision-maker, work with them to share your analysis, the alternatives explored, and your logic behind choosing one of them. Given the superficial level of the analysis, you may field some questions and tweak your model a bit, but together you should either reach a conclusion or decide to deepen the analysis.

Once a decision is made and an agreement on the path forward has been decided, it’s time to plan and execute on the implementation.

Evaluate

A good way for you to develop your decision-making skills is to keep your analysis close by, and schedule some time with yourself in the future to review it once implementation is ongoing or close to done. This will give you the benefit of having observed some consequences of the decision and help identify some areas that you may have glossed over in your analysis. Maybe your estimates were too conservative. Maybe you forgot to take into account another business unit that needed to be kept in the loop. Maybe you ended up needing to course correct midway through.

The trick is to honestly assess what you could have known at the time of going through your First Loop, do a small review of your process, and use what you’ve learned to hone your intuition.

Make sure you’re subscribed so you don’t miss the next article. Next week, we’ll dive deeper into activities mentioned in the structured decision process, but aren’t part of the First Loop.

Have you ever used a structured approach to decision-making? I’d love to hear about your experience in the comments or in my DMs!