Episode 45 — Decision Trees and Expected Value, Verbally
In Episode Forty-Five, “Decision Trees and Expected Value, Verbally,” we explore a way to reason through branching choices without heavy math or specialized tools. Picture a story that forks at key moments, with outcomes that depend on what you choose and what the world delivers. A simple narrative can carry everything a diagram would show, yet it stays accessible to any stakeholder. This approach lets you weigh alternatives under uncertainty, explain the moving parts, and land on a recommendation you can defend. The point is not to prove certainty, but to surface structure so judgment becomes clearer, calmer, and faster.
Begin by framing the decision with strong edges. Name the choice you control, the main alternatives available, and the external events that can change the result. For a vendor selection, the choice might be incumbent, challenger, or deferral. The chance events could include delivery reliability, quality performance, and regulatory approval. Keep the frame tight enough to analyze in a conversation, yet rich enough to capture what truly drives consequences. This first step draws the boundary of the decision tree before any branches appear, and it stops the discussion from drifting into unrelated risks.
Now describe the paths with plain language, like a script you could read aloud. Start with one alternative and sketch what happens if the favorable event occurs, then what happens if the unfavorable event occurs. Continue until each path reaches a clear consequence, such as a delivery date, a cost envelope, or a go or no go milestone. Use short, specific phrases to mark the forks, such as supplier meets yield or supplier misses yield. If a path loops back through rework, say so and name what would trigger that detour. Your story is the tree.
Assign rough probabilities thoughtfully, and do it transparently. Ask what fraction of similar efforts succeed at each fork, or what a well informed outsider would assume if they had your evidence. Avoid treating every branch as fifty fifty; that is a default for ignorance, not insight. Encourage a short debate about why a probability should move up or down, and point to drivers that would shift it further. If people resist numbers, use verbal anchors like unlikely, possible, and likely, then translate them to an agreed band. The target is coherence more than precision.
Estimate payoffs as ranges rather than single paychecks. Costs, benefits, and time savings all wander within plausible bounds. Give each end state a credible low, a central view, and a credible high, tied to the conditions that make those values real. If a favorable branch delivers savings, say the savings are modest if learning is slower, average if learning is normal, and strong if reuse lands early. Treat penalties the same way. By naming ranges for each path, you give yourself a wider lens that still keeps detail under control.
Compute expected value conceptually instead of reaching for a calculator. The idea is simple. Multiply the value of each path by the chance of that path, then add the pieces to see the average result you would expect over many trials. In conversation you can approximate by mental math. If a good outcome with high value is likely, that weight pulls the average upward. If a bad outcome with high cost is possible but rare, it still pulls downward. The numbers become a compass rather than a cage, pointing to the stronger direction.
Include both one time effects and recurring impacts in the payoffs. A signing bonus is a one time benefit; a lower run rate is a recurring benefit. A license true up can be a one time cost; a maintenance fee is recurring. Say which is which, then estimate the period over which recurring impacts will matter. This keeps a large, short term gain from overshadowing a steady, long term drain, or the reverse. When stakeholders hear the cadence of costs and benefits, they can judge alignment with budgets, investment horizons, and strategic patience.
Run quick sensitivity checks by nudging probabilities and payoffs and watching the direction of change. Ask what happens if the likelihood of on time delivery slips by one notch, or if the cost saving narrows by a small percentage. If your recommendation flips under tiny nudges, you are living on a knife edge and should seek more information. If it holds under reasonable movement, you have robustness. Speak these checks aloud. They demonstrate humility about the inputs and confidence in the structure, both of which build credibility in the room.
Do not ignore non monetary effects like time, reputation, and strategic option value. Time saved could accelerate revenue or reduce exposure elsewhere. Reputation gained with a reliable partner can reduce future friction. Option value appears when a choice opens new paths tomorrow, such as a modular design that unlocks faster releases. Translate these effects into the same narrative form as money. If conversion to dollars feels forced, keep them as explicit notes with weight in the final call. A clear note beats a false number every single time.
Acknowledge dependencies and correlations that link branches across options. Two suppliers might depend on the same raw material, or two internal teams might draw on the same scarce expertise. If those drivers move together, the tree’s branches are not independent. Say so, and reflect that coupling in your verbal probabilities and ranges. This prevents the misleading comfort that comes from assuming diversification you do not actually have. When dependencies are strong, the expected value of a portfolio of choices may be lower than the sum of its parts.
Summarize the recommendation in a few sturdy sentences that echo the tree. Start with the option you favor and the average outcome it delivers. State the key drivers that make it win and the one or two risks that deserve ongoing attention. Name the runner up and why it fell short, using the same language of probabilities and ranges. Finally, point to the small set of actions that would strengthen the choice, such as a pilot to tighten a probability or a clause to cap a payoff tail. The logic should fit on a single breath.
Close with a verbal pre mortem to reinforce readiness. Imagine six months have passed and the decision underperformed. Ask which branch went the wrong way and which assumption was off. Then name the early signals that would have warned you and the contingency that you will keep ready to respond. This practice transforms the tree from a static artifact into a living plan. It shows that your decision is not a bet you hope to win, but a managed exposure you intend to steer with eyes open and hands steady.
Simple trees clarify choices because they turn uncertainty into shareable structure. You frame the decision, tell the branching story, assign honest likelihoods, and tie outcomes to ranges, cadence, and consequences. Expected value gives you a center; variance and dependencies reveal the shape; sensitivity checks test resilience. With that, a recommendation emerges that respects facts and acknowledges limits. The team leaves aligned on what must go right, what might go wrong, and how you will respond. That is the purpose of the method and the measure of its value.
