When Is It Worth It?

We’ve done the machinery. Lesson 1 gave us the one honest question — would this answer change my action? Lesson 2 built the ceiling: EVPI, what a perfect answer is worth (160 on our launch call). Lesson 3 priced a real, noisy test: EVSI, netted against what it costs. Now the strategic payoff — the question a sharp decision-maker actually asks before spending a dime or a day on research: when is any of this worth it?

The short answer is a formula you can run in your head, and it has exactly three moving parts. Get a feel for them and you’ll stop researching decisions that were already made — and stop shooting from the hip on the ones that matter.

Recap in one line: value comes only from averted mistakes

Here’s the whole course compressed into a sentence. Information is worth something only when it can change your action — and a changed action is only worth something when the action you’d have taken otherwise was a mistake. No mistake averted, no value. Full stop.

That single idea unpacks into three conditions, because a mistake has to be both possible and expensive to be worth avoiding:

• A mistake is only possible if you’re genuinely unsure (Condition 1) and the decision is close enough that new evidence could tip it (Condition 2).
• A mistake is only expensive if the stakes are big (Condition 3).

Multiply those together and you get the value of the information. Zero out any one of them and the product is zero. Let’s take them one at a time.

Condition 1 — genuine uncertainty

Intuition. If you already know the answer, asking again is theatre. Value of information lives at the fog, not in the clear.

Why. Think of your prior — your belief before looking. If it’s balanced (50/50), a good answer genuinely moves you: half the time it says one thing, half the time the other, and you’re primed to act on either. But if your prior is lopsided — 97% sure — then even a perfect answer confirms what you already believed in almost every world, and you’d take the same action anyway. You paid for a report that only told you what you’d already have done.

This is exactly why EVPI shrinks as the prior gets one-sided. Value of information is largest near maximal uncertainty (the 50/50 knife-edge) and falls away toward either extreme, bottoming out at zero when you’re 100% certain.

Mini-example. On our launch, the prior is P(good) = 40% — genuinely uncertain, so there’s real room for a test to matter. Now imagine the prior were 3%. A survey flashing green would nudge you a little, but almost certainly not enough to overturn a 97%-likely bad market. You’d walk away either way, and the survey earned its fee telling you nothing you’d act on.

Condition 2 — the decision is close

Intuition. Information is a tie-breaker. If the game is already a blowout, the tie-breaker does nothing.

Why. What matters isn’t just uncertainty about the state of the world — it’s whether that uncertainty sits near the line that separates your actions. If two options are near-tied in expected value, a small shove flips you, and evidence is precious. If one option dominates by a mile, no realistic answer moves you across the line, so the information is worth ≈ 0 even if you’re genuinely unsure about details that don’t change the verdict.

Worked example — close call vs. blowout. Our launch is a close call by design. Run the priors:

In the close call, Launch and Walk are only 20 apart. A survey that shifts your belief a little can genuinely flip you from Walk to Launch — which is why the perfect answer here is worth EVPI = 160 and even a noisy 80%-accurate survey is worth EVSI = 92. In the blowout, Launch beats Walk by 330. What survey result could possibly drag +330 below zero? None you’d believe. So on the blowout, VOI ≈ 0 — you already know what to do; go do it. Same product, same uncertainty about demand, wildly different value of looking, purely because one decision is close and the other isn’t.

Condition 3 — high stakes

Intuition. The same nudge is worth more when there’s more on the line. Flipping a coin for a dollar isn’t worth a phone call; flipping it for your house is.

Why. Value of information scales with the size of the payoff swing around the decision. Two decisions can be equally close and equally uncertain, but if one moves $300,000 and the other moves $30, the first is worth researching and the second isn’t. Roughly, double the stakes and you double the VOI — the probability of flipping is unchanged, but each flip is now worth twice as much.

Mini-example. Keep our launch exactly as close and uncertain as it is, but shrink every payoff by a factor of 1,000: Launch+Good = +$400, Launch+Bad = −$300, in dollars. The decision is still a genuine 40/60 near-tie — but now EVPI is 16 cents. Nobody sane commissions a market survey to protect 16 cents. Just pick one and move on; the cost of thinking exceeds the cost of being wrong.

The kill-switch: value is a product, so any zero wins

Put the three together and you get a back-of-envelope formula for the value of information:

VOI ≈ (how likely the answer is to flip your action) × (how much the flip is worth).

The first factor is uncertainty and closeness rolled together — you only flip if you’re unsure and the decision sits near the line. The second factor is the stakes. It’s a product, which is the whole point: zero out any single factor and the entire value collapses to zero. You don’t need all three to be weak — one is enough.

That gives you three fast “value is zero” tells, plus a fourth we met back in Lesson 1:

• Lopsided prior — you’re already 95%+ sure. (Kills Condition 1.)
• Dominant action — one option wins by a mile. (Kills Condition 2.)
• Tiny stakes — being wrong barely costs anything. (Kills Condition 3.)
• Unactionable answer — you couldn’t or wouldn’t change course even if you knew. (The change-my-action test from Lesson 1 — if nothing you learn can move you, the “flip probability” is zero by construction.)

Hit any one of these and stop researching. You’re about to buy a report you already know the ending of.

See the value collapse for yourself

The lab below is the same VoiLab from last lesson, but now use it as a kill-switch demonstrator. It starts at a maximally uncertain, close call — a 50/50 prior with symmetric ±300 payoffs — where the test is worth the most you’ll ever see. Watch the three bars (act now on the prior; buy the test, with its info-gain tinted and its price shaded off; and the perfect-info ceiling), the decision line, the live verdict, and the EVPI / EVSI / Net tiles.

Then break a condition and watch the value crater: drag the prior toward either extreme (both signals start pointing the same way, so no flip is possible), or shrink the gain and loss (Condition 3 evaporates). Either move sends EVSI and the Net crashing toward zero.

The deeper reason: information is a real option

This course sits on top of asymmetry and optionality, and now we can cash that in. Gathering information before you commit is buying a real option — and that’s not a metaphor, it’s the exact same payoff shape.

An option is the right, but not the obligation, to do something later. When you run a pilot, order a sample, check references, or read the survey before you commit the big money, here’s the trade you’re making: you pay a small, known cost (the test) up front, and in return you get to keep the good outcomes and drop the bad ones. If the news is encouraging, you launch and keep the upside. If it’s grim, you walk away and dodge the downside. You’ve clipped off the bad tail while holding onto the good one.

That asymmetry — capped, known cost; open-ended benefit from avoided disasters — is a convex payoff, the signature of optionality from the prerequisite course. “Look before you leap” isn’t folk wisdom; it’s literally the instruction to buy an option before exercising it. The test premium is what you pay for the right to change your mind with better information.

The flip side — and it’s the one people forget. An option only has value if you can still act on what you learn. Buy the survey after you’ve already signed the lease, hired the team, and printed the boxes, and the “right to change course” is worthless — you’ve turned the option into a receipt. Information bought after you commit isn’t a call option; it’s just an expensive way to find out how the story ends. Buy the option before you lock in, or don’t buy it at all.

Practical close: the four-question triage

Before you spend money, time, or a meeting on “should we research this?”, run four questions. Each one guards a condition; a single no means put the calculator away and just decide.

1. Am I genuinely unsure? (Condition 1 — is the prior balanced enough that an answer could move me?) If you’re already 95% sure, no.
2. Is the decision close? (Condition 2 — are the top options near-tied, so a nudge flips me?) If one option dominates, no.
3. Are the stakes big enough to matter? (Condition 3 — is being wrong actually expensive?) If it’s pocket change, no.
4. Could I actually act on the answer? (The Lesson-1 change-my-action test, in option form — do I still have the right to change course?) If I’m already committed or can’t move, no.

Only a yes to all four earns the cost of looking. It’s a fast filter that kills most “we should study this” reflexes before they eat a week — and it points your scarce research budget at the handful of decisions that are uncertain, close, consequential, and still open.

Where this leaves us

You now have the strategic core of the whole course: value of information is a product of uncertainty, closeness, and stakes, gated by whether you can still act — and it’s best understood as buying an option before you leap. Run the four-question triage and you’ll spend on research only when it can actually change what you do.

But the model has sharp edges, and the final teaching lesson is about the places it quietly lies to you. Next up: the traps — the cost of delay (while you gather data, the world moves and the opportunity decays), analysis paralysis (researching becomes a way to avoid deciding), and the accurate test of the wrong question (a survey that’s 99% reliable about something that was never going to flip your action). Great tools, used carelessly, fail in style. Let’s go learn how.