Start With the Base Rate

The mistake that feels like careful thinking

When you face a decision or prediction, the natural move is to focus on the details in front of you. You gather information, weigh the particulars, and build a picture of this situation. That feels thorough. It often isn't.

The problem is that your specific case belongs to a larger category of similar cases, and those cases have a track record. Ignoring that track record—or treating it as less important than your fresh analysis—is one of the most reliable ways to be wrong with confidence.

This is what Daniel Kahneman called the inside view: reasoning from the details of the case at hand. The corrective is the outside view: asking first how things usually go for cases like this one.

The outside view starts with the base rate.

What a base rate is

A base rate is simply the frequency with which something happens in a reference class—a group of comparable situations.

If you're wondering whether a new restaurant will still be open in two years, the base rate is the proportion of new restaurants that survive two years. If you're estimating whether a software project will finish on time, the base rate is how often similar projects finish on time. If you're assessing whether a business merger will create value for shareholders, the base rate is how often mergers do.

These numbers are often discouraging. Most new restaurants close. Most software projects run late. Most mergers destroy shareholder value. The people involved in each of those cases believed their situation was different. Sometimes they were right. Usually they weren't.

The base rate doesn't tell you what will happen in your case. It tells you where to start.

How to apply it

The process has three steps.

Step one: Identify the reference class.

Before you can look up a base rate, you need to decide what category your situation belongs to. This is harder than it sounds, because you can define the category narrowly or broadly, and the choice matters.

If you're launching a mobile app, your reference class could be all new apps, new apps in your category, new apps by first-time founders, or new apps with your level of funding. Each gives a different base rate. The right choice is usually the most specific class for which you have reliable data. If you have solid data on first-time founders in your category, use that. If you only have rough data on apps in general, use that and note the uncertainty.

The temptation is to define the class so narrowly that your situation looks unique. Resist it. Uniqueness is usually an illusion, and it's a convenient one—it lets you ignore inconvenient frequencies.

Step two: Find or estimate the rate.

Sometimes base rates are published. Survival rates for businesses, on-time rates for construction projects, success rates for clinical trials—these exist and are findable. Use them.

When published data doesn't exist, you can estimate. Talk to people with experience in the domain. Look at analogous cases. Even a rough estimate—I think this works out maybe a third of the time—is more useful than no anchor at all, because it forces you to confront the outside view before you get lost in specifics.

Step three: Adjust for what's genuinely different about your case.

Once you have a base rate, you can reason about your specific situation—but now you're adjusting from an anchor rather than building from scratch.

Ask: what features of my case push the probability up from the base rate? What features push it down? Be honest about which differences are real and which are flattering stories. We have a great team is something almost everyone believes about their own team. We have a signed contract with a major customer before launch is a concrete differentiator that might actually move the number.

The adjustment step is where your specific knowledge matters. But it comes second, not first.

Why this order matters

If you reason from specifics first and consult the base rate afterward, you'll anchor on your initial impression and use the base rate selectively—accepting it when it confirms your view, dismissing it when it doesn't. This is motivated reasoning dressed up as analysis.

Starting with the base rate forces you to confront the outside view before your inside-view story has fully formed. It's harder to dismiss a 70% failure rate when you encounter it before you've fallen in love with your plan.

This is especially important for making better predictions. Forecasters who consistently outperform others tend to anchor on base rates and adjust modestly. They're suspicious of large adjustments, because large adjustments usually reflect overconfidence in the specifics rather than genuine information.

A worked example

Suppose you're deciding whether to pursue a legal claim. Your lawyer has reviewed the facts and thinks you have a strong case. You feel wronged. The evidence seems clear.

Before you go further, ask: what percentage of cases like this one—same type of claim, same jurisdiction, similar facts—result in a favorable outcome for the plaintiff? If your lawyer doesn't know, that's worth noting. If the answer is 30%, you now have a starting point.

Then adjust. Is your evidence stronger than typical? Does the defendant have resources to outlast you? Is the judge's record relevant? Each factor moves you up or down from 30%. Maybe you end up at 40%. Maybe 25%.

Now you're reasoning well. You haven't ignored the specifics—you've placed them in context. A 40% chance of winning might still be worth pursuing, depending on the stakes and costs. But you're making that decision with your eyes open, not on the basis of a story you told yourself about how clear-cut your case is.

The habit to build

Make what's the base rate? a reflex. Ask it before you analyze, before you plan, before you commit. It takes thirty seconds and it will save you from a category of error that feels, from the inside, like careful thinking.

If you're working on goals that depend on accurate self-assessment, goal-setting that actually sticks covers how outside-view thinking applies to planning specifically. And if you want to track how well your probability estimates hold up over time, Brier scores give you a concrete way to measure it.

The base rate won't always be right about your case. But it's almost always the right place to start.