When Data Stalls You

Key takeaways

• Data can freeze a decision as easily as it can free one, and it usually freezes when people quietly disagree about something they have not said out loud.
• Daniel Kahneman's inside view, reasoning from the specifics of your own case, feels natural and tends to mislead. The outside view, treating your case as one of a class and asking what usually happens, feels unnatural and tends to help.
• The explanation loop, asking the data for one more cut to explain a wobble, is often a way of putting off a decision rather than improving it.
• Base rates are dull and useful: what normally happens to ventures like yours is stronger evidence than the story of why yours is different.
• The particular paralysis of flipping between frames has a name, zozobra, and its source is often social rather than statistical.

Data can stall a decision as easily as it can free one

We tend to assume that more data moves a decision forward. Often it does the opposite. A team that cannot agree gathers numbers, and the numbers hand everyone fresh material to disagree with. I have sat through launch meetings postponed three times, each time for "more analysis", when the real blocker was that two senior people held different and unspoken views about what the company was for. No amount of conversion data was going to resolve a disagreement that was never about conversion.

The inside view feels natural and tends to mislead

Kahneman drew a distinction worth carrying around. The inside view looks at your problem from inside its own details, your plan, your team, your particular advantages, and forecasts from there. The outside view sets most of that aside, treats your project as one instance of a wider class, and asks what happened to the others. He noted that the outside view is an unnatural way to think, because it asks you to forget what you know about your own case and see yourself as a point in a distribution. People hate doing this and resist it, which is exactly why it carries information the inside view does not.

The outside view asks the boring, useful question

Base rates are the outside view made practical. What share of new product lines at your size break even in year one? How long do subscription tiers like yours usually last? The answers are unexciting and a little deflating, which is why founders hurry past them towards the reasons their own case is special. Sometimes it genuinely is. More often the base rate was trying to say something and the inside story drowned it out. There is a reason people feel calmer about a ten percent risk once they learn the average is higher: a number only becomes meaningful next to its reference class, and without one we are guessing with confidence. This is the narrative fallacy from the Your Brain posts, met out in the field, because the vivid inside story will always beat the dull outside number in a straight fight for attention.

The explanation loop is often avoidance in a lab coat

There is a specific failure I have learned to watch for in myself. A metric wobbles, so we ask why. The first cut raises a question, so we ask for another. Three weeks later we have a forensic account of a blip and no decision. The loop feels like diligence. Underneath, it is frequently a way of not choosing, because as long as there is one more cut to run, the moment of commitment can be deferred. Data can only tell you what has already happened. At some point you have to decide what to do next, and no further slice of the past will take that step for you.

When you cannot settle on one frame, the problem may be social

Spanish has a word for what the rest of this is circling: zozobra, the unease of being unable to settle into a single point of view, of wobbling between frameworks and never relaxing into one. In a business it shows up as a team that re-litigates the same launch from a new angle every week. The tell is that the analysis never converges, because the thing stopping convergence is not in the data. It is an unspoken disagreement about goals, or trust, or who carries the blame if it goes wrong.

Eisenhower said that plans are useless but planning is indispensable, and something similar holds for analysis that has started to spin. The work of analysing clarifies, right up until the point where it quietly becomes a substitute for deciding. Telling those two states apart in the moment is hard. The organisation posts later in the series, on why capable companies make poor decisions and on what happens when a room agrees too quickly, are partly about the conditions that turn useful analysis into expensive stalling. What I do not have is a clean rule for the exact moment the loop should stop. I am not sure one exists.