Most of the value is not in the headline percentage. It is in the reasons and in the individual twins, because those are what survive contact with a decision. A number tells you what happened; a reason tells you what to do about it.
What it contains
The distribution
Each option in your action space with the share of the population that chose it, and a range around that share.
Leave a tip 63% ±2.4
No tip 37% ±2.4
Read the range as seriously as the number. Two options whose ranges overlap have not been separated by this simulation, however different their headline shares look. If separating them matters, a larger sample will narrow the ranges — with diminishing returns.
The ranges account for the fact that a sample is weighted to match your population's composition. Weighting is what makes a few thousand twins speak for a market, and it costs some precision: a result where a small number of people carry a lot of weight has wider ranges than the raw count would suggest. That is correct, and the population page warns you when a group is thin enough for it to matter.
The reasons
The twins' own reasons, grouped into the rationales that recurred, with the share of each option they account for:
Leave a tip · 63%
41% protecting the next order
28% the driver is the one who loses out
19% it is pre-selected and not worth changing
12% other
This is the part of a result that is hard to get any other way. It comes from what each twin said, in its own words, grouped — not from a summary written about them. Every rationale opens onto the twins who gave it, so a cluster you do not believe can be checked in about ten seconds.
The subgroups
The same answer broken out by parts of your population — age, region, income, whatever your population is stratified on. "63% overall, 41% among under-25s" is often the finding, and it is free: the twins already carry those characteristics.
Groups too small to say anything about are suppressed rather than shown with a misleading number.
The takeaway
One line, saying what the result means. It is written from the distribution and the rationales, and it never contains a number that was not computed — every figure you see anywhere in a result comes from counting the twins.
The twins themselves
Every individual, with its identity, its choice, its reasons and the parts of its identity that drove them. Filter to the people who chose a particular option and read a few. You can ask any of them a follow-up, or ask the same follow-up of a whole group at once.
What was excluded
Any twins that did not resolve, and why. They are excluded from the percentages rather than quietly counted, and the number is shown. A run that lost a meaningful share of its sample says so.
What the twins were told
The environment for this run: every fact, its source, and whether it was established or assumed. Collapsed by default, and worth opening the first few times.
How to read one
Start with the twins, not the percentage. Read five who chose the leading option and five who did not. You will learn more in three minutes than from the distribution, and you will immediately know whether the population is right.
Then the rationales. These are what you will repeat to other people.
Then the distribution and its ranges, for the ranking and the rough magnitude.
Check confidence before you act on the exact figures.
Then the subgroups, which is usually where the actionable difference is.
What to do with it
Use it to rank options, size a gap, and find the reasoning you had not considered. Do not use it as a substitute for the one real experiment you can actually afford to run — see Limits.
If two options came out close, the useful next step is rarely a bigger sample. It is a comparison: change one thing and see which direction it moves.