A population is made of twins. A simulation puts a question to a sample of them and adds up what they each decide. So the twin is the unit the whole product is built from, and the quality of a twin is the ceiling on the quality of everything above it.
The word matters. A twin is not a persona in the marketing sense — not an archetype assembled from intuition and given a stock photo. Every twin is built from evidence about people who actually exist, and every line of that evidence can be traced back to where it came from. See Identity and Source.
What a twin holds when it decides
Four things, and nothing else:
Its identity. Who this person is: their circumstances, how they behave, what they have said. This is stable. The same twin carries the same identity into every simulation it appears in.
The environment. What is true in the world it is deciding in — prices, norms, what is available, what has recently happened. Every twin in a given simulation gets the same environment, which is what makes their disagreements about them rather than about what each of them happened to know.
The date. The moment the situation is set in. A twin reasons from what was true then and knows nothing after it.
The decision. The scenario it is in, the question it is being asked, and the action space it must choose from.
What a twin is not told
It is not told that it is in a study, that it is one of several thousand, or that anything is being measured. People answer differently when they know they are being surveyed, and a twin that knows it is a twin starts performing rather than deciding.
It is also not told which population it belongs to. A population shapes a twin's decision by determining which people are asked and which environment applies — never by telling the twin what group it is in.
What a twin gives back
Every twin returns four things:
- The action it chose, always one of the options you listed. A twin cannot answer off the menu, and cannot decline to choose.
- Why, in its own words, in the first person and in a few sentences. This is the part worth reading. The percentages tell you what happened; the reasons tell you why, and the why is what survives contact with a decision.
- A short label for its reason, which is how reasons get grouped across thousands of twins into the recurring rationales you see in a result.
- Which parts of its identity drove the choice, so you can see the connection between the evidence and the decision rather than taking it on trust.
Some twins also record what they would have done if their first choice had not been available, which is the fastest way to notice a missing option in your action space.
Asking a twin more
A twin's decision is not the end of the conversation. You can open any twin in a result and ask it a follow-up: why not the other option, what would have changed your mind, what would you do at a different price. It answers as the same person, with the same identity and the same situation in front of it.
You can do this across a group as well — pick the 37% who chose one option and ask all of them the same follow-up. That is the closest thing to a focus group in the product, and it is often where the useful finding is.
Two things to know about how twins remember. Within one line of questioning, a twin remembers what it has already told you. Across simulations, it does not: the same twin answering a question in March and again in June is not carrying the March answer into June. That is deliberate. A twin that remembered its previous answers would start agreeing with itself, and consistency between simulations would stop being evidence that its identity is well-built.