EnneaFox methodology: how the assessment is built and scored

Design goals

We set four goals for the assessment. It should ask about motivation rather than surface behavior, because the Enneagram defines types by what drives a person rather than by what they do. It should show uncertainty honestly instead of forcing a confident label onto an ambiguous pattern. It should be fully rule-based so that anyone can understand why a result came out the way it did. And it should claim no more than the evidence on the Enneagram supports.

Item design

The assessment version currently live, called ef-v2, contains 60 items: 54 rated statements and 6 forced-choice discriminators. All items were written for EnneaFox and are not adapted from any published or proprietary inventory.

Rated statements

• Six statements per type, for 54 in total.
• Each statement describes a motivation, a habitual reaction, or a priority. We avoided statements that read as virtues, because people tend to endorse flattering items regardless of type.
• One idea per statement, no double negatives, and no clinical vocabulary.
• Exactly one statement per type is reverse-worded. Agreeing with it counts against that type.
• Statements within a type were checked so that no two are near-paraphrases of each other.
• Answers use a five-point scale: Strongly disagree, Disagree, Neutral, Agree, Strongly agree.

The statements are presented in rounds that rotate through all nine types in a different order each round, so no two neighboring statements belong to the same type and the reverse-worded items are spread out rather than clustered.

Forced-choice discriminators

Six either-or items each contrast one pair of types that are often confused: 2 and 9, 4 and 5, 5 and 6, 6 and 9, 3 and 8, and 1 and 6. Each item presents a situation and two reactions written to be equally acceptable. The display order of the two options is shuffled for each attempt.

Scoring rules

Scoring runs on our server using whole-number arithmetic, so the same answers produce the same result on any device. In plain terms:

1. Each rating becomes a value from minus two to plus two, with "Neutral" as zero. Reverse-worded statements are flipped.
2. Each type's values are averaged. A type with fewer than four answered statements is left unscored.
3. Any type whose average is within 0.20 points of the leader is a contender.
4. If there are at least two contenders, each discriminator that contrasts two of them adds 0.15 points to the type you chose. Discriminators between types that are not both contenders are ignored.
5. Types are ranked by their adjusted score. Exact ties are broken by the unadjusted average, then by the number of answered statements, then by type number, so the order is always the same. An exact tie at the top is flagged and labeled close.
6. The best-fit type is the top-ranked type, provided all nine types were scored.

Clarity

The lead is the gap between the top type and the second type. A lead of 0.40 points or more is labeled clear. A lead from 0.20 up to 0.40 is moderate. A lead under 0.20 is close, and the report lists the contenders.

Wings

The wing check compares only the two types on either side of your best-fit type on the Enneagram circle. If one neighbor scores at least 0.20 points higher than the other and its score is not negative, the report shows a wing leaning. If the gap is smaller, the wing is balanced. If both neighbors score below neutral, the wing is unclear. We present wings as exploratory, because the wing concept has very little research behind it.

Alignment

Each type's adjusted score is shown on a descriptive 0 to 100 scale, where 50 corresponds to answering "Neutral" on average. It describes how strongly you endorsed that type's statements. It is not a probability that you are that type, and it is not a comparison with other people.

Response style

The report notes patterns that make self-report harder to read: giving the same answer to most statements, agreeing with nearly everything, disagreeing with nearly everything, or answering fewer than 48 of the 54 rated statements. These notes are cautions, not accusations.

Thresholds are design choices

The numbers above, such as the 0.20 contender margin and the 0.40 clear lead, were set by us when designing ef-v1. They are reasonable starting points, not cut points derived from a validation study. We say this plainly because a threshold that sounds precise can imply a certainty that does not exist.

Observations

Each rated statement has a matching reflection written in advance, such as "You said a small mistake can stay with you long after everyone else has moved on." When you answer a statement strongly in the keyed direction, its reflection can appear in your report. Observations describe what you said. They are not generated by AI and they do not make judgments about you.

Versioning

Every attempt stores the assessment version and scoring revision that produced it. Items in a shipped version are frozen. If we improve wording, change an item, or adjust a threshold, that becomes a new version with a new identifier, and earlier results remain tied to the rules that created them. A report always reflects the version you actually took.

What we do not claim

• We do not claim the assessment has been validated in a published study.
• We do not report reliability coefficients, norms, or comparisons with other takers, because we have not collected the data that would justify them.
• We do not claim that nine distinct types are an established scientific fact.
• We do not claim wings, stress and growth lines, or centers have research support. We label them as tradition.
• We do not offer therapy, mental health screening, or a diagnosis, and the assessment is for adults only.
• We do not let generative AI decide a score, a type, or an observation.