Anslation Persona
How Persona works. Every rule in plain words.
Every Persona result places you on each pattern's own scale from 0 to 100, read in bands, by default 0–34, 35–65 and 66–100. When your top patterns are within 0.5 points of each other, the result says Even profile instead of picking one. Confidence is Strong at 85% of the asked questions and 8 answers, and Moderate at 50% and 4.
By Anslation Editorial Team · Checked against the scoring code on
How is a Persona result worked out?
Each rule below is the one the scoring code follows today, with the numbers it uses.
Positions, not grades
Each pattern's position is where your answers sit between the lowest and the highest total its questions allow, from 0 at one end to 100 at the other. Statements worded the other way round count in reverse. A question whose answers would all give a pattern the same value is not counted for it, and a pattern none of your answers reached shows no score, never 0.
Bands
Each pattern's scale is split into bands written by the test's authors. The standard set is 0–34, 35–65 and 66–100. Positions keep two decimal places, so one can fall between two bands: it takes the nearer band, and exactly halfway, such as 34.5, goes to the higher one.
Near a band edge
A position within 3 points of the edge between two bands is marked as near a band edge, because a small change in your answers could move it into the next band.
Even profile
When two or more of your top patterns are within 0.5 points of the highest one, the result says Even profile and names them all instead of picking one to lead. The rule is applied once, where the result is scored, so every page that shows the result agrees.
Confidence
Confidence comes from how many of the asked questions you answered. Strong needs at least 85% of them and at least 8 answers; Moderate needs at least 50% and 4 answers; anything less is Limited. It describes how much a result rests on, not whether it is right.
Strengths and watch areas
By default a pattern at 66 or above is listed as a strength, and one at 34 or below as a watch area. A test version can set its own lines. Both describe where answers sit on that test's scale.
Fixed rules
No language model writes or scores anything in Persona. Scores, readings, practice suggestions and team figures all follow fixed rules.
One pattern, three bands
Example pattern: Planning
Its bands and their stretch of the scale
- Adapts as it goes 0–34
- Balanced planning 35–65
- Plans ahead 66–100
| Confidence | Asked questions answered | Answers, at least |
|---|---|---|
| Strong | 85% or more | 8 |
| Moderate | 50% or more | 4 |
| Limited | Anything less | – |
How are tests built and checked?
How a test is put together before anyone takes it, and how its questions are checked once people have.
Adaptive tests
An adaptive test asks part of its scored questions: 65% of them by default, rounded up, and at least 5 unless it has fewer. Its facts line says "Up to" before its number of questions.
Skill signals
A pattern can feed one of 7 work skills: the one its test's maker chose for it, else the one whose words its code and name use or, failing that, its description (in a result, also the name of its band). A pattern that matches none feeds a general "Pattern awareness" signal instead, which has no skill page. A skill's signal averages the positions of the patterns that feed it, and a pattern you did not answer is left out rather than counted as 0.
Reliability checks
For test authors, Studio works out Cronbach's alpha for each pattern, with a Feldt 95% interval, once at least 10 people have answered every question of it. Alpha from 0.80 reads Consistent, from 0.70 Acceptable and from 0.50 Weak; lower is Inconsistent, and below 0 Items disagree. Studio previews never count.
Publishing
A version cannot be published while a pattern has no bands or its bands leave a gap or overlap, it asks for consent without saying what the taker agrees to, its safety notes are missing, it has fewer than 6 active questions, a question is not tied to a pattern, or a blocking review item is open.
Versions
A published version is frozen. Changes go into a new version, and every result keeps the scoring of the version it was taken on. A preview of a draft in Studio is nobody's result and never counts.
Who sees results, and what is recorded?
Who can see what once a result exists, and what is recorded while you answer.
Group figures
Group figures need at least 5 people other than the person reading them. Head counts are rounded down to a multiple of 5, and there is no ranking of people and no list of lowest scores.
Share links
A result is shared only when its owner shares it. Every share link expires: after 30 days unless the owner picks another time, and never later than 365 days. The shared page shows the test, when it was taken, the headline and a band per pattern, never positions, scores or answers.
Answer time
How long each answer takes is recorded, so test authors can find questions people find confusing. It feeds group statistics about the test, never anyone's result.
What Persona never does
Self-insight only.
Self-insight, not a diagnosis or a hiring decision.
No type labels.
Results show patterns on scales, in words such as "suggests" and "may prefer", never a fixed type.
No ranking of people.
There is no ranking of people and no list of lowest scores. Team views show status and group figures only.
No guesses about who you are.
Persona does not infer protected characteristics, such as gender, age or religion, from your answers.
No promises about your future.
A result describes how you answered today. It does not predict what you will achieve.
No language model.
Your answers are never sent to a language model. Scoring, readings and practice suggestions follow fixed rules.
Words used on these pages
- Pattern:
- One thing a test looks at, such as planning or listening, shown on a scale of its own, between two named ends when the test names them.
- Position:
- Where your answers sit on a pattern's scale, from 0 to 100. It is not a grade; where the scale names two ends, neither end is better.
- Band:
- A named range of a pattern's scale, with a short description written by the test's authors.
- Even profile:
- What a result says when its top patterns are within 0.5 points of each other, instead of naming one to lead.
- Confidence:
- How much a result rests on, from how many of the asked questions were answered: Strong, Moderate or Limited.
- Reliability:
- How well the questions on a pattern agree with each other, checked by test authors with Cronbach's alpha.
- Skill signal:
- How often a work skill shows up in your answers across tests, as the average position of the patterns that feed it: emerging, steady or strong, never a grade.
- Group floor:
- The smallest group a team figure may describe: 5 people other than the person reading it.
- Grow:
- Where you pick one pattern to work on and try practice steps chosen by fixed rules from your results, and see how often each skill shows up over time.
- Studio:
- Where Persona managers write a test, preview it the way a taker sees it, check it against the publish checklist and publish it as a version that never changes.
- Library:
- Every test you can take once your account has Persona access, with the tests someone has assigned to you and the packs that group them.
- Mission:
- One test, or a ready-made template of tests, that a Persona manager assigns to people with a due date; each person accepts it before they start.
References
- Cronbach, L. J. (1951). Psychometrika, 16(3), 297–334. doi:10.1007/BF02310555
- Feldt, L. S. (1965). The approximate sampling distribution of Kuder-Richardson reliability coefficient twenty. Psychometrika, 30(3), 357–370. doi:10.1007/BF02289499