What You'd Actually Like DoingWhat You'd Actually Like Doing. Every free career quiz ends with a numbered list of six. On forty answers, the difference between fourth and first is usually smaller than the error in measuring it.

Methodology: Measures the direction and strength of a person's pull towards six kinds of work, through original preference-rating and forced-choice items. Construct statement: It measures which kinds of work you are drawn towards and which you would rather avoid, across six facets of what a working day can be made of. It does not measure what you are good at, what you would be paid for, what you should study, whether you would be hired, or what you are worth — an interest is a pull, not an ability, and the two come apart often enough that treating them as one is the most common error in career advice. Declared response instruction: preference, not capability — every activity asks whether the respondent would seek that work out or avoid it, never whether they are good at it, and the report repeats that sentence rather than assuming the reader remembers it. Item format: thirty-six rating_scale items on a six-point scale with no midpoint, six per facet, of which exactly three per facet are reverse keyed so that agreeing with everything cannot produce an interested-in-everything profile; and four multi_select weeks of work, each listing six parts spanning all six facets, which supply a choice-based signal to set against the self-report. Scoring: ratings are reverse-keyed, then ipsatised within the person across all thirty-six, so response style and general enthusiasm drop out and only the shape of the profile is reported; elevation is reported separately as a descriptive fact. A respondent whose ratings carry no variation has no shape to report, so no ranking and no top interest is produced for them and the sitting is routed to the response-quality composite. Reliability is estimated as McDonald's omega from item count and an assumed average inter-item correlation, and is a stated target rather than a measurement; the standard error of each facet mean sets a sixty-eight and a ninety-five per cent band, and facets whose ninety-five per cent bands overlap are reported in the same tier rather than ranked against each other. No chance correction is applied and none is possible: this is a preference profile and not a test, so there is no key, no percentage correct and no chance score. No percentile is shown anywhere; the reference distribution on the report is modelled from the declared response distribution, labelled as a model with its simulated n, and carries no rank. Construct areas drawn on: the long-standing finding that vocational interests are structured rather than a list, and are better read as relative pull than as absolute level; ipsative and within-person standardisation as the accepted control for acquiescence and extreme responding in preference data; balanced keying as an acquiescence control; the well-documented separation of interest from measured ability, which is why no item asks about competence; the stability of interests through early adulthood and their usefulness for exploration rather than for prediction of success; the difference between expressed and inventoried interests, which is what the four weeks of work are for; measurement-error banding and the refusal to rank differences smaller than the standard error; social-desirability resistance through facets that are equally respectable to endorse; item fairness across markets, so that no activity assumes a vehicle, a garden, domestic help, a particular school system or a religious practice; and plain-English readability limits so that reading difficulty never becomes part of what is measured. No profession, industry, qualification, tool or company is named anywhere in the instrument, because an interest inventory that names jobs measures what a person has heard of. All items are original works, no trademarked instrument, branded framework or published interest taxonomy is named or reproduced, and no affiliation with any source is claimed. AssessAll original design.