How the panel scores a page

Every panel answers the wording below, exactly as it is written here. That is what makes two runs comparable: a run against a reworded statement still produces a number, and the number silently means something else. The statements are handed to the panel by the engine and rendered on this page from the same source, so the two cannot drift apart.

Six agreement statements

A five-point scale, every point labelled

The psychometric literature converges on five to seven points. Below five, reliability and information measurably fall away; above seven, and especially at eleven, the extra points add noise without adding reliability. Five is the most widely used, and labelling every point rather than only the two ends keeps respondents interpreting the scale the same way, which is what holds measurement error down.

A 0–100 score would claim a precision a single shopper cannot supply. Nobody can tell 61 from 64. A labelled ordinal scale is the finest grain the answer actually has.

One verdict

Each persona also says where it landed, in shopper’s words. The three counts across the panel are what the report prints beside the means, because a panel of three ready and two bouncing is a different page from a panel of five warming to the same average.

Each persona answers one purchase-intent question, 0–10:

The PPI is the panel’s plain mean of those answers, to one decimal.

Juster’s eleven-point probability scale predicts real purchase better, but only once it has been calibrated against actual purchase data, which a synthetic persona cannot supply. The PPI is a directional index, and it is reported as one. It is not a conversion forecast.

The report prints each statement’s mean on the 1–5 scale it was answered on, and beside it the number of panel members who agreed — the standard top-two-box definition, Agree plus Completely agree.

Nothing is rescaled to 100. A mean of 4.4 sits at 4.4 on a five-point axis, and the label the scale gives that point is the vocabulary used to describe it.

• False precision. No 0–100 certainty score from a single simulated shopper, and no percentage over a panel of five.
• Expert-judge framing. Personas answer as the shoppers they represent. None of them is asked to review the page as a marketer.
• An average that hides the split. The per-statement means, the agreement counts and the ready / warming / not-yet breakdown are all reported, because the distribution is where the finding is.
• A first run pretending to be a comparison. A delta against nothing does not exist, and the report leaves it blank rather than printing a zero.

• Number of Response Options, Reliability, Validity, and Potential Bias in the Use of the Likert Scale — a literature review
• Should all scale points be labeled? — MeasuringU
• Intent scale translation (the Juster scale) — overview
• Bias and variability in purchase intention scales — Journal of the Academy of Marketing Science