Market research tool
Penalty analysis for JAR scales
How results are checked
Named statistical review pending. Selected automated reference cases are published on the validation page.Biggest penalty
0.67weighted
Sweetness: too much
40.8% of respondents · liking drops 1.65 points
Penalty analysis shows how much overall liking a product loses when an attribute is "too little" or "too much" on a Just-About-Right scale. For each side it computes the mean drop — liking of the just-right group minus liking of that side — and multiplies it by the share of respondents on that side. Sides chosen by at least 20% of respondents are conventionally worth acting on.
Run penalty analysis on your data
Open a respondent-level file with an overall liking score and one column per JAR attribute, then pick the scale type. Everything runs in this browser tab — free, no account, and the file is never uploaded. "Use sample data" loads a synthetic snack test with sweetness, saltiness and crunch JAR scales.
CSV, Excel, SPSS, Stata, SAS, R, Parquet, SQLite or JSON. The file is read in this browser tab and never uploaded.
Reading the penalty plot
Each point is one side of one attribute. The horizontal axis is the share of respondents on that side; the vertical axis is the mean drop in liking. Points in the upper right — many respondents and a large drop — are the priorities. Points left of the dashed action line affect too few people to justify reformulating, however large their drop; a high point there is often noise from a small group.
A negative mean drop means that side actually liked the product more than the just-right group, which usually signals a confusing scale or a small group rather than a real preference. To see how JAR answers differ by segment, crosstab them in the workspace.
The method
Formula
Weighted penalty = (mean liking JAR − mean liking side) × proportion on side
where
- JAR group
- respondents at the scale midpoint (3 on a 5-point, 2 on a 3-point, 4 on a 7-point scale)
- side
- respondents below (too little) or above (too much) the midpoint, analysed separately
- mean drop
- mean liking of the JAR group minus mean liking of that side
- test
- optional Welch two-sample t-test of the JAR group against that side, with Welch–Satterthwaite degrees of freedom
Worked example
In the synthetic sample, 600 respondents rated overall liking (1–9) and three 5-point JAR attributes. For sweetness, 38.2% said "just about right" (mean liking 6.42), 40.8% said too sweet (mean 4.77) and 21.0% not sweet enough (mean 6.02).
"Too sweet" has a mean drop of 1.65 points and a weighted penalty of 1.65 × 40.8% ≈ 0.67 — the largest in the study. "Not salty enough" follows (35.7%, drop 0.99, penalty 0.35). Crunch is balanced: both sides sit just above 20% (23.3% and 22.3%) with small drops of 0.22 and 0.35.
The action: reduce sweetness first, then raise saltiness; leave crunch alone.
Using it well
When to use it
Use it when
- Use: Product tests or concept tests with Just-About-Right scales and an overall liking question.
- Use: Deciding which attributes to reformulate, and in which direction.
- Use: Comparing several prototypes on the same attributes.
Not the right tool when
- Avoid: Attributes measured on intensity or liking scales without a 'just about right' midpoint.
- Avoid: Very small samples — with fewer than about 20 respondents on a side, the mean drop is unstable.
- Avoid: Attributes that move together (for example sweetness and flavour strength) — penalties overlap and cannot simply be added.
How to interpret it
Rule of thumb
Act on sides that are both common (at least 20% of respondents) and costly (a mean drop of roughly half a point or more on a 9-point scale); rank them by weighted penalty. Ignore big drops from small groups unless the significance test supports them.
Questions
Frequently asked questions
What is a mean drop in penalty analysis?
The mean drop is the average overall liking of respondents who rated an attribute 'just about right' minus the average liking of respondents who rated it too little (or too much). It estimates how much liking is lost among people who find the attribute off.
Why is 20% the action threshold?
It is a widely used convention in sensory and consumer research: a side chosen by fewer than about 20% of respondents affects too few people to justify changing the product, even if their liking drops sharply. The threshold is adjustable in the tool.
What is the weighted penalty?
The mean drop multiplied by the proportion of respondents on that side. It combines how much liking is lost with how many people lose it, so attributes can be ranked by their total impact on mean liking.
Which significance test is used?
Optionally, a Welch two-sample t-test compares the liking of the just-right group with that side, without assuming equal variances. With survey weights, the tool uses weighted means and variances with Kish effective sample sizes, which is an approximation.
What JAR scales are supported?
3-point (2 = just about right), 5-point (3 = just about right) and 7-point (4 = just about right) scales coded 1 to the number of points. Answers outside that range are excluded and counted.
Sources and next steps
References & further reading
- [1]Rothman, L., & Parker, M. J. (Eds.). (2009). Just-About-Right (JAR) Scales: Design, Usage, Benefits, and Risks (ASTM Manual MNL63). ASTM International.
- [2]Lawless, H. T., & Heymann, H. (2010). Sensory Evaluation of Food: Principles and Practices (2nd ed.). Springer.
- [3]Welch, B. L. (1947). The generalization of 'Student's' problem when several different population variances are involved. Biometrika, 34(1–2), 28–35.
Try it on your own data — free, no signup
Upload a CSV, Excel, SPSS, Stata, SAS, R, Parquet or JSON file. Everything runs in your browser; your file never leaves your device.
Open the workspace →