Ranking report
The analysis of the ranking test (ISO 8587): the Friedman test, a rank-sum table, and homogeneous letter groups.
What it is
The ranking report is the output of the ranking test (ISO 8587), in which each panelist places the samples in order according to a criterion (least to most sweet, least to most bitter, least to most liked). In chocolate making, it lets you rank several samples along a single axis without asking for an absolute score for each one.
Unlike descriptive tasting, scores are not averaged here: you work with positions (ranks). That is why the analysis is not an ANOVA but the Friedman test, the right test for ordinal data with blocks (each panelist is a block that ranks the same set of samples).
How to get there
- URL:
https://app.areacacao.com/sensory-analysis/evaluations/<id> - How to reach it: open the ranking tasting under Evaluations and scroll down to the Results section.
Pro plan or above, or a tasting pack with credit left.
What the report shows
Friedman test
The report opens with the statistical verdict of the Friedman test, which answers a single question: did the panel rank the samples consistently, or is the order indistinguishable from chance? The report shows:
| Value | What it means |
|---|---|
| χ² | The Friedman statistic. The larger it is, the sharper the differences in position between samples. |
| Degrees of freedom (df) | They depend on the number of samples (number of samples minus 1). |
| p-value | The probability of obtaining this order by chance. If p < 0.05, there are significant differences between samples. |
A highlighted box puts the result into words: significant differences (the panel ranked the samples reliably) or no significant differences (you cannot claim the panel tells the sample order apart).
Rank-sum ranking table
A table orders the samples by their rank sum: the positions each panelist assigned to a sample are added up. Lower sum = better position on the ranked criterion (the sample most panelists placed first accumulates the lowest sum).
| Column | What it holds |
|---|---|
| Rank | The sample's place in the ranking (1 = best positioned). |
| Blind code | The sample's 3-digit code. |
| Name | The sample's internal name: this report does show it next to the code. |
| Rank sum | The total of positions assigned by the panel. The lower it is, the better the position. |
| Group | The homogeneous-group letter (see below). |
Homogeneous letter groups
A letter (ᵃ, ᵇ, ᶜ) appears next to each sample. These are the homogeneous groups: samples that share a letter do NOT differ significantly from one another in order; those that share no letter do differ. It is the quick way to see which samples ended up tied for your panel and which truly stood apart.
The report also indicates the number of panelists who ranked all the samples: the Friedman analysis only counts those who completed the entire ranking, because it requires complete blocks. If anyone didn't rank every sample, a note says how many panelists were discarded. With fewer than 2 panelists who ranked every sample, instead of the report you'll see the notice "Ranking needs at least 2 panelists who have ranked every sample."
Trend along the declared order
With at least 3 samples and 3 panelists with a complete ranking, the report adds Page's trend test: it checks whether the panel ranks the samples following the order in which you entered them when creating the tasting (for example, the days of a shelf-life study or an increasing cocoa %). The box says The panel ranks the samples following your declared order, …in reverse of your declared order, or No trend along your declared order, with L, z, and p, and repeats the order that was tested. It only makes sense if that order was the one you expected beforehand; if you entered the samples without a planned order, ignore it. Below the minimum you'll see "The trend test needs 3 samples and 3 panelists with a complete ranking."
How to interpret it
- A p-value below 0.05 confirms the panel ranked the samples consistently: the order is not coincidence.
- Read the table from lowest to highest rank sum: at the top is the sample best positioned on the criterion you ranked.
- Use the letter groups so you don't over-interpret: two consecutive samples that share a letter are, for practical purposes, tied for your panel, even if their rank sums are not identical.
- A no significant differences verdict does not prove the samples are equal: it means that, with this panel, no reliable order has been detected. Increasing the number of panelists gives the test more power.