Understanding Results

After generating an assignment, Shibutz presents a detailed statistics page so you can evaluate how well the algorithm balanced your classes. This page is your go-to resource for reviewing the quality of each placement run before sharing results with staff.

Overall Assignment Statistics

The top of the results page shows a high-level summary of the entire assignment. This gives you a quick sense of how balanced the outcome is across all classes:

Total students placed — confirms every student was assigned to a class.
Number of classes — the classes that received students.
Overall balance score — a quick indicator of how evenly students were distributed across all dimensions (gender, scales, friend requests).

Per-Class Breakdown

Below the overall summary, each class has its own statistics card. These cards let you drill into the specifics of every class:

Student count — the number of students assigned to this class. Compare across classes to spot any imbalances.
Gender ratio — the percentage of boys and girls in this class relative to the target balance.
Scale averages — the mean score for each scale (social, emotional, behavioral, learning) in this class compared to the overall average.
Origin school distribution — how students from different feeder schools or groups are spread across this class.

Friend Placement Rates

One of the most important outcomes for students and parents is whether friend requests were honored. The results page shows:

Total friend requests — how many preferences were submitted.
Requests honored — the number (and percentage) of friend pairs that ended up in the same class.
Unmet requests — pairs that could not be placed together, often due to conflicting restrictions or class size limits.

A friend placement rate above 70–80% is typical for a well-configured assignment. If the rate is significantly lower, consider whether too many restrictions are limiting the algorithm's flexibility.

Scale Distribution

The scale distribution section visualizes how the four student scales are spread across classes:

Social — measures social interaction tendencies.
Emotional — reflects emotional regulation and wellbeing indicators.
Behavioral — captures behavioral patterns in classroom settings.
Learning — represents academic engagement and learning readiness.

Ideally, every class should have a similar average for each scale. Large deviations between classes suggest that one class may be carrying a disproportionate load in a particular dimension.

Interpreting Your Results

A good result typically means:

Class sizes are within one or two students of each other.
Gender ratios are close to the overall school ratio in every class.
Scale averages per class are close to the global average.
The majority of friend requests were honored.

A result that needs improvement might show:

One class with notably more students than others.
A large gap in scale averages between classes.
A low friend placement rate (below 50%).

What To Do If Results Aren't Satisfactory

If the assignment doesn't meet your expectations, you have several options:

Re-generate — run the algorithm again. Each run explores different configurations and may produce a better balance.
Adjust constraints — review your restrictions and preferences. Removing or loosening conflicting rules gives the algorithm more room to optimize.
Review pre-assignments — if you have many pre-assigned students, they reduce the algorithm's flexibility. Consider whether all pre-assignments are necessary.
Check settings — verify your assignment settings (max students per class, balance thresholds) are realistic for your cohort.

Once you're happy with the results, head to the Exporting to Excel guide to download and share your class lists.

All Docs