The New Batch Is Bimodal: 17 Skills Above 8, 10 Below 5
The average score of the forty new skills is 6.79. That number is almost useless on its own.
Here is the distribution, in half-point bands and coarser buckets.
| Score band | Skills |
|---|---|
| 9.0 and above | 6 |
| 8.0 – 8.9 | 11 |
| 7.0 – 7.9 | 3 |
| 6.0 – 6.9 | 5 |
| 5.0 – 5.9 | 5 |
| 4.0 – 4.9 | 5 |
| below 4.0 | 5 |
Seventeen of the forty score at least 8.0. Ten score below 5.0. The middle is thin: only three skills land between 7.0 and 7.9. The batch is not a bell curve around 6.79; it is two clusters with a gap in the middle.
The median is 7.20, the mean is 6.79, and the standard deviation is 2.00. The mean is pulled down by the thin tail, not by a broad middle.
The two clusters
At the top: jira at 9.5, k12-lesson-differentiation at 9.3, k12-check-for-understanding at 9.2, k12-lesson-plan-creation at 9.1, dynamic-pricing-ecommerce at 9.1, and deal-qualification-gates at 9.0.
At the bottom: ecommerce-ab-testing at 3.2, conversion-rate-optimization and customer-feedback-analysis at 3.3, and shopify-seo at 3.5.
The top cluster is mostly methodology: education, revenue operations, a well-engineered Jira integration, and one e-commerce strategy skill. The bottom cluster is mostly short how-to files that describe a topic without giving an agent a procedure.
Why the average is the wrong summary
If you only read 6.79, you would conclude that the batch is mediocre. If you read the bands, you learn something more useful: the batch contains real material and filler, and the difference is visible at the skill level.
This is the same lesson as our earlier long-tail analysis: the directory is not one distribution. It is a collection of clusters, and averages flatten them.
The next posts break the clusters down by dimension: workflow, trigger quality, and the split between tool integrations and methodology.