How is Learning Progress read (Request, Client, Country, ASN network axes); why is the lowest axis decisive?

Short answer

Learning Progress is the ratio of the evidence a suggestion has accumulated to the threshold and appears as a green bar and a percentage in the lists on the AI Learning & Teacher Tracking tab. In AI Learning, a separate ratio is calculated for every threshold defined in the Learning Methods window: Request = number of requests / Required Requests, Client = different clients / Required Clients, Country = different countries / Required Countries, ASN network = different ASNs / Required ASN Networks; a threshold left empty does not produce an axis. The percentage in the list is the lowest of these ratios, and the suggestion is applied automatically only when the lowest axis reaches 100%. The lowest axis being decisive is deliberate: with the default thresholds, 1,000 requests from a single IP raise the Request axis to 10,000% but leave the Client axis at 20%, and the suggestion is not applied. The breakdowns are seen separately in the preview and can go above 100%. In Teacher there is a single axis: the number of Teacher clients that feed the suggestion; one client is enough.

Axes

Axis	Ratio	Threshold field (default)
Request	Number of requests / threshold	Required Requests (10, required)
Client	Different clients / threshold	Required Clients (5, required)
Country	Different countries / threshold	Required Countries (1)
ASN network	Different ASNs (the /23 block for an address whose ASN cannot be found) / threshold	Required ASN Networks (empty = no axis)
Teacher	Number of Teacher clients	None; 1 client is enough

• Example (default thresholds, 1,000 requests from a single IP, a single country): Request 10,000%, Client 20%, Country 100% → Learning progress 20%.
• The bar in the list fills up at 100%; the percentage label can exceed 100%. In the preview, the breakdown of each axis is written separately (preview of the selected suggestion).
• Automatic applying requires the lowest axis to reach 100%; Min Progress (%) in the AI Learning Tracking view affects only the listing (What are the Min Progress (%), Max Suggestions and Line Limit advanced settings for?).
• In a cluster, the ratios of each device are calculated with its own records; traffic split between two devices has to fill the threshold with the data of a single device.
• Each axis is calculated only if its threshold is defined. Since the threshold of the time slot axis is present neither in the interface nor in WEB-CLI, this axis does not appear in the preview.
• Applying manually: Can I apply AI Learning's suggestions manually without waiting for it to apply them automatically (Learn Selected)?

Common Misconceptions

• "The percentage is the average of all the axes." It is the lowest axis.
• "A value above 100% is an error." The breakdowns can grow to multiples of the threshold; what is decisive is the lowest axis.
• "The default thresholds filter out traffic that comes from a single country." The Required countries default is 1; traffic that comes from a single country fills this axis. To require multiple countries and ASNs, the thresholds are raised (Can an attacker fool AI Learning by making their traffic look widespread; how is this risk reduced?).

How it is done in the documentation

• AI Learning & Teacher Tracking · How Is Learning Progress Read?
• Learning Methods · Required Requests
• WAF · When Is the Learning Finished?

Related questions

• What is AI Learning and how does it work?
• Can an attacker fool AI Learning by making their traffic look widespread; how is this risk reduced?
• What are the Min Progress (%), Max Suggestions and Line Limit advanced settings for?