Design: UI when there are few/no results

[matejminar]

Problem: We have certain product workflows today where customers can "drill down" into their transaction data to look for specific samples by using a search function:

• Starting from the Transaction Summary
• Some also go directly to Discover

Our biases try to offer a good baseline of samples out of the box, but we cannot guarantee them for very specific queries.

Proposed solution:
When customers are presented with no or few samples after a search within Discover and the Tx Summary page, give them an option to temporarily collect samples that match the given criteria.

For the POC the following is sufficient:

• the user can easily create a new investigation bias
• the user has a way to know whether the collection is in progress of not
• the user has a way to get notified about the results (when we've collected enough samples or how many we collected by the time the 1h time interval expired).

Other Notes:

• The bias will only support sampling for transactions, not full tracing
• In case the query cannot be transformed into a rule (relay condition), we need to handle this in the UI
  • We can display which tag is not supported and suggest the user to drop it
• once the collection starts, it stops either once it collects 100 samples or for maximum 1h (whichever comes first)
  • The number of samples is currently hardcoded to 100, this can change in the future
• While the bias is active, we need to show a visual indicator that the collection of sampling is in progress
• Customers can have multiple investigation biases concurrently (there is a high abuse limit)
• For now, there temporary biases cannot be deleted/deactivated (This would be problematic because a new rule that matches the exact conditions of a pre-existing rule, so it can get messy).
• We want to notify users once we have collected enough samples/the collection has stopped
  • The user should still specify who should be notified about updates (for now just the user who created it)
  • Even if we haven't collected the 100 samples, we should inform them about how many samples were indeed collected
• Performance issues are another starting point Example 1 and Example 2, but those we should likely tackle with automated biases and not in scope of the POC.

[Jesse-Box]

Marking this ticket as done

Copywriting is still TBD

Figma
https://www.figma.com/file/fRiNjr8CL4Cdhgf6esv5Pm/Specs%3A-Investigation-Bias?type=design&node-id=35%3A7313&mode=design&t=05AUk3jf9ZyQHzw5-1