Analytics for Thunder

[sahandilshan]

Problem

Thunder has zero visibility into user behavior and system operations.
We cannot answer basic questions:

"How many users logged in today?"
"Which features are being used?"
"What's causing errors?"
"Where do users struggle in our flows?"

Impact:
Product decisions made without data
Slow incident debugging
Cannot measure success or adoption
Missing optimization opportunities

What We Need to Decide

To implement analytics for Thunder, we need to make decisions across four key areas:

1. What to Publish?
   Questions:

   • Which events initially? (user actions, system operations, errors)
   • What's mandatory vs nice-to-have?
   • What should we NOT publish? (PII, sensitive data)

2. How to Publish?
   Questions:

   • Technical approach? (listener pattern vs message queue)
   • Performance requirements? (latency, throughput)
   • Reliability needs? (acceptable loss rate)

3. Where to Publish?
   Questions:

   • Output destinations? (analytics DB, files, cloud, dashboards)
   • How many destinations initially?

4. How to Visualize?
   Questions:

   • Does Thunder provide dashboards?
   • Target audience? (engineering, product, business, end users)
   • What's the MVP? (dashboard, raw data, API)

[sahandilshan]

The solution of this have been discussed here #578. We need to finalise on the scope of this feature by addressing the aforementioned four questions.