pull down to refresh

Hello Stackers,

I noticed a significant discrepancy between the data shown on my profile dashboard and raw engagement tracking.

This led me down a rabbit hole and ask myself:

  1. How are SN analytics actually calculated under the hood?
    Specifically, I'm questioning the latency and aggregation logic.
  2. Are we relying on real-time stream processing, or is there a batch-processing delay?
  3. How are nested replies factored into overall post engagement metrics, and how is bot traffic being filtered out?

Why This Matters

To make Stacker News a better hub for quality content, we need reliable, transparent metrics.

Who is this crucial for:

Content Creators: Understanding what truly resonates with the community.

The Ecosystem: Ensuring rewards align with actual value generated.

Proposed Improvements & Roadmap

I believe we can take SN analytics to the next level by making them more transparent and predictive:

Open Source Metric Definitions: A simple public breakdown of how "Views," "Votes," and "Time Spent" are measured to eliminate discrepancies.

Predictive Engagement Score: Implementing a machine learning model using historical data to forecast which posts are likely to trend, helping stackers find high-value content faster.

Real-Time Dashboard Updates: Moving toward a real-time data pipeline to reduce lag between interaction and data reflection.

Call to Action for the Devs

@niftynei @keyan @ek, could you shed some light on the current architecture?

Specifically:

  1. What database is driving the analytics?
  2. Is there a documented API for these metrics?
  3. What are your thoughts on current metrics?
  4. Have you noticed similar discrepancies?

Let’s discuss how to make our data as robust as our community.

  1. why are you tagging nifty and ek?
  2. point the bot (you used to write this post) at our open source repo and ask it these questions
reply
  1. I thought they are part of the SN dev operational teams.
  2. What do you mean? I can't recall what do you refer? And is this some sort of a forbidden topic?
reply
  1. they aren't
  2. ask chatGPT questions about our code: https://github.com/stackernews/stacker.news
reply