Onboarding, News, and Metrics: Completing the Social Loop for AI Platforms - #3

If discovery and ranking are the engine, onboarding is the ignition—and metrics are the dashboard. Here’s how to complete the loop.

Personalise at Sign-Up (Without Friction)

Collect just enough to be useful:

• Handles/Links (optional): GitHub, Twitter/X, LinkedIn—used to infer interests and graph links.
• Interests & Tasks: Pick from curated lists (e.g., “LLMs ≤ 7B,” “Segmentation”) so users don’t have to type.
• Regional cues: Offer popular local creators (e.g., Indian sign-ups see /segmind).

Immediately construct the first feed from these choices so the Home screen is warm.

Add a Lightweight “News” Surface

Many orgs with repos also publish engineering posts elsewhere. Pull headlines into a News tab scoped to the creators/companies a user follows. It’s a small lift that centralizes attention and reduces context-switching.

Close the Loop with Input → Output Metrics

Input metrics (lead indicators):

• Follows (from justified recs), profile visits, reactions
• Feed scroll depth, clicks to models/datasets/spaces

Output metrics (business outcomes):

• Time spent, D1/D7/D30 retention
• Creator conversion: consumers who later upload models, create Spaces, or publish datasets

Focus experiments on input metrics that most strongly predict retention and creator conversion. For example, if “2+ Bulk Follow packs” at day 0 lifts D7 by 8–12%, double down and expand packs by topic and region.

Why This Matters

AI platforms win by density: dense graphs, dense discussions, dense learning loops. Onboarding personalization + News + metric discipline ensures the graph gets denser every week, not just every launch.

The document below is a more detailed discussion of how Hugging Face can implement a feed that can draw on many different signals and become the home feed of choice amongst their large and growing user base.