Curated Discovery Growth Engine

The Curated Discovery Growth Engine shifts the responsibility of discovery from a centralized algorithm to the platform's supply side (the creators or power users). This approach builds a social graph based on trust and goodwill rather than "sellable eyeballs," leading to higher-quality user acquisition and stronger network effects.

The Core Principle: The Control Framework

When designing discovery features, prioritize the "Control Principle." Every decision should be evaluated by whether it grants more agency to the users or the platform:

• Writer/Provider Control: Do they choose who is recommended?
• Reader/Consumer Control: Is the source of the recommendation transparent?
• Platform Role: Facilitate the connection rather than forcing the match.

Implementation Workflow

1. Identify Organic Cross-Pollination

Look for evidence that users are already trying to help each other grow.

• Signals: Are users manually linking to each other? Are they guest-posting or mentioning peers in comments?
• Action: Map these manual behaviors to identify where a native product feature could reduce friction.

2. Design the "Opt-in" Curation Flow

Instead of a "People You May Know" algorithm, create a curation tool for your power users.

• Simplicity: Allow users to pick 5-10 peers they genuinely admire.
• Placement: Insert the recommendation prompt at a high-intent moment (e.g., immediately after a user subscribes or completes a transaction).
• Transparency: Clearly state why the user is seeing the recommendation (e.g., "Lenny recommends these 3 newsletters").

3. Build the Viral Feedback Loop

Close the loop by notifying the person being recommended.

• The "Goodwill" Notification: Send an automated update to the user being recommended: "User X is now recommending you to their audience."
• Reciprocity Trigger: This notification serves as a natural prompt for the recipient to recommend the first user back, creating a self-sustaining growth loop.

4. Use the "Product Lab" Pilot Method

Before a wide release, test the feature with a hand-picked group of power users to ensure it doesn't "cheapen" the brand.

• Create a "Product Lab": A group of ~100 high-engagement users.
• Run the feature as an optional beta to see if it drives "low-intent" vs. "high-intent" growth.
• Monitor "Open Rates" or "Retention" of users acquired through this path to ensure quality remains high.

Examples

Example 1: Marketplace Recommendation

• Context: An Etsy-like marketplace where sellers want to support other artisans.
• Input: A seller picks 5 other shops they personally buy from.
• Application: After a customer buys a ceramic mug, the "Thank You" page says, "The Maker of this mug also loves these 3 shops."
• Output: High-intent traffic flows between shops with similar aesthetics, increasing the "Network-driven" sales metric.

Example 2: SaaS Integration Discovery

• Context: A B2B tool with a developer ecosystem.
• Input: A developer of a popular plugin recommends 3 complementary plugins.
• Application: When a user installs the first plugin, a modal suggests the curated list.
• Output: Increased stickiness of the platform as users build a more robust, interconnected toolset based on developer trust.

Common Pitfalls

• Fearing "Too Many Steps": Leaders often worry that an opt-in flow has too much friction compared to an automated algorithm. However, the "costly signal" of a manual recommendation often results in higher-quality conversions that outweigh the drop-off.
• The "Low Intent" Myth: Assuming that discovery-driven users will be low quality. If the recommendation comes from a trusted source (e.g., a writer the user just paid), the intent remains high.
• Ignoring the Robin Hood Effect: Failing to realize that early power users want to "share the wealth." If you don't provide a way for big players to boost smaller players, you miss out on a massive organic growth lever.
• Over-Automating: Avoid the temptation to "autocomplete" recommendations for users. This erodes the trust that makes the curated discovery engine work.