The Brief
Trading 212 needed a strategy to boost engagement in their in-app community feature. The challenge: propose a minimal, practical plan to surface relevant community posts via a single, highly personalised push notification. I had two pages to define the problem, design the selection logic, show worked examples with real data, define notification copy, and outline a roll-out plan with success metrics.
Constraints: Ship fast (single sprint), keep it simple and auditable, prioritise safety compliance over volume, and leverage only first-party user data (followed authors, holdings, viewing history).
Unlock retention with “Community Pulse” notifications
Every client is looking for an edge and our community is a goldmine of timely, actionable investment insight. Currently too much of that high-value content goes unseen. If a client follows a top author or holds a crucial stock like AMD and a staff analyst posts a critical update, the user should know about it and be extremely interested due to the relevance of the content.
Failure to deliver this personalised content in a timely manner hurts us in two ways:
- Retention: It reduces the stickiness of our platform, making us feel like just another trading app, not a community hub.
- Engagement: It slows the adoption of our community feature, limiting the positive network effects that drive organic growth.
We will transform passive observers into active, retained community members by ensuring every user receives one perfect post that is relevant to them, exactly when they need it.
Strategic focus and guardrails
- Relevance stands out: Personalised notifications based on user context are proven to deliver significantly higher click through rate (CTR) than generic alerts. We will validate the precise lift in the A/B test.
- Trust is paramount: Notifications must be rigorously pre-filtered for policy violation risks. We must protect the brand above all else.
- Keep it simple: The model must be straightforward, auditable and deployable within one sprint, minimising engineering workload for maximum immediate return.
The logic behind “Community Pulse”
We are deploying a new, single push notification; Community Pulse – triggered only when a high value community post is created that meets a user’s specific, proprietary criteria. The post is selected by a relevance hierarchy model. The model calculates a total score for each post using weighted components and relevance bonuses to ensure precision and safety.
| Component | Rule | Score | Rationale |
|---|---|---|---|
| Safety filter | If reports_24h > 0 OR risk_score >= 1 | Exclude post from scoring pool | Eliminate spam and violations |
Base |
Sum of freshness and engagement | Score += (Recency + Engagement) | Prioritise fresh and trending content |
Relevance1 |
If == Followed author | Score += 1.5 | Leverage user intent |
Relevance2 |
If == Ticker match | Score += 1 | Ties content to capital |
Trust |
If == Staff post | Score += 1 | High quality authority content |
| Final score | = Base + Relevance1 + Relevance2 + Trust |
Highest wins | Most relevant (tiebreaker == newest wins) |
Notification copy to maximise click through rate
The copy will leverage the psychological principle known as the Zeigarnik Effect (the desire to complete unfinished tasks) by truncating compelling post content. This creates a “curiosity gap” which compels the user to tap the notification and complete the story.
| Type | Relevance | Example copy | Hook |
|---|---|---|---|
| Followed author | Mentions author | @Alex: I’m selling all my TSLA because the new… | Trusted opinion is truncated, creating urgency and curiosity |
| Staff / Ticker | Mentions trusted source and stock | AMD: Staff analyst explains why the chip sector is missing… | Knowledge gap to be filled by opening notification. |
Roll-out plan and success measurement
We are not comparing Community Pulse against existing push notifications due to the specific goal of this feature. We will measure community engagement uplift (reactions, responses and onward journeys) over a 1 day, 7 day and 30 day window. We can proceed to form data-driven hypotheses based on patterns identified to establish a future roadmap.
Frequently asked questions
This is a living document. Please direct questions to [email protected] and you will be answered directly, with common queries to be answered in this document.