WhatsApp
Case Study.
A Promotions filter for WhatsApp. Marketing messages go into their own tab. Delivery codes, order updates and real conversations stay in Chats.
Promotions and delivery codes shouldn't compete.
In India, shopping updates, offers, and ordinary conversations often arrive in the same WhatsApp inbox. Blocking a brand may stop promotions — but it can also interrupt useful updates from that same business.
“I get 7–10 promotional messages in a day, but I still need to find my Amazon or Flipkart delivery code.”
Classify messages, not businesses.
A single business can send both an offer and a delivery code.Explore the proposed feature.
This is a working interaction model, not a static screenshot. Switch inbox modes, open conversations, send messages, simulate delivery updates, and test settings.
One sender.
Two different purposes.
Amazon sends both a sale offer and a delivery confirmation. Promotions routes the sale; Chats preserves the delivery update. Opening either shows one original conversation.
The business account isn't permanently labeled as promotional.
10 product decisions.
Every design choice came from a specific product trade-off. Expand a decision to see the reasoning, constraints, and the alternative we deliberately didn't choose.
Use existing template categories.
Instead of analyzing every private message with AI, the proposed system starts with existing approved business-template categories and a small set of inbox-state rules.
A WhatsApp Flow is an interactive experience; the approved template used to send it may be Marketing or Utility. A customer opening a conversation from an Instagram or Facebook ad is not automatically a Marketing-template message. In the MVP, free-form replies without trustworthy classification stay in Chats.
Separate delivery updates from discounts.
Today, templates containing both utility and promotional material can be classified as Marketing. Our proposal requires Meta to enforce clearer separation before approval. This is a new policy dependency, not an existing guaranteed safeguard.
Launch blocker until mixed-template safety is validated.No category? No automatic hiding.
Inbox views reference messages, never split history.
Only meaningful interactions reset the 7-day window.
Search, access, and cleanup undo remain available.
Product requirements.
A concise PRD that distinguishes MVP behavior, important dependencies, and what's intentionally out of scope. Tap a requirement to read its acceptance criteria.
MVP includes
- Promotions filter and clean Chats view
- Automatic template-category routing
- One conversation with two entry points
- Customizable promotional notifications
- 7-day active-conversation lifecycle
- Opt-in beta, controlled rollout
- Optional metadata-backed historical cleanup
Not in V1
- AI content inspection and classification
- SMS categorization
- Blocking or deleting messages
- Personalized offers or recommendations
- Entire business-account labeling
- Automatic sorting of free-form replies
Risks and feasibility.
The interface can be designed quickly. Production readiness depends on internal message metadata, template-policy enforcement, and a migration path that doesn't surprise users.
UI and conversation state
New filter, message-level views, notification controls, and seven-day active-state logic are standard product engineering patterns.
Consumer metadata access
Business categories exist, but category data must be reliably available to the consumer client for routing.
Mixed-template safety
Critical transactional content inside Marketing templates must be prevented or safely handled before broad deployment.
| Product risk | Level | How the proposal responds |
|---|---|---|
| Delivery codes hidden by Marketing classification | Critical | Separate templates by policy; audit existing templates; gate rollout on safety results. |
| Free-form promotions remain in main Chats | Medium | Accept MVP limitation; quantify residual clutter before expanding scope. |
| Users fear conversations have disappeared | High | Onboarding explanation, global search, original thread preservation, cleanup undo. |
| Business message engagement decreases | High | Track read rates, support replies, blocks, and commercial guardrails. |
| Cross-device inconsistencies | High | Consistent routing semantics and cross-device quality assurance. |
| Historical category metadata missing | Medium | Offer cleanup only for messages with trustworthy recoverable metadata. |
How we would measure success.
The measure isn't the number of messages moved. It's how much unwanted clutter disappears from Chats — without losing important communication.
Explore the assumptions.
Change the assumptions to see the potential reduction in promotions appearing in Chats. These aren't observed research results.
Reduction in unwanted promotional messages appearing in the main Chats inbox vs control.
No confirmed pilot misroutes; larger statistically adequate validation before scale.
Proposed ≥99% correct routing of eligible Marketing template messages.
Inbox satisfaction, feature retention, support replies, business engagement.
Interview 10–15 frequent WhatsApp users about real promotional overload; usability-test both inbox views with mixed-brand messages; then run a privacy-preserving randomized treatment/control experiment. Do not present fictional prototype messages or model inputs as measured outcomes.
Less clutter.
Same conversations.
Start with business template categories Meta already uses. Keep Utility and Authentication messages visible. Preserve one original thread, even when the message appears in different inbox views.