Loading…
Message against no message
Per 1,000 people · against people whose WhatsApp Meta blocked (131049): same list, same app, no message
Both channels bring people into the Diwali chat at a similar rate. WhatsApp readers turn chats into calls more often, so WhatsApp does more per message; push reaches 2.5× as many phones.
Headline numbers
Android sends on 6 Nov, 09:49–15:41 IST
Time to tap
Message sent → first tap on the Diwali link
Funnel
Big number = conversion from the step before · people
WhatsApp
Push
Opened the Diwali chat, card and call count any activity by these people in the range, from any entry point.
Where messages drop off
Pick a channel · reasons add up to the loss
Effect on all outbound calling
6 Nov · every outbound call, not only Diwali-chat calls · range: low end adjusted on 5 Nov, high end on the week before
Tap rate by send hour
Batches sent later in the day tapped less; part of this may be batch order
iOS
Too few iOS sends to read a lift
Definitions
- Targeted: Android users on the Diwali send list. Reached the phone: WhatsApp delivered or read (incl. sent without a receipt yet); push accepted for a registered device (from a random sample of the send ledger).
- Tapped the link: a Diwali link open (Mixpanel deeplink, campaign diwali_2026) after the person's send. Push and WhatsApp share one link tag, so taps are split by which list the person was on.
- Message against no message: compares each channel with people whose WhatsApp was blocked by Meta's marketing limit. The audiences were chosen by different rules, so read each channel against that group, not head to head.
- Dates are IST days. The campaign went live on 5 Nov at 16:00 IST, so 5 Nov is a part day, and the latest day runs to the data cut shown at the top.
- Platform is the phone of the Equal user who opened the chat or set up the call.
- Full job means the assistant did all four parts on the call: wished (advance or Diwali day), said the "what", asked "anything to convey back?" and closed warmly. It is an estimate: the share of graded calls with all four parts, applied to every call where the receiver spoke at least twice.
- Sources: user-services read replica (calllog Diwali chats with nudge diwali_2026, outbound calls, transcripts, call events, scheduled tasks, users), Mixpanel (Diwali theme and app activity, message link taps), comms send ledger (WhatsApp and push status).