Every English reel belongs to one cell: who it's for, their moment, the pain, the end result, the emotion, one creative signal and one point of view. Low kcal stays as proof on every ad and leads only one cell. Nothing is generated until you OK this map.
Each cell leads with a different signal, so the EN test reads ten hooks, not one. The tier comes from a blind evidence check of the live EN ads: EXPAND builds on a proven EN winner, RETEST keeps a proven signal in a new voice, TEST is your new signal (CN-only or untested in EN).
One EN reel per approved CN reel type, each its own creative with its own ad_id (minted before generation). The opening line is a draft for the cell; real-creator reels get no authored lines. Jev pick both orders agreed · Jev best fit orders split, best average · no natural fit, or the cell was full.
| Reel type | Cell | EN opening line (draft) | Placed by | EN version needs |
|---|
Tap "Go with your pick", or "Different" with a note. That's all I need before making the EN reels.
| All WhatsApp ad sets | 7D | 14D | 30D |
|---|---|---|---|
| EN · spend / chats / RM per chat | RM2,430 / 316 / RM7.69 | RM5,235 / 684 / RM7.65 | RM13,682 / 1,914 / RM7.15 |
| CN · spend / chats / RM per chat | RM2,780 / 265 / RM10.49 | RM6,038 / 713 / RM8.47 | RM14,192 / 1,701 / RM8.34 |
| EN traced orders (RM spend per traced order) | 5 (RM486) | 12 (RM436) | 28 (RM489) |
| CN traced orders (RM spend per traced order) | 6 (RM463) | 13 (RM464) | 36 (RM394) |
EN buys cheaper chats; CN turns chats into sales a little better (same as September). EN website ad sets: about 0 purchases.
| EN ad (30D) | Its opening | Spend | Chats | RM/chat | CTR | Traced 14D |
|---|
Source files: ~/.zennith/state/mirra-oct-2026/en-map/ (map-v2.json, en_map_jev.py, allocate.py, jev/ receipts) · handover HANDOVER-EN-REELS-STRATEGY-2026-10-14.md (v2). Decisions here are saved with your name and time and copied to /approvals as the record.