Context
Aladdin JSC runs four restaurant brands across more than 50 outlets. Ads bring guests into Messenger; when a guest books a table in the chat, it is logged as a purchase. Once the guest arrives, that booking is checked against the restaurant's records and sales, and marketing is measured on the revenue these guests bring against a monthly target.
The problem
From January to June, marketing revenue at Long Wang reached only 64–85% of target, and leads fell short as well. The real question was where the next budget should go: which promotion, which outlet and which creative actually brought in paying guests.
What I changed
Measure the whole funnel, by promotion.
Every campaign name carries its promotion, so data pulled from Meta can be read promotion by promotion and outlet by outlet: lead cost, purchases, cost per purchase and close rate, not cost per purchase alone.
Plan the month by promotion.
Each promotion has its own target cost per purchase. Starting from the monthly budget the company sets, I split spend by promotion and outlet, with more for outlets whose sales are lagging, and forecast the purchases each split should bring, allowing for cost per purchase to rise as a promotion scales.
Watch it daily.
A Google Sheets tracker pulls Meta data every 5 minutes and checks each outlet group against its daily purchase and cost-per-purchase targets, next to offline bookings and revenue. Budget moves mid-month, not at month end.
Let the numbers brief the creative.
Angles and promotions that brought paying guests got more variations; creators were kept or cut on cost per purchase.
Fix the plumbing.
Website tracking, Google Maps listings and fanpages set up properly; Messenger conversations matched to ad IDs to separate ad-driven guests from organic ones; and purchases reported in Ads Manager checked against real orders in Pancake, since conversions sent through the API are often missing, wrong or credited to the wrong ad.
The result
Jul–Sep against Jan–Jun (monthly averages, February left out because ads were largely off): purchases ×2.2, cost per purchase −44%, spend +22%. Marketing revenue reached 138%, 122% and 114% of target, and at Long Wang and Tian Long the share of Messenger leads who booked rose from 20–28% to 29–34%.
Long Wang: from 64–85% of target to 114–138%
Marketing revenue vs monthly target, Jan–Sep 2026. Blue: the three months after the new approach.
Show the numbers
| Month | % of target |
|---|---|
| Jan | 71% |
| Feb | 73% |
| Mar | 85% |
| Apr | 73% |
| May | 64% |
| Jun | 79% |
| Jul | 138% |
| Aug | 122% |
| Sep | 114% |
Long Wang: cost per purchase 44% lower
Monthly cost per purchase as an index: Jan–Jun average (excl. Feb) = 100. Lower is better. Blue: the three months after the new approach.
Show the numbers
| Month | Index (Jan–Jun = 100) |
|---|---|
| Jan | 113 |
| Feb | 92 |
| Mar | 94 |
| Apr | 105 |
| May | 112 |
| Jun | 83 |
| Jul | 46 |
| Aug | 61 |
| Sep | 64 |
Tian Long
The same approach ran at Tian Long from August: 124% and 115% of target in Aug and Sep, with cost per purchase 44–48% below target.
My role
A team result: sales, outlets and promotions all played a part. My scope: Meta Ads planning and buying, budget split, promotion analysis, creative briefs, tracking setup and reporting.