Mayora
Management Trainee · 2024–2025
Mayora is an Indonesian FMCG multinational — Kopiko, Danisa, Roma are among the brands. The China team operates across traditional retail and O2O instant retail platforms: Meituan Flash Sale, JD Daojia, Taobao Flash Sale.
I joined as a management trainee, rotating across O2O e-commerce, marketing, sales, and NTM. Most of my substantive work landed in O2O instant retail — a channel the company was still figuring out. There was no existing SOP, no historical performance data to benchmark against. For a lot of what I worked on, I was the first person to do it.
CNY Campaign — Meituan Flash Sale
This was Mayora's first-ever instant retail campaign. Two product lines ran in parallel: gift boxes (the volume driver, tied to CNY gifting behavior) and new products (category expansion, using the holiday as a launch window). Total approved budget: ¥230K across RTB ad placements and platform vouchers.
My line manager set the macro direction. I handled the day-to-day: audience targeting strategy, budget allocation, monitoring, optimization, and daily reporting. No prior Meituan campaign to reference, no playbook. Every decision had to be reasoned from first principles.
Before launch, I mapped four targeting hypotheses based on behavioral logic rather than guesswork:
Cookie and biscuit buyers — the highest category affinity segment. Lowest conversion barrier, our core base. Gift-giving and care audiences — CNY gifting occasion matched with gift box product. Direct intent match for the send-as-gift use case. Existing customers — brand familiarity drives repurchase during CNY, and re-engagement cost is low. Dairy transaction buyers — adjacent category hypothesis: biscuits and milk as a common consumption pairing. This was always a test.
The dairy hypothesis turned out to be wrong. Adjacent category logic sounded reasonable but didn't hold up in practice — the incremental ROI was insufficient. We cut it in Week 1 and reallocated the budget to segments that were already outperforming.
The campaign ran two structured optimization rounds. Round 1 (end of Week 1): shifted gift box targeting from broad to precision insight plans with CPM bidding, expanded gift box cities from 3 to 5 (added Nanjing, Guangzhou), cut the dairy audience, and increased budget allocation to gift-giving segments. Round 2 (sprint phase, Jan 13–17): scaled gift box GMV daily caps from ¥500 to ¥800 to ¥1,000 to ¥3,000, extended ad hours from afternoon-only to full day (07:00–24:00), and continued pruning low-performing segments.
A few clarifications on the numbers. The 600% for new products is growth against the January pre-campaign daily average — not year-over-year. I validated this against 2024 same-period data and isolated campaign-attributed sales from organic to control for seasonality. The gift box ROI of 7.22 is the full-window comprehensive figure. The daily direct ROI was stable in the 4–5 range; the peak reflects the cumulative effect as compounding audiences built up over the campaign window.
Three audience segments drove the majority of gift box performance: existing customers (ROI 38.7), cookie transaction buyers (ROI 56.4), and gift-giving and care audiences (ROI 7.3). The hypothesis-driven approach — define each segment's logic upfront, validate mid-campaign, cut what doesn't work — was the right framework.
What I'd do differently: pre-define spend ramp criteria before launch instead of making those calls on intuition. Early in the campaign, I held back on scaling high-performing audiences — intuitive caution about spending too fast. By the time I tried to ramp aggressively in the final sprint, the campaign window was closing and I never found the ROI ceiling. The threshold for "when to ramp" should have been set in advance as a data trigger, not a judgment call in the moment.
Daily O2O Operations
Beyond the CNY project, I was running day-to-day operations across Meituan Flash Sale, JD Daojia, and Taobao Flash Sale for all Mayora brands — monitoring sales and conversion data daily, coordinating SKU uploads and listing updates with the TP partner, and writing weekly data reports for my LM.
Built the product matrix for key categories — mapping daily small-pack SKUs, gift boxes, and seasonal items to channel strategy, and drafting the launch cadence and promotional configuration for seasonal windows. Contributed to the Y25 annual planning: Y24 Meituan Flash Sale revenue was ~¥19.8M, Y25 target was ¥30M (+51.5%), with a ¥1.61M precision marketing budget at a 5% expense ratio.
Reflection
Running a campaign with no historical data and no SOP is a different kind of pressure from having a playbook to execute. Every decision — which cities to launch in, which audiences to test, when to scale — had to be reasoned from first principles rather than precedent.
What I found useful: treating every targeting decision as a hypothesis with an explicit validation plan before launch, not after. The dairy audience cut would have been cleaner if I'd pre-specified "cut if Day 7 incremental ROI is below X" rather than "we'll see how it goes." The SOP I documented afterward — campaign rhythm to resource confirmation to execution to performance review — captures the process. But the more important artifact is the hypothesis log: what you assumed before launch, and whether the data confirmed or refuted it.