MeituanStrategy Analyst
Beijing · 2025.05 — nowCompetitive research across the service chainThe fallback strategy it identified is now piloted on Meituan Waimai and Hotel & Travel.
8 / 16sub-businesses
Details
As LLM customer service scaled (coverage reached 50% across all businesses by the end of 2025), studied Ctrip, Alibaba, ByteDance and Ele.me across the portal → AI service → human service chain.
Responsible for 8 of 16 sub-businesses; led alignment, interview design, expert sourcing and the benchmark database.
Identified the fallback of routing unresolved AI cases to a human at the survey step; now piloted on Meituan Waimai and Hotel & Travel, balancing AI coverage with experience.
New product shapes for AI serviceDefined the two shapes of human oversight and their cost logic; cycling and power-bank shipped on cost target.
Details
Mapped supervision, quality inspection and agent assist across competitors; defined human oversight as pre-handoff (live tagging for data flywheel, human-in-the-loop for multimodal gaps) and post-handoff (one person supervising many AI sessions via an inspection model) — two shapes, each with its cost logic.
Benchmarked coverage and closure rates to guide adoption; cycling and power-bank teams shipped it on cost target.
Service resourcing structureVisited two suppliers; the crowdsourcing study supported doubling the managed cloud-agent pool.
2research reports
Details
As AI reshapes staffing and outsourcing, owned the resourcing agenda: the Q3 2025 industry resourcing survey, BPO market research and crowdsourced-operations research.
Framework from demand shifts → BPO competition → responses → Meituan's gaps; visited two BPO suppliers, surfaced supplier tiering (Alibaba) and order splitting (Pinduoduo); delivered the B-team report.
Conclusions back the move to BPO 2.0 and cloud agents; the selection-and-retention study helped double the managed cloud-agent pool.
Service strategy for high-value usersDrove proactive-service iterations on Waimai and Hotel & Travel, saving six figures monthly.
7businesses, 5 companies
Details
Self-initiated under pressure on high-value experience goals; built a tiering — entitlements — validation framework, benchmarked 7 businesses at 5 companies (88VIP, JD Plus and others) on four dimensions, and tested every product including Meituan's own.
The report spread through KPs and drove proactive-service iterations on Waimai and Hotel & Travel, such as early outreach on cooking, pickup and delivery anomalies.
Testing exposed rule conflicts behind poor resolution for high-value users; the fixes lifted AI resolution for high-value users and save six figures of labour cost monthly.
Team AI workflowsMeeting notes went from one person-day to ten minutes, spreading to the recruiting team.
5people use it daily
Details
With Claude Code, turned interview-note processing, research gathering, competitor monitoring, database upkeep, report templates and test frameworks into reusable team Skills.
Meeting notes went from one person-day to the ten-minute level; five people use it daily and it spread to the recruiting team.
