Water softeners are a high-ticket, high-repeat niche that refused to grow. Years of conventional research kept returning ideas the client already knew — because framed surveys only let consumers answer inside the brand's own assumptions. Whatever lies outside the frame never surfaces.
Every step reproducible
- 01TrendAI
Collect 100k authentic UGC posts
TrendAI captured ~100k water-softener posts across Xiaohongshu, Douyin and e-commerce reviews — no questions asked, just listening first.
- 02ezData
Clean to 50k quality texts
ezData deduplicated and de-spammed the corpus down to 50k analyzable texts.
- 03ezSummary
Co-create a 4-level coding frame
A co-created 1–4 level coding frame let AI quantify every verbatim — each tag auditable — at 86–92% parsing accuracy versus the ~60% industry norm.
- 04ezData
Crosstab segments × metrics
Segment-by-metric crosstabs peeled down to audiences, scenarios and purchase barriers: who buys, where they use it, and exactly where the funnel breaks.
- 05
Reverse-engineer the strategy
The voices pointed straight at the fix: "filter hassle" was the core barrier, driving a redesign from 6-month to ~2-year filter life.
- 86–92%VOC parsing accuracy (industry norm ~60%)
- 10万→5万100k posts collected, 50k quality texts kept
- 半年→2年Core barrier fixed: filter life redesigned
- 新场景Pet & plant scenarios the brand never anticipated
- 可复用Coding frame became a reusable, priced data asset
"Conventional research returned what we already knew. This time, the data grew things we never thought of."