Baixiang's 100-to-3 Flavor Funnel: Why Product-Only Innovation Resets Yearly

Baixiang Food's product strategy director Tian Lu revealed that R&D collects over 100 creative directions yearly from social comments, e-commerce reviews, and field research, yet only 2-3 flavors make it to national offline channels. The industry typically reads this as a low success rate, but the ratio stays constant year after year, indicating a structural limit rather than a filtering problem.

The Four-Stage Validation Funnel

Baixiang uses a rigorous four-step process before national rollout:

Feasibility proof : Lab feasibility ≠ production line compatibility. Cilantro noodles required whole freeze-dried cilantro stems that couldn't crush during transport, so a protective tray was added to packaging.

Internal screening : High-frequency needs feed innovation contests and cross-department brainstorms; promising directions go to blind tests and e-commerce pre-sales, letting actual orders dictate batch size.

External consumer testing : R&D VP Xu Qing insists every new product undergoes external consumer tests with 1,000+ samples before launch.

Online small-batch launch : Track repurchase rate, user reviews, and organic sharing. Only after passing these does a product enter offline distribution.

Crucially, rejection happens after validation — make, test, sell, then decide. This prevents meeting-room politics from overriding evidence.

Three Decision Types, Only One Door Open

User feedback (demands, usage hacks, corrections) enters the system but only triggers product decisions (recipe, spec, process changes). Examples:

Cilantro noodles: green dough + 5 whole freeze-dried cilantro + tray.

"Duoban" weight correction: 70g → 110-120g, packaging halted and replaced within a month.

Regional flavors (Wuhan hot-dry noodles, Lanzhou beef noodles) reverse-engineered into parameterized sauce packs (sesame paste viscosity, alkali ratio, pickle crunch) reusable across SKUs.

Service decisions (how to handle a specific user's hesitation, complaint, or churn) and resource allocation decisions (shifting budget from promotion to delivery improvements) never change. Feedback is categorized by product, not by person; the same team that sees a comment cannot see who said it or their history.

Parameterization as Reusable Asset

Value innovation requires both customer value and sustainable delivery. Baixiang achieves the latter by translating "tasty/authentic" into measurable parameters: sesame paste viscosity, alkali-to-water ratio, chicken breed & growth cycle, equipment settings. These parameters are reusable assets — a 50g half-tendon sauce pack serves three regional SKUs at three price points. However, innovation stops at product form; purchase, usage, and service paths remain identical to legacy products.

Online as Testbed, Offline as Commitment

E-commerce head "Shark" notes online shelves are infinite, low-cost experiment spaces; offline shelves are finite, high-cost commitments. Online tests (repurchase, review gap, organic sharing) de-risk offline placement. Distributors receive a validated result, not a gamble.

Transaction Signals vs. Interaction Signals

Two signal streams return after a cycle:

Transaction signals (repurchase, returns, negative reviews, sell-through) → improve the product.

Interaction signals (what a person asked, hesitated on, why they left) → improve the relationship.

Baixiang captures the first stream well (Cilantro 2.0 cut freeze-dry cost, adjusted price band, became a perennial SKU). The second stream is lost: product managers see comments but not commenters; e-commerce teams feed supply chain by SKU, not by user. Consequently, each year restarts from public chatter rather than from known relationships.

Conclusion: Goods Assets Remain, People Assets Vanish

After one cycle Baixiang retains reusable parameters, a repeatable process, and a stabler supply chain — all "goods assets." No "people assets" (identifiable, re-engageable users) remain. The funnel ratio is structurally fixed because only one of three decision doors opens. True demand-driven operation requires every piece of feedback to find its matching door: product, service, or resource allocation.