12 KiB
首席增长官 (Chief Growth Officer) — Recommended Product Roadmap
A risk-sequenced roadmap. The organizing principle is not "what to build next" but "what to prove next." Each phase exists to retire a specific risk and is gated on evidence before the next phase unlocks.
Guiding principles
These are the design decisions that shape everything below. They are deliberate departures from the original plan.
- Sequence by risk, not by feature. The original plan built outward (1 engine → 5 engines → industry infrastructure). The three things that can actually kill this company are data access, willingness-to-pay at a workable price, and whether founders trust and act on AI recommendations. The roadmap retires those first.
- The wedge must stand alone. Phase 1 has to be a gross-margin-positive business on its own economics. It cannot be a loss-leader that only works if a heroic upsell assumption holds.
- Depth before breadth. Do not launch five engines. Add them one at a time, starting with the engine that compounds the existing data moat, and only after multi-engine synergy is demonstrated rather than asserted.
- Earn the "decision-maker" claim; start as decision-support. Position as a co-pilot first. Move up the autonomy ladder only as accuracy and trust are proven. Selling autonomous strategic decisions on day one creates a trust and liability bar the product can't clear yet.
- The real moat is private-data lock-in plus the strategic-intent filter. Build deliberately toward those. Treat the knowledge graph as a depth play in one vertical, not a defensibility claim against platforms that own the raw data.
The trust ladder (runs across all phases)
The product climbs this ladder over time. Pricing power and the "CGO" positioning are earned by moving up it, not asserted at the start.
| Rung | What the AI does | Phase |
|---|---|---|
| 1. Inform | Surfaces signals, ranks pain points, flags anomalies | Phase 0–1 |
| 2. Recommend | Proposes specific actions with reasoning and confidence | Phase 1 |
| 3. Draft | Produces ready-to-use artifacts (content, plans) the human approves | Phase 2 |
| 4. Orchestrate | Coordinates multi-step workflows across functions, human-in-the-loop | Phase 2–3 |
| 5. Operate | Executes within guardrails, human-on-the-loop | Phase 3+ |
Phase 0 — Validation & De-risking
Months 0–4 · Thesis: "Earn the right to build the platform."
This phase did not exist in the original plan, and its absence is the single biggest gap. The goal is to spend as little engineering effort as possible while proving the assumptions the whole business rests on.
What to do
- Resolve the data question before anything else. Map exactly how competitor review and social data can be obtained legally and durably: official platform APIs, licensed third-party data providers (e.g. established e-commerce/social analytics vendors), direct data partnerships, and what is realistically off-limits via scraping. This answer determines the entire product surface. Treat it as a gating investigation, not a footnote.
- Run a concierge (done-with-you) version of the Product Innovation Engine. Deliver the insight output manually, AI-assisted behind the scenes, to 5–10 design-partner beauty brands. Charge real money for it. The point is to learn whether founders act on the recommendations and whether they'll pay, not to ship software.
- Validate the strategic-intent filter. Confirm that parameterizing 战略意图 (e.g. "premium ingredient-led" vs "value/price-led") meaningfully changes the output and that customers perceive and value the difference. This is the most defensible idea in the original plan; prove it early.
- Seed the knowledge graph in ONE narrow sub-category. Not all of beauty. Pick something like skincare serums or sunscreen and build real depth with domain experts.
- PIPL / data-handling readiness check for the eventual private-data ingestion (客服对话, SOV, loss data).
Exit gates (must hit to fund Phase 1)
- A validated, repeatable, legally defensible data pipeline for the core radar.
- ≥ 5 paying design partners; documented evidence that founders changed a real decision based on the output.
- Clear signal on willingness-to-pay and at what price point.
Risks retired
Data access feasibility · willingness-to-pay · whether the core insight is actionable.
Phase 1 — The Wedge: Product Innovation Engine as a standalone business
Months 4–12 · Thesis: "One product that pays for itself."
Productize what worked in Phase 0. This is the customer entry point and must be a real business on its own, independent of any future upsell.
What to build
- The SaaS workbench: 战略配置中心 (strategic config / the intent filter), 外部市场雷达 (external market radar), and the AI决策工作台 (decision workbench) — positioned at rungs 1–2 of the trust ladder (inform + recommend).
- The professional tier's private-data diagnostics (本品体验诊断, 流失归因, 区域/渠道预警) — this is where the durable lock-in begins, so prioritize a clean, low-friction data-connection experience.
Fixes to carry in from the critique
- Reposition as a co-pilot, not an autonomous CGO. "增长副驾" / decision-support framing. The "never-quits CGO" vision is the destination, not the day-one promise.
- Re-price to match value. The 8,800元/year entry price anchors the product as a cheap data tool and undercuts the entire positioning. Raise the floor, or move to a value/usage-aligned model, so the price signals "strategic system" rather than "SaaS trinket." Phase 1 must be gross-margin positive at this price with realistic CAC.
- Fix the tier logic. The flagship workbench query ("where are we weaker than competitor X?") requires the customer's own data, which the public-only basic tier lacks. Re-draw the basic/pro line so the headline feature isn't hollow in the tier most people buy — e.g. make a light private-data connection part of the entry experience.
GTM (absent from the original — build it here)
- Founder-led sales + design-partner referrals as the initial motion.
- The founder's growth methodology as content marketing into the beauty-founder community — this doubles as the corpus that differentiates the AI.
- Define and instrument CAC and payback target explicitly. A high-touch onboarding (config wizard + private-data integration) against a low price is the fastest way to negative unit economics; price and motion must be designed together.
Exit gates (must hit to fund Phase 2)
- Logo retention and net revenue retention above target (set concrete thresholds with the team).
- CAC payback under ~12 months on the entry product alone.
- Positive gross margin on Phase 1 in isolation.
- Documented % of customers who repeatedly act on recommendations (the leading indicator of expansion appetite).
Risks retired
Standalone unit economics · go-to-market repeatability · early trust.
Phase 2 — Depth, then ONE adjacent engine
Months 12–24 · Thesis: "Prove synergy with two engines before claiming a five-engine moat."
This is the largest strategic departure from the original. Do not launch four engines at once. That is five separate hard products for a company that has shipped one. Instead:
What to build, in order
- Go deeper in the wedge first. Expand the knowledge graph to more sub-categories; raise recommendation accuracy and confidence calibration. Depth compounds the moat faster than breadth.
- Add exactly one engine: Content Operations (内容运营). It is the natural second engine because it directly consumes product-innovation insights — this is the exact synergy the original plan used as its showcase example (insight → "清爽不粘腻" content strategy). It's also where beauty brands spend heavily and where AI generation has real leverage.
- Build the orchestration layer for real — and measure the synergy. Prove that two engines together produce more value (retention, expansion, outcomes) than two engines bought separately. This is the empirical test of the "core moat" the original plan merely asserted. Climb to trust-ladder rung 3 (draft).
Pricing & expansion (now earned, not assumed)
- Model conversion from single-engine to multi-engine as an explicit, measured cohort variable — not the original "vast majority will convert" assumption. Track it; price the bundle off observed synergy value.
- Avoid the original 25–40x price cliff between tiers. Build a smoother expansion path so growth within an account is a series of natural steps, not one improbable leap.
Sequencing the remaining engines
Decide Ads vs User Ops vs Full-chain Ops as engine #3 based on data — which adjacency your Phase-2 customers actually pull you toward, and which compounds the data you already hold. Add them one at a time, each behind its own value gate.
Exit gates (must hit to fund Phase 3)
- Demonstrated, quantified two-engine synergy (cohort with both engines materially out-retains/out-expands single-engine cohorts).
- A defensible position (share, retention, reference base) within the beauty vertical.
- Expansion revenue that is observed and repeatable, not modeled on assumption.
Risks retired
Whether synergy is real and monetizable · expansion economics · orchestration as a genuine capability.
Phase 3 — Platform & the "全域" vision
Months 24–36+ · Thesis: "Become the system of record for growth decisions in beauty — then, and only then, expand the surface."
Only now is the original "首席增长官全域版" vision credible, because the company has earned a defensible beachhead.
What to build
- Cross-platform data unification (天猫 / 京东 / 抖音 / 私域 / 线下) on a single data foundation — the real cross-silo moat, which the original positioned as a day-one barrier but is actually a Phase-3 capability.
- The full orchestrator and higher-autonomy operation (trust-ladder rungs 4–5, human-on-the-loop).
- Complete the engine matrix as data and demand justify each one.
Treat the "industry data services" second curve with caution
The original plan's idea of selling anonymized insights to OEMs, raw-material suppliers, and investors is a different business with real channel-conflict risk: your brand customers may not want their behavior, even anonymized, informing competitors or suppliers. Validate customer consent and trust implications before pursuing it; it can quietly undermine the core product's lock-in. Park it as an option, not a committed milestone.
Possible second vertical
A second industry (e.g. a different consumer category) is a Phase-3 option, gated on the beauty knowledge graph and playbook being genuinely repeatable rather than founder-dependent.
What to keep, cut, and defer from the original plan
Keep
- The wedge / land-and-expand shape.
- Beauty-first single-vertical focus.
- The strategic-intent (战略意图) filter — the strongest idea in the plan; make it the centerpiece of the differentiation story.
Cut or recast
- The day-one "autonomous CGO that never quits" claim → recast as the top of the trust ladder, earned over time.
- The five-engines-at-once Phase 2 → one engine at a time, gated on proven synergy.
- The 85折 bundle math error and the 25–40x tier cliff → rebuild pricing as a smooth, value-anchored expansion path.
Defer
- Cross-platform unification → Phase 3 (it was framed as an early moat; it isn't).
- Industry-data second revenue curve → optional Phase 3+, pending channel-conflict validation.
The decision questions that actually matter
The original plan's open questions (is pricing right, is the cadence right, is beauty confirmed) are second-order. These are the ones that gate the business:
- Can we obtain the radar's data legally, at scale, and durably? (Gates Phase 0.)
- Will founders pay a value-anchored price and actually act on the output? (Gates Phase 1.)
- Does multi-engine synergy create enough measured value to justify expansion pricing? (Gates Phase 2.)
- What is our CAC and payback, and does the sales motion repeat? (Gates everything.)