Aligning a Million-Person Organization: 7 Strategic & Organizational Mechanisms

Large Organizations Inevitably Lose Focus — Not a Capability Issue

Communication paths grow as n(n-1)/2: 100 people = 4,950 links; 1,000 = 500,000; 10,000 = 50 million; 1,000,000 = 500 billion. Headcount grows linearly, coordination cost grows quadratically. Brooks' Law: adding people to a late project makes it later.

Three inevitable phenomena appear:

Goal dilution: Group strategy "improve fulfillment experience" becomes "reduce complaint rate" at business group, "cut response time 10%" at department, "clear daily tickets" at team. Four layers later, no one remembers the original why.

Resource spreading: Everyone carries three tasks, each gets one-third attention. Output isn't three units; it's near one.

Meetings turn into debates: Sales says big promo must be first, marketing says conversion can't wait, supply chain says inventory turnover is vital. Every demand is "must do." Prioritization degrades to loudest voice, highest rank, or boss's pet.

This is entropy increase — it happens daily without external push. Only two counters: directional convergence via strategic mechanism, structural convergence via organizational mechanism. The former teaches the organization to do the right things; the latter teaches it to do things right.

Strategic Mechanism: First Make a Million People "Do the Right Things"

Called "strategic mechanism" internally. It's one job: put limited resources, with an investor's lens, precisely onto the most important strategic projects and high-priority demands. Seven steps in fixed order.

Step 1: Plan First, Then Focus

Resources are sprinkled like pepper because strategy stays at slogan level. "Strengthen middle-platform capability" or "improve user experience" cannot guide a single concrete allocation.

Two mature frameworks exist:

Huawei's "Five Looks, Three Determines" — Look at industry trend, market, competition, self, opportunity; then Determine strategic control points, targets, strategies. The case company trimmed it to "Five Looks, Four Determines" : Five Looks (direction, market, customer, capability, competitor) + Four Determines (direction, initiatives, targets, owner). The extra "owner" seeds Step 5: every strategic initiative has an owner from birth.

OKR — O is direction, KR is initiative/playbook.

They use a "Strategic House." Positioning: "tech & service enterprise based on supply chain." The house breaks that into floors: positioning at top, then targets, then initiatives. One picture shows three curves (see image):

First curve: Centralized e-commerce core business

Second curve: Decentralized e-commerce

Third curve: Overseas expansion, etc.

One Strategic House, three curves. Resource destinations are visible at a glance. Planning's sole purpose: give resource allocation evidence — decide by Strategic House, not by who shouts loudest.

Framework choice is not the point. Tools abound: BCG Matrix, GE 9-cell, SWOT, Ansoff, BLM, McKinsey 7S, Balanced Scorecard… each consultancy brings new vocabulary. None is the essence. Essence is picking one that reaches consensus and keeps running in your organization, then executing ruthlessly. Tencent, Huawei, JD each have different playbooks; the one your organization understands is the good one.

Hard constraint for this step: strategic initiatives ≤ 5 . More than five means no strategy.

Step 2: Consolidate Similar Items, Identify "Strategic Projects"

Strategy yields 5 initiatives, but 300 projects run simultaneously. Need a consolidation step.

First, classify projects into three types:

Operational (create value now) — directly tied to business metrics: GMV, conversion, fulfillment cost, gross margin.

Product Research (layout for users & future) — future capability, product incubation, architecture upgrade, platformization, tech debt repayment.

Maintenance (risk control, sustain experience) — stability, security compliance, daily ops & bug fixes.

Then define "strategic project" — meets at least two of three criteria:

Cross-functional: spans functions/departments, single business unit can't push alone.

Complex: long cycle, high uncertainty, needs phased validation.

Strategic: directly carries one of the strategic initiatives above.

Three consolidation scenarios (most skipped, most value):

Same goal: Dept A and Dept B each have "improve search conversion" → merge.

Same approach: Three teams building "user profiling" → merge into one platform project.

Overlapping resources: All fighting for same algorithm engineers → must merge or sequence.

Just this "consolidate similar items" — one full project inventory per year — typically frees 10–15% redundant investment. No new headcount; squeezed from waste.

Boundary: strategic projects are not business units' own projects. Business projects = "my KPI"; strategic projects = "organization's #1 engineering" — they get resource priority, escalation authority, and #1 staffing. Hard constraint: strategic projects ≤ 10 .

Step 3: Demand Management — Shift Perspective Before Rules

Most painful layer: requires R&D teams to change self-definition.

Face reality: business side inherently lacks cost awareness. Not an attitude problem; mechanism vacuum. Traffic, marketing, procurement are measured on growth, conversion, turnover; R&D headcount is an "invisible bill" to them. If you don't show the bill, they'll never calculate it.

So demand management step one isn't scheduling — it's perspective shift, from "order-taker" to "investor":

Reaction to demand: Order-taker → schedule immediately. Investor → check return first, then schedule.

Prioritization basis: Order-taker → who screams loudest, who ranks highest. Investor → strategic weight × business value ÷ investment cost.

Cost ownership: Order-taker → R&D's own cost problem. Investor → business & R&D share.

Measurement: Order-taker → how many demands delivered. Investor → how much business result generated.

With perspective set, rules follow. Demand grading not by urgency but by strategic level:

P0: CXO-level, group-strategic — if not done, strategy fails.

P1: Business-strategic — impacts annual business targets.

P2: First-level department optimization demands.

Demand management governs the entire R&D resource pool, not one business line or team. When resources truly run short? Process re-ranks openly — not private backdoor jumps — and publishes the reasoning. Transparent ranking is itself the best management mechanism.

Step 4: Resource Management — "Dividing the Pork" Needs Rules Too

Rules done, now the most feeling-hurting step: how to cut the cake.

First cut uses 721:

Business demands: 70% — support current year operating targets.

Product Research layout: 20% — architecture upgrade, platformization, tech debt, capability building.

Maintenance: 10% — stability, security compliance, daily ops.

Second cut inside the 70%: by business value, not headcount, not history, not loudness.

Clean corollary: want more resources? Trade business value. Show higher YoY growth, GMV growth, conversion lift, fulfillment cost drop — any metric proving "one more dollar still grows."

Accompanying action: resource billing — each quarter give business side a bill: your demands consumed how many R&D person-days, equivalent cost, what results produced. Purpose: grow cost & investor mindset in business.

721 is guardrail, not railroad. An org in tech-debt explosion forcing product research to 20% accelerates collapse. Right way: set at year-start per org state, e.g., 65/30/5, no arbitrary mid-year breach; adjustment requires collective resolution. Example: a category business group plans 50% growth logic, signs military pledge, ties to appraisal — can fully get matching people/money/resources!

Step 5: From Strategy to Demand, Every Layer Has a "No.1 Position"

Resources allocated, who owns results?

Simple answer: from strategy to project to demand, every layer sets a No.1 position. Business No.1, Product No.1, Dev No.1, Test No.1 — four roles named in responsibility matrix, published externally.

No.1 position means three things: decision authority (within mandate, direct call), resource scheduling authority (can pull cross-functional resources, counterparts must respond), result accountability (if this fails, first person asked is me).

Metrics also layered:

KPI (floor, ≤3 items) — too many KPIs = no focus. Note: suits stable industries; tech companies chasing growth/innovation rarely use.

OKR (stretch) — carries strategic projects' ambitious goals; 70% completion counts as success.

Most revealing metric: strategic project resource guarantee rate — promised P0 resources vs. actual delivered. Promised 40%, actual 20% → mechanism leaks at first execution layer.

Step 6: Execution Transparency — Agile Is Rhythm, Transparency Is Leverage

Rules set, how does delivery run?

Use agile iterations for rhythm (typically two-week sprints), but key isn't stand-ups — it's publicity:

Demand pool open: what this sprint does, why these, who requested. Value-stream oriented.

Progress & commitments open: what was promised, current status, any deviation.

Risks open: where stuck, who needed for support, estimated impact scope.

Why is transparency a lever? Because most org coordination cost comes from information asymmetry causing repeated communication and guessing. If a business stakeholder can self-serve see demand rank, why it's there, they don't need to ask three times a week; R&D doesn't explain three times. Both sides save time.

Step 7: Tools Come Last

Biggest pitfall: reversing the order.

Pain point appears, first reflex: "buy a system." System lives six months, fields half-filled, demands still sorted in WeChat groups because who-goes-first, who-decides rules were never clarified. System just solidifies chaos.

Tool's correct place is the finish line, not the starting line. When strategy, projects, demands, resources, responsibilities are crystal clear, tools merely freeze their relationships into data flows. Then selection resistance lowest, adoption highest, because everyone's behavior already runs smoothly in the mechanism; system just saves a few forms.

Organizational Mechanism: Then Make a Million People "Do Things Right"

Strategic mechanism solves "where to hit." After group, sub-group, business groups, and middle/back offices complete strategic consensus and breakdown, direction is set. But a harder question remains: what force makes a million people move in that same direction?

That's what organizational mechanism answers. This company's organizational management goal is crystal clear: build an efficient, agile, customer-driven group ecosystem platform. Three layers:

Organizational Form (rights/responsibilities/benefits positioning & functional config): Strategic Control Group + Modular Organizational Units

Organizational Construction (core orientation for building org): Strategy-driven, clear goals, functional integration, streamlined hierarchy

Organizational Mechanism (mechanisms for efficient org operation): Decision mechanism, collaboration mechanism, innovation mechanism

Most meat is in organizational form.

"Strategic Control Group" answers what HQ controls: HQ sets direction, allocates resources, appoints No.1 positions — does NOT replace business in operating decisions. If reversed (HQ controls both direction and operations), organization grows an extra sluggish brain.

"Modular Organizational Units" answers how units combine: each unit has clear rights/responsibilities/benefits, standard interfaces, can be reassembled as business changes — not "one adjustment breaks bones." Modularity's value isn't that each block is huge; it's that blocks can be disassembled, recombined, replaced. True agility at massive scale isn't about approval speed; it's about whether the organization itself can be reassembled.

Beyond three layers, one concrete landing action deserves separate mention: Strategic BP & Strategic Meeting.

Path:

Strategic BP (Business Planning) → Strategic Meeting (facilitate consensus) → Consensus outputs targets / battlefields / must-win battles → Identify strategic projects & generals/commanders → Assign project managers to track.

This chain's value: turns "consensus" from adjective into action sequence. Consensus forced by process, not by one meeting's cheer: write targets, draw battlefields, pick must-win battles, name generals/commanders, assign project managers. Five actions done, consensus truly formed.

And "generals/commanders + project managers" pairing exactly meshes with Step 5's "No.1 position mechanism": generals/commanders own winning; project managers own rhythm. One owns outcome, one owns process.

Back to Ren Zhengfei. He has a more vivid metaphor:

Running a venture is like dragon dancing: head must rise — that's direction; body must move with power — whole dragon dances alive.

Strategic mechanism is the head — decides where dragon flies. Organizational mechanism is the body — decides if dragon can move. For a million people to march in step, head and body need a transmission mechanism; head straining alone is useless.

How to Know You're Doing It Right

Three weekly self-checks:

If you cannot say "this year we only do three things" and actually only do those three, you haven't focused. Strategic initiatives >5, strategic projects >10 — same disease.

If No.1 positions this week didn't personally kill at least one project and didn't personally re-prioritize demands once, mechanism is idling. Focus is action, not slogan.

If one org adjustment takes >1 month to land, or every adjustment breaks bones, it's still cement, not Lego. A million-person org's agility shows first in structural reorganizability.

Conclusion: The Larger the Plate, the More You Need Mechanisms

Back to the opening question: tens of thousands scale, how to land mechanisms?

Answer may be counter-intuitive: precisely because the plate is large, you must rely on mechanisms. When small, trust, tacit understanding, shoulder-taps work; when large, all fail — only mechanisms remain.

Strategic & organizational mechanism upgrade — a million people can succeed; you can too. Welcome to exchange.