01 · Industry Research Methodology
Research the industry before you research any stock — this is the sell-side analyst's first lesson, and the one retail investors most often skip. This article lays out a five-step framework for industry research, the top-down and bottom-up paths, industry classification conventions, criteria for judging a "good industry," a source checklist, and common pitfalls — all in one place, ready to use.
1. What Problem Does Industry Research Solve
Industry research answers not "which stock will rise," but four more fundamental questions:
| Question | Dimension | Typical Sub-questions |
|---|---|---|
| How big is the industry, and is it still growing? | Size & growth | What is the market size? Where is the ceiling? What's the 3-5 year growth rate? |
| Who is capturing the money? | Competitive landscape | Is concentration high? Who leads? Is the landscape improving or deteriorating? |
| On what basis does the industry earn money? | Industry chain & business model | Which segment keeps the profit? Asset-heavy or asset-light? |
| Where are we now? | Prosperity & cycle | Which stage of the cycle? Where is the penetration rate? What are the catalysts? |
⚠️ Without answering these four questions, your stock analysis is an anchorless boat
Skip these four questions and your stock analysis drifts anchorless: you may own a good company sitting on a shrinking industry; you see a fat gross margin but not the upstream price hike about to punch through it.
The output of industry research should compress into one sentence, e.g.:
- A certain baijiu (Chinese liquor) industry: stable size (volume down, price up), excellent landscape (CR5 above 70%), asset-light with high margins, prosperity driven by consumption and inventory cycles.
- A certain solar PV industry: large size but decelerating growth, deteriorating landscape (overcapacity, price war), capital-intensive, currently in capacity-clearing phase.
2. The Five-Step Framework
① Market Size and Growth (TAM Estimation)
TAM (Total Addressable Market): how big the market would be if every potential customer adopted the product.
| Method | Approach | Example |
|---|---|---|
| Top-down | Total population/volume × penetration rate × average spend | Electric two-wheeler market ≈ urban population × ownership rate × average price |
| Bottom-up | Sum of incumbent players' revenue + estimate of untapped market | Third-party payments ≈ sum of payment providers' revenue + unpenetrated merchant space |
| Benchmarking | Reference per-capita spending/penetration in mature overseas markets | Incremental headroom as China's per-capita insurance premiums converge to Japan/Korea levels |
Key points for estimation:
- Split size into "volume" and "price": volume follows penetration rate; price follows consumption upgrades or price wars.
- Multiply size by time: what is it at 3, 5, and 10 years out, and is growth shifting gears.
- Size is an "elasticity," not a constant: small changes in methodology can swing results by multiples, so treat market-size figures as ranges, not precise values.
② Competitive Landscape and Concentration
| Metric | Meaning | How to Read It |
|---|---|---|
| CR3 / CR5 / CR10 | Combined share of the top 3/5/10 firms by revenue | Higher = more concentrated; track year-over-year changes to see whether concentration rises or falls |
| HHI (Herfindahl-Hirschman Index) | Sum of squared market shares of each player | > 2500 highly concentrated; < 1500 fragmented (see Article 03) |
| Share change | Annual gain or loss in the leader's share | Sustained share gains = landscape converging toward the leader |
③ Industry Chain Position and Bargaining Power
- Is the industry positioned upstream, midstream, or downstream in the value chain? What is its bargaining power over suppliers and customers?
- Compare concentration up and down the chain: if the upstream is more concentrated and the downstream more fragmented, the midstream's profit pool gets squeezed from both ends most easily.
- Does this segment have pricing power? Can cost increases pass through smoothly? (See Article 02)
④ Business Model and Economics
| Trait | How to Judge | Investment Implication |
|---|---|---|
| Asset-light / asset-heavy | Fixed assets as % of total assets, capex intensity | Capital-heavy businesses have big profit swings but hard exits; strong cyclicality |
| High margin / low margin | Gross and net margin levels | High margins typical of brands/software/pharma; low margins typical of contract manufacturing/trade |
| Cyclical / growth / defensive | Whether earnings swing with the economy and prices | Determines valuation method and timing approach (see Article 04) |
| Cash flow quality | Does operating cash flow cover profit? | An industry-wide pattern of "growing revenue without cash" signals receivables/advance-funding models |
⑤ Sector Prosperity and Catalysts
- Is the industry in an upswing, downturn, or bottom zone? Which way are product prices, inventories, and operating rates moving?
- Build the 6-12 month catalyst list: policy implementation, new-product cycles, price hikes, exports, capacity clearing, etc.
- Catalysts signal "when to act"; the previous four steps establish "whether it's worth acting" — never reverse the order.
💡 Output of the five-step framework: the "one-page industry sheet"
The framework produces a single A4 "one-page industry sheet": size range, growth rate, concentration, chain position, one-line business model, prosperity position, and catalyst list. Fill this template for every new industry you study; keep it up for a year and you'll have your own industry database.
3. Top-down vs Bottom-up
| Dimension | Top-down | Bottom-up |
|---|---|---|
| Starting point | Macro → industry → company | Company → industry → macro |
| Logic | Pick the right "track" first, then pick stocks within it | Find "good companies" first, then examine their industry |
| Best for | Sector rotation and macro-driven markets | Deep single-stock work, bottom-up value investing |
| Weakness | Ignores company quality — right track, wrong stock | Ignores systemic industry risk (policy, cycles) |
| Typical use | Deciding "which sector to allocate to next" | Long-term tracking of a few companies' fundamentals |
- Mature practice combines both: screen 3-5 industries worth studying top-down, find the best companies within them bottom-up, then use a top-down lens to time entries.
- For ordinary investors, start bottom-up within the industries you know well, then gradually widen coverage — far more effective than chasing hot themes everywhere.
4. Industry Classification Conventions
Before doing industry research, learn how industries get "numbered." Three systems dominate:
| System | Maintainer | Features | Common Uses |
|---|---|---|---|
| SWS Industry Classification | Shenwan Hongyuan (China) | Level 1/2/3 hierarchy; de facto standard for A-share research, e.g. "SWS L1 – Electrical Equipment" | A-share sector quotes and industry indexes |
| CITIC Industry Classification | CITIC Securities (China) | Similar to SWS, ~30 level-1 industries | Sell-side research, industry indexes |
| GICS | S&P/MSCI (global) | Unified global standard, 11 sectors; standard for US-stock funds | US stocks, global allocation, ETF classification |
SWS level-1 industries (partial): agriculture/farming/fishery, food & beverage, pharmaceuticals & biotech, electronics, computers, electrical equipment, machinery, automobiles, nonferrous metals, coal, steel, petroleum & petrochemicals, basic chemicals, banks, non-bank financials, real estate, building materials, home appliances, light manufacturing, textiles & apparel, telecom, transportation, utilities, commerce & retail, social services, defense, media, environmental protection, beauty care, conglomerates.
Practical conventions:
- The same company may be classified differently across systems; always note which classification a cited figure uses.
- SWS level-3 industries are granular enough (e.g., "baijiu" is a level-3 industry under the level-1 Food & Beverage): use level 3 for single-stock research, level 1 for allocation decisions.
- Classifications get revised (e.g., "Beauty Care" was added recently), so don't be surprised by historical data breaks.
5. Criteria for Judging a "Good Industry"
Four criteria
| Criterion | Meaning | Traits of a Good Industry |
|---|---|---|
| Perpetual demand | Demand persists long-term and resists substitution | Basic needs: food, clothing, housing, transport, healthcare, energy, power |
| Stable landscape | Competition converges toward leaders | Stable oligopoly, few price wars, high entry barriers |
| Bargaining power | The industry is strong within its value chain | Can raise prices, pass through costs, no dependence on a single customer |
| Policy-friendly | Long-term policy support exists | Innovation encouraged, demand has policy backstops, not repeatedly targeted |
Counterexamples (same criteria, negative cases)
| Criterion | Counterexample Industry (illustrative) | Why It Fails |
|---|---|---|
| Perpetual demand | Traditional film | Demand fully substituted by digital technology; the industry went to zero |
| Stable landscape | Early home-appliance chains; recent solar polysilicon | After the landscape broke down, endless price wars crushed even the leaders' profits |
| Bargaining power | Small pharma vs. medical-insurance negotiations; contract manufacturers vs. big clients | Power sits with the counterparty; margins are persistently squeezed |
| Policy-friendly | Tutoring ("double reduction"), online games (license tightening) | One policy turn instantly rebuilds the industry's logic |
Judge with a "ten-year view": today's textbook "good industries" (baijiu, banks, appliances) were once "new tracks" ten or twenty years ago; today's hot sectors (AI, robotics) may not be on the list ten years from now. Good industries are filtered dynamically, not memorized.
⚠️ Counterintuitive: good industries are filtered dynamically, not memorized
Good industries are screened dynamically, not memorized. Today's textbook answers (baijiu, banks, appliances) were "new tracks" a decade or two ago; today's hottest sectors (AI, robotics) may drop off the list within ten years — so always judge with a "ten-year view," and never mistake today's celebrity sector for a permanently good industry.
6. Source Checklist for Researching an Industry
Ordered by priority, starting with free authoritative sources:
| Tier | Sources | What You Get |
|---|---|---|
| Official data | National Bureau of Statistics, MIIT, customs authorities | Output, output value, import/export, investment data |
| Industry bodies | Industry associations (auto, passenger-car, steel, fund associations, etc.), exchanges | Monthly/annual industry data, production/sales volumes, inventories |
| Corporate disclosures | Listed companies' annual/interim report "industry" sections; IPO prospectus industry chapters | Leaders' official framing and data on the landscape and market size |
| Broker research | In-depth industry reports (start with deep dives, not daily notes) | Size estimates, competitive landscape, industry-chain maps, earnings forecasts |
| Primary information | Conference minutes, expert calls, trade shows, company visits | Real signals on orders, prices, operating rates, expansion plans |
| Price data | Commodity price services, Bloomberg/Wind-type terminals, metals portals | Real-time and historical series of product prices, spreads, and inventories |
Three disciplines for gathering material:
- Official first, brokers second, self-media last: broker reports mostly copy official and association data anyway — trace to the original whenever possible.
- Let data drive conclusions; don't hunt data to justify conclusions: build the logical framework first, then test whether the data supports it; reversing the order is self-deception.
💀 Iron rule: let data drive conclusions; don't hunt data to justify conclusions
Use data to derive conclusions, not to decorate them. Framework first, then verify against data. If you fix the conclusion before opening the data, the data becomes mere evidence-hunting props — reversing this order is the most common cognitive trap in industry research.
- Build a "monthly routine," not a "look-it-up-when-needed" habit: refresh key industry data (prices, output, inventory, sales) on a fixed monthly cadence — trends matter far more than single points.
7. Common Pitfalls in Industry Research
Pitfall 1: Treating industry prosperity as a company thesis
A booming industry ≠ every company in it makes money. In 2020-2021 EV-sector prosperity was extreme, yet some companies still over-expanded and watched gross margins slide quarter after quarter. Confirm the company's relative position within the industry first, then apply the industry logic — never substitute industry logic for company logic.
Pitfall 2: Ignoring technological disruption
Extrapolating today's landscape ten years forward is the most dangerous habit. Solar replacing thermal power, EVs replacing combustion cars, smartphones replacing feature phones — in every such wave, the incumbent industry's "leader structure" was its most fragile asset. Always ask when researching an industry: could a new technology bypass it entirely? If your industry sits on a substitutable track, no amount of structural stability saves you.
Pitfall 3: Extrapolating today's landscape 10 years out
"This industry's CR5 is very high, so it's safe" — that holds only if demand, technology, and policy stay frozen. Shrinking demand triggers fresh consolidation, technology shifts rewrite the landscape, policy turns reshape entry barriers. Landscape is a point-in-time reading, not a constant; test the source of its stability before extrapolating.
Pitfall 4: Taking conclusions without checking definitions
"A trillion-yuan market" — under which definition? Exports included? Ex-factory price or retail price? Different scopes differ by multiples. "30% growth" — whole industry or listed-company sample only? Conclusions with unclear definitions have zero comparison value.
Pitfall 5: Substituting "I think" for data
Nothing kills industry research faster than gut feel: "I think this demand will explode." Whether demand explodes depends on four datasets resonating: penetration curves, orders, capacity, and policy — a lone "feeling" is not evidence.
8. A Practical Checklist for Retail Investors
When researching a new industry, complete these eight steps:
- [ ] Write the industry's definition and boundaries in one sentence (what it does, what it sells, who buys)
- [ ] Estimate market size using two methods; state ranges, not precise values
- [ ] Check CR3/CR5 and recent changes; determine the direction of concentration
- [ ] Draw a simplified industry-chain map (upstream → midstream → downstream) and mark the fattest profit segment
- [ ] Characterize the business model (asset-light/heavy, margin level, cyclical profile)
- [ ] Place the prosperity position (price, inventory, utilization, penetration — four signals)
- [ ] List catalysts expected in the next 6-12 months
- [ ] Self-check with the "ten-year view": will this industry still exist in ten years? Could it be disrupted?
⚠️ Risk Warning
⚠️ Risk Warning
Industry research is a tool for raising win rate, not a weapon guaranteeing profit. Size estimates can be badly wrong (especially on new tracks), landscapes can be rewritten overnight by technology or policy, prosperity judgments can be humiliated by black swans, and industry data carries lags and definitional padding. No methodology eliminates rotation risk or systemic risk. This is educational methodology content, not investment advice; however deep your research, face uncertainty with position sizing and stop losses.