The operating manual corresponding to Article 1 of the Research Constitution's investment philosophy—"find irreplaceable companies in industries with upward prosperity." Industry chain structure is public knowledge; digging it out does not equal alpha; the value of this skill lies inturning "irreplaceability" from an adjective into an evidence-backed label, and providing a calibration of the extent to which it is priced by the market.
0. Three disciplines
- Every industry chain fact (who supplies whom, share, capacity, yield, certification progress, capacity expansion cycle) must have an evidence id (announcement / research report / news / industry data endpoint); if the source cannot be stated, write "to be supplemented"; do not fill in from impression.
- The irreplaceability label has only four values:
tech_moat(technology / process / patent / know-how that others cannot catch up with in the short term) /capacity_moat(process maturity / capacity layout / customer certification cycle that make it impossible for rivals to catch up, with share concentrated during the dividend period) /both/待补. Each label must be followed by one sentence of evidence. - This skill only answers "which layer of the chain it is at, on what basis it is irreplaceable, and to what extent the market has priced it," not "whether to buy."
1. Six-step drill-down (ask the same four questions at each layer)
Treat the recognized leader as the "demand entry point," do not stop at the leader, and move upstream along the supply chain:
需求入口(整机 / 云厂 / 终端龙头)→ ① 关键部件 → ② 核心器件 / 芯片 → ③ 关键材料 → ④ 衬底 / 耗材 → ⑤ 设备 / 检测 → ⑥ 设备的核心零件
Four questions for each layer:
- Is it irreplaceable?(technology / process / patent; or certification cycle / capacity ramp-up period)
- Is supply rigid?(how long the capacity expansion cycle is, how long yield ramp-up takes, whether there are alternative routes, how long new entrants need)
- What is the concentration?(oligopoly / duopoly / fragmented; share evidence)
- Has it been noticed?(market cap, number of research reports covering it, number of institutions in consensus expectations—chokepoints that "everyone knows" are usually already priced)
Stop at the layer where "irreplaceable + rigid supply + concentrated + low attention" hold simultaneously; for A-shares, record both lines of "leader + the deepest layer that can be researched" (whether there is a tradable target is only a factual record, not a conclusion).
2. Physical hard sieve (constraints more reliable than business narratives)
| Constraint type | What to look at | Evidence source (registry endpoint example) |
|---|---|---|
| Capacity expansion cycle | How long it takes for a new production line from investment to full production; equipment delivery time | Announcements (fetch_announcements / cninfo_announcements), research reports (em_reports / em_industry_reports) |
| Yield / process maturity | Yield ramp-up curve, unit cost, batch stability | Research reports, company interactive Q&A (cninfo_irm), news (em_stock_news / rss_news) |
| Certification cycle | Verification time required to enter a major customer's BOM, cost of switching suppliers | Announcements, research reports, industry news |
| Alternative routes | Whether a technology route switch would bypass this layer (route discontinuity risk) | Industry research reports, overseas peers (yahoo_news / sec_filings etc.) |
| Material / resource monopoly | Is global supply concentrated in a few regions / companies (China-exclusive physical chokepoints: some rare-earth permanent magnets, specialty chemical materials) | Industry data, macro endpoints |
| Industry classification and prosperity | Industry classification, peer comparison, prosperity direction | sw_industry、em_industry_comparison、em_concept_blocks |
Physics does not lie: prioritize verifiable constraints such as "capacity expansion cycle / yield / certification cycle" rather than narratives such as "good brand / ecosystem / track".
3. Calibration: the harder the chokepoint, the more expensive
- Once the irreplaceability of a chokepoint is widely recognized, valuation will have fully priced it in ("the harder the chokepoint, the more expensive"); the labels of this skill should be viewed together with valuation's PEG / percentile:Hard chokepoint + high PEG is the norm, not an opportunity signal.
- Four questions on expectation gap(to judge whether there is an expectation gap in the "new direction" dimension): ① Is the direction new (cognition not yet formed) ② Will supply fall short of demand (new demand × rigid supply) ③ Is there a listed target in this market ④ Has it already been speculated on(the most critical — most hot directions are stuck at this question). Identification signals: on the demand side, "route switching / everyone betting on"; on the supply side, "sold out / lengthening lead times / price increase letters".
- Default assumption and falsification: for any "new direction", firstassume it has already been priced by the market, then use question ④ to falsify — the evidence is the gain / turnover / number of research report coverages / consensus expectation revisions of targets related to that direction (available from registry endpoints); if evidence cannot be obtained, write "degree of pricing to be supplemented", and do not conclude "already speculated / not speculated" based on impressions. Under this premise, the usable analytical perspectives are: ① when the direction is real + there is an order fulfillment path, analyze "certainty of fulfillment"; ② for targets with real business that have not yet been conceptualized, analyze "the gap between fundamentals and pricing"; ③ when the track has been fully priced, only compare "degree of performance fulfillment + whether growth can digest valuation" (back to the valuation framework). These three areanalytical frameworks, used to organize evidence and adjudication points, not screening or buy/sell instructions.
4. Output format (write into the extra_findings of the profile / risk stage or the "inference" section of the report)
Industry chain position table:
| Layer | Segment | Company role | Irreplaceability label | Evidence (ev id) | Supply rigidity (capacity expansion / yield / certification) | Concentration | Attention |
|---|
Add three lines of conclusions:
- Label:
tech_moat / capacity_moat / both / 待补+ one sentence of evidence; - The biggest "bypassed" risks (technology route discontinuity / customer in-house development / overcapacity), one each, go into the counter-evidence list of catalyst-risk;
- Adjudication points: what data coming out would change the label (e.g., a competitor passing certification, new capacity going into production, route-switching orders landing).
5. What not to do
- Do not give "core pool / recommendation / add position" type conclusions; do not translate the labels of this skill into actions.
- Do not use model memory to fill in share, capacity, or customer lists; write "to be supplemented" for cells without evidence.
- Do not cite external personal views or social media statements as facts; they can only serve as clues, and clues must be corroborated by data from registry endpoints before entering the report.