The short answer: create a research record简要答案：建立一份研究记录
To research a stock, start with a narrow decision question, split it into a question tree, and attach every material claim to a dated primary source. Keep reported facts, your calculations, assumptions, and interpretations in separate fields. Then search deliberately for contrary evidence and set filing- or event-based update triggers. Analysis may produce a snapshot of business quality or valuation; research is the maintained evidence system that makes that snapshot reproducible and revisable.
研究一只股票时，先提出一个边界清楚的判断问题，再把它拆成问题树，并让每项重要观点连接到注明日期的第一手来源。把已披露事实、自己的计算、假设和解释分栏记录；随后主动寻找反证，并设置由财报或事件触发的更新节点。分析可以形成某个时点的商业质量或估值快照，研究则是持续维护证据、让快照能够复现并随新信息修订的系统。
Research continues after the analysis ends分析结束之后，研究仍在继续
A stock analysis usually has a defined endpoint: calculate a valuation range, compare margins, or write an investment memo. Stock research begins earlier and lasts longer. It decides which questions matter, collects the documents needed to answer them, records provenance, notices conflicts, and reopens conclusions when new evidence arrives. Treating the two as synonyms encourages a polished answer without a durable trail.
股票分析通常有明确终点，例如算出估值区间、比较利润率或写完投资备忘录。股票研究开始得更早、持续得更久：它先决定哪些问题值得回答，再收集所需文件，记录资料出处，识别矛盾，并在新证据出现时重新打开旧结论。若把二者当成同义词，很容易得到一份看似完整却没有持久证据链的答案。
A dated conclusion, model, comparison, or scenario built from the evidence available at that moment.
分析产出基于当时可得证据形成的、带日期的结论、模型、比较或情景。
Questions, source locations, claim status, rejected explanations, open gaps, and the next observable checkpoint.
研究档案问题、来源位置、观点状态、已排除解释、未解决缺口，以及下一项可观察核验。
Turn the broad idea into a question tree把宽泛想法拆成问题树
Write the root question so that evidence could change your answer. “Is this a great company?” is too elastic. “Can the company replace a departing distributor without increasing customer dependence over the next four reported quarters?” identifies a mechanism, a horizon, and observable evidence. Under that root, create branches for the business mechanism, dependencies, management actions, financial consequences, and disconfirming signals.
根问题必须写成能够被证据改变答案的形式。“这是不是一家伟大公司”过于宽泛；“公司能否在未来四个披露季度内替代离开的经销商，同时不提高客户依赖”则明确了机制、时间范围和可观察证据。随后在根问题下建立分支：商业机制、关键依赖、管理层行动、财务后果与反向信号。
| Branch | Question to answer | Evidence that would move it |
|---|---|---|
| Mechanism | Who chooses, pays, and can switch? | Contracts, channel disclosures, customer behavior |
| Dependency | What must remain available or approved? | Supplier terms, licenses, key-customer concentration |
| Execution | What has management actually completed? | Dated milestones, not undated aspirations |
| Counter-case | What would make the thesis wrong? | A measurable threshold and a future filing location |
| 分支 | 需要回答的问题 | 能够改变判断的证据 |
|---|---|---|
| 商业机制 | 谁做选择、谁付款、谁能够切换？ | 合同、渠道披露、客户行为 |
| 关键依赖 | 哪些资源、批准或关系必须持续存在？ | 供应条款、许可证、主要客户集中度 |
| 执行进度 | 管理层实际完成了什么？ | 带日期的里程碑，而非没有时点的愿景 |
| 反向情景 | 什么情况会证明原观点错误？ | 可衡量阈值与未来披露位置 |
If a branch cannot name a source, an observable signal, or a decision it could affect, it is probably background reading rather than a research task.
范围测试如果某个分支说不出来源、可观察信号或它会影响的判断，那么它更可能只是背景阅读，而不是研究任务。
Build the source log before writing the story先建来源日志，再写公司故事
Start with the regulator or exchange filing system, then the issuer’s filed reports, exhibits, proxy materials, and dated investor communications. Commentary can surface questions, but it should not silently become proof. Save the exact form, reporting period, filing date, amendment status, page or section, URL, and the sentence or table the claim relies on. When a later filing changes the disclosure, preserve the earlier version and mark it superseded rather than overwriting history.
先从监管机构或交易所披露系统入手，再查看发行人提交的定期报告、附件、委托投票材料和注明日期的投资者沟通。评论文章可以帮助发现问题，但不能悄然变成证明。日志应保存准确表格类型、报告期、提交日期、修订状态、页码或章节、URL，以及观点所依赖的句子或表格。后续文件改变披露时，应保留旧版本并标记“已被替代”，而不是覆盖历史。
- Use one source ID everywhere, such as S-2025-10K-1A, so a memo claim can be traced without searching again.
- Quote sparingly; record enough surrounding context to avoid detaching a risk, exception, or accounting definition.
- Separate absence from contradiction. “Not disclosed” does not prove “does not exist,” while a later filing may genuinely conflict with an earlier claim.
- 为每个来源设置统一编号，例如 S-2025-10K-1A，让备忘录中的观点无需重新搜索就能追溯。
- 少量摘录，并保留足够上下文，避免把风险、例外条件或会计定义从原意中剥离。
- 区分“没有披露”和“存在矛盾”。未披露不能证明不存在，而后续文件确实可能与早期说法冲突。
Run three passes, then challenge the file完成三轮研究，再挑战整份档案
- Triage the company. Confirm the listed entity, share class, fiscal year, segments, and latest filing set. Write the root question before reading management’s narrative.
- Trace each branch. Move from business and risk sections into the relevant notes, exhibits, proxy disclosures, and prior filings. Add a source ID as soon as a claim enters the file.
- Reconcile time. Compare what management previously said would happen with what later filings report. Distinguish a missed milestone from a milestone whose definition changed.
- Challenge the leading explanation. Search the same filing for exceptions, concentration, termination rights, related-party details, amendments, and language that became less specific.
- Close the session. Record the strongest supported answer, the strongest contrary evidence, the largest gap, and the next dated trigger. Do not mark the whole company “researched.”
- 快速建档：确认上市主体、股份类别、财政年度、分部和最新披露组合。在阅读管理层叙述之前，先写下根问题。
- 追踪分支：从业务与风险章节进入相关附注、附件、委托投票披露和历史文件。任何观点一旦进入档案，就立即添加来源编号。
- 对齐时间：把管理层过去声称会发生的事项，与后续披露的实际结果比较；同时区分“里程碑未完成”和“里程碑定义被改变”。
- 挑战首选解释：在同一份披露中搜索例外条款、集中度、终止权、关联方、修订内容，以及措辞是否变得更含糊。
- 结束本轮：记录最强支持答案、最强反证、最大缺口和下一项带日期的触发器。不要把整家公司标记为“研究完成”。
Research file example: a distributor exit研究档案例：经销商退出之后
The following dossier is entirely hypothetical and does not describe a real security. Northstar Instruments sells testing equipment through distributors and a small direct-sales team. The research question is whether a channel transition is reducing dependence or merely moving it to fewer direct customers.
以下档案完全虚构，不对应任何真实证券。Northstar Instruments 通过经销商和一支小型直销团队销售测试设备。研究问题是：渠道转型是否正在降低依赖，还是仅仅把依赖从经销商转移到更少的直销客户。
| Evidence class | Dossier entry | Permitted use |
|---|---|---|
| Reported fact | Hypothetical annual filing: largest customer represented 24% of sales versus 15% one year earlier; one national distributor agreement ended after year-end. | Cite these as disclosed figures and events only. |
| Calculation | Concentration increased 9 percentage points; 24% divided by 15% equals a 60% relative increase. | Show the formula and avoid calling either measure customer churn. |
| Assumption | The ended distributor previously served some accounts now buying direct. | Label it unresolved until customer or channel detail supports it. |
| Interpretation | The transition may have improved direct access while increasing exposure to one buyer. | Keep “may”; the facts do not establish bargaining power or retention. |
| 证据类别 | 档案记录 | 允许得出的内容 |
|---|---|---|
| 已披露事实 | 虚构年报：最大客户销售占比由一年前的 15% 升至 24%；一家全国经销商协议在年末后终止。 | 只能把它们引用为已披露数字与事件。 |
| 研究者计算 | 集中度提高 9 个百分点；24% 除以 15% 得到相对增幅 60%。 | 展示公式，不能把任一结果直接称为客户流失。 |
| 研究假设 | 终止合作的经销商过去服务了部分现已转为直销的账户。 | 在客户或渠道明细支持前，必须标为未解决。 |
| 研究解释 | 转型可能改善了直接触达，同时增加对单一买方的暴露。 | 保留“可能”；现有事实无法证明议价能力或留存率。 |
The counter-case is equally important. If the next filing shows that the largest-customer share falls below 18% while several named channel replacements become active, the dependency concern weakens. If concentration stays above 24%, the terminated agreement remains unreplaced, or disclosure stops distinguishing direct and distributor sales, the concern strengthens or becomes harder to test. The file therefore ends with three update tasks: inspect the next concentration note, search exhibits for new or amended distribution agreements, and compare management’s channel language word for word with the prior filing.
反向情景同样重要。如果下一份披露显示最大客户占比降至 18% 以下，同时多个有明确名称的替代渠道开始运营，那么依赖风险会减弱。如果集中度仍高于 24%、终止协议没有替代，或者披露不再区分直销与经销商销售，风险就会增强，或变得更难验证。因此档案以三项更新任务结束：检查下一份客户集中度附注；在附件中搜索新增或修订的经销协议；逐字比较管理层渠道措辞与前一份文件。
It does not prove that direct selling is better, that the large customer will leave, or that the stock is attractive. It demonstrates how one research question stays open, testable, and updateable.
该示例不能证明什么它不能证明直销一定更好、最大客户一定会离开，也不能证明股票具有吸引力；它只演示如何让一个研究问题保持开放、可检验且可更新。
Use a claim ledger to prevent narrative drift用观点台账防止叙述漂移
Give each important claim a status: supported, partly supported, contradicted, superseded, or open. Add one best source, one contrary source or test, an owner, and a next-review date. When wording changes from “will” to “expects,” or a previously reported measure disappears, record the change itself as evidence. Do not rewrite the old note to make the history look consistent.
为每项重要观点设置状态：已支持、部分支持、被反驳、已替代或未解决。每条观点都要附上最佳来源、一个反向来源或检验、负责人和下次复核日期。当措辞从“将会”变为“预计”，或过去披露的指标消失时，应把这种变化本身记录为证据，不要改写旧笔记来制造前后一致。
- A calculation is reproducible only when its inputs, units, periods, and formula remain visible.
- An assumption earns attention in proportion to how much the conclusion changes when it is removed.
- An interpretation should name at least one plausible alternative and the observation that would distinguish between them.
- 只有输入、单位、期间和公式都可见，计算才可复现。
- 一个假设值得投入多少精力，取决于移除它之后结论会改变多少。
- 解释至少要写出一个合理替代原因，以及能够区分两者的观察信号。
Update on events, not only on the calendar按事件更新，而不只按日历更新
Use three clocks. Review event filings, material announcements, leadership changes, financing, litigation, or regulatory decisions when they appear. Reconcile open claims at each periodic report. Once a year, rebuild the source map from the latest annual filing and archive stale assumptions. Price movement alone can justify refreshing valuation inputs, but it does not automatically change evidence about the business mechanism.
研究档案需要三只时钟。第一，重大事件披露、重要公告、管理层变更、融资、诉讼或监管决定出现时立即复核；第二，每份定期报告发布后调节所有未解决观点；第三，每年用最新年报重建来源地图，并归档陈旧假设。股价变化可以触发估值输入更新，但不会自动改变关于商业机制的证据。
A new filing directly touches a root question, threshold, contract, approval, or named dependency.
立即重开新披露直接涉及根问题、阈值、合同、批准事项或已命名依赖。
The information is interesting but cannot yet alter a claim or identify a new test.
稍后复核信息虽然有趣，但尚不能改变观点，也无法提出新的检验。
Stop a research round without pretending certainty结束本轮研究，但不假装已经确定
A research round can stop when every decision-relevant branch has a source or an explicit gap, the leading interpretation has survived a genuine contrary test, and the next update trigger is recorded. The deliverable is not “everything about the company.” It is a dated answer whose evidence strength, unresolved questions, and revision path are visible to another reader.
当每个影响判断的问题分支都有来源或明确缺口，首选解释经受了真实反向检验，并且下一次更新触发器已经记录时，本轮研究可以结束。交付物不是“关于公司的全部知识”，而是一份带日期的答案；另一位读者能够看到证据强度、未解决问题和未来修订路径。
Organize a multi-file research packet整理多文件研究资料包
When annual reports, interim filings, exhibits, and investor materials are spread across files, InfiniSynapse Stock Explained can help turn the uploaded packet into a plain-language brief and surface locations relevant to a research question. Use it to organize document comparisons, create a first-pass claim list, and identify passages that deserve manual review.
The tool does not guarantee that every disclosure was collected or that a located passage supports your interpretation. Confirm company identity, form type, period, units, page references, amendments, and all material claims against the original uploaded documents. Keep your external source log and update triggers outside the generated summary so the research record remains auditable.
当年报、阶段性报告、附件和投资者材料分散在多个文件中时，InfiniSynapse“一眼看懂这只股票”可以把用户上传的资料包整理成通俗摘要，并定位与研究问题有关的内容。适合用于组织文件比较、建立第一版观点清单，以及找出值得人工精读的段落。
工具不能保证所有披露都已收集，也不能保证定位到的段落支持你的解释。必须回到原始上传文件核对公司主体、表格类型、期间、单位、页码、修订状态与所有重要观点。外部来源日志和更新触发器应在生成摘要之外继续维护，才能让研究记录保持可审计。
Upload the latest complete filings plus the earlier documents needed to test how management’s claims changed. Ask one narrow research question and verify every important output at its cited location.
上传最新完整披露，以及用于检验管理层说法变化的历史文件。一次提出一个边界清楚的研究问题，并在引用位置核验每项重要输出。
Open Stock Explained打开“一眼看懂这只股票”Questions about maintaining stock research关于持续维护股票研究的常见问题
Use enough history to cover the mechanism you are testing. Three to five annual periods may reveal continuity, but acquisitions, a new reporting segment, or a business-model change can make older periods less comparable. State the chosen window and why it fits the question.
股票研究档案应包含多少年的披露？应覆盖足以观察所研究机制的历史。三到五个年度期间可能显示连续性，但收购、新分部或商业模式变化会削弱更早期间的可比性。需要写明所选时间窗口，以及它为何适合当前问题。
It is blocking when a missing document or undisclosed definition could reasonably reverse the answer to the root question. Mark the claim open, describe the missing evidence, and avoid replacing it with an industry average or management assertion.
什么样的研究缺口足以阻止形成结论？如果缺失文件或未披露定义有可能推翻根问题的答案，这就是阻断性缺口。应把观点标为未解决，说明缺少哪项证据，不要用行业平均值或管理层主张替代。
Record guidance as a dated management statement, not as a reported outcome. Preserve its definition and range, then create a later checkpoint that compares the guidance with what was actually reported without quietly changing the original target.
管理层指引应如何进入证据日志？把指引记录为带日期的管理层陈述，而不是已实现结果。保留其定义和区间，再建立后续核验点，把指引与实际披露结果比较，不能悄然改变原目标。
Use the issuer’s home-market regulator or exchange as the primary filing source and identify the applicable report forms, language, currency, and accounting framework. A U.S. depositary receipt may add SEC filings, but it should not cause you to ignore the home-market record.
如何研究美国以外市场上市的股票？以发行人本土监管机构或交易所为主要披露来源，并确认适用的报告表格、语言、币种和会计框架。美国存托凭证可能带来额外 SEC 披露，但不能因此忽略本土市场记录。
Not by itself. Refresh dated price-dependent calculations and check whether a filing, announcement, or market event explains the move. Reopen the business question only when new evidence touches its mechanism, dependency, threshold, or counter-case.
股价大幅波动是否应该触发完整研究更新？不应仅凭价格波动就重做全部研究。先更新依赖价格的计算，并检查是否有披露、公告或市场事件解释波动。只有新证据触及商业机制、关键依赖、阈值或反向情景时，才重开相应业务问题。
Primary sources and research boundary第一手来源与研究边界
These official U.S. sources explain where to locate common public-company filings and what several filing types contain. A company-specific file should use the regulator, exchange, accounting framework, and disclosures applicable to that issuer; SEC forms are not a universal template for every market.
This guide is an educational research method, not individualized investment advice, a completeness guarantee, or a forecast. A traceable dossier can reveal what supports a conclusion and what remains unknown, but it cannot remove business risk, disclosure limitations, market uncertainty, or the possibility that new information changes the answer.
以上美国官方资料说明了常见上市公司披露的查找位置，以及若干表格包含的内容。研究具体公司时，应使用适用于该发行人的监管机构、交易所、会计框架和披露制度；SEC 表格并不是所有市场的通用模板。
本文提供的是教育性的研究方法，不是个性化投资建议、完整性保证或收益预测。可追溯档案能说明哪些证据支持结论、哪些问题仍然未知，但它不能消除经营风险、披露限制、市场不确定性，也不能排除新信息改变答案的可能。