When preparing collaboration materials, the most awkward part is often not that you can't find anything, but that you don't dare to write down what you do find. The list runs a full page—an organization, Project A, several titles—and it looks like a lot of information, but once it actually has to go into a formal document, you start hesitating over every item: Where did I see this sentence? If someone asks, what do I answer with? After a few rounds of hesitation, the page has been cut down to nothing but the title.
The problem is usually not whether you searched enough, but how you organize things. This article breaks down a set of habits for organizing publicly available materials. The core idea is just one sentence: anything that is to be written into formal materials must have a process that can be traced back.
01 Four steps: criteria, separate search, cross-check, flag doubts
▲ Figure 1 | From defining the criteria to flagging doubts, the output of the four steps is a verifiable fact list, not an evaluative conclusion
Before starting, define the criteria: what exactly needs to be verified this time? Is it the timeline of organizations and positions, public projects the person has participated in, or publicly searchable achievement records? If the criteria are not fixed, the search will diverge—you search a lot, but not a single field is complete.
Then search separately along three lines: person, organization, and project. As you search, record the source for each item, down to the page or entry. The third step is cross-checking: for the same matter, at least two sources must agree; records from a single source are marked separately and not rushed into use as conclusions. Parts that don't match are put on a separate page and flagged as unresolved; do not arbitrarily pick one or make a judgment.
The output of the entire chain is a fact list, not a character evaluation. These are two completely different things: every item in the former can point back to a public source, while the latter cannot.
02 What can go into the dossier, and what cannot
▲ Figure 2 | Dossier fields fall into three layers: what can be included, what should be included cautiously, and what should not be included; the boundaries must be written clearly in advance
The usability of a dossier depends in large part on what it excludes.
What can go into the dossier is content that can be traced back to a public source for every item: the public timeline of changes in organization and position, public projects the person has participated in, and publicly searchable achievement records. What should be included cautiously is material that exists in multiple versions—such as records with the same name, projects whose names are written inconsistently, and public pages that are slow to update. Such content should be listed side by side with sources noted, without choosing between versions.
There are three categories that should not be included, and this must be stated very firmly: information obtained through non-public channels does not go into the dossier, unverified secondhand accounts do not go into the dossier, and content involving personal privacy does not go into the dossier. These three are not suggestions; they are the bottom line for this kind of material. Once the boundaries are clear, the dossier can be used with confidence—users do not need to guess which entries are clean.
03 "I know" and "I can prove how I know"
▲ Figure 3 | The difference between verbal verification and traceable verification: the former relies on secondhand accounts, the latter relies on traceable sources
This is the most critical distinction in the article. "I know something" and "I can prove how I know it" are two different things.
Information obtained by asking around verbally has a source of "my friend said," and its traceability becomes unclear after a while; if someone else asks, the result may even be different. This cannot count as verification—only as having asked around. Traceable verification is different: each item notes a public source, the original page can be opened at any time, and anyone following the same path will get the same result. The value of the former decays over time; the value of the latter is stable.
A phenomenon often mentioned in academic discussions of reproducibility is similar: even when data and methods are made public, a considerable proportion of studies are still difficult to reproduce accurately; many public materials lack label descriptions and data-processing documentation when submitted, making it hard for replicators to reconstruct the entire process. The conclusion itself may be correct, but if the process cannot be traced back, it is hard to take seriously. Due diligence materials face more direct questioning than this, so the requirement for traceability is only higher.
▲ Figure 4 | Every fact passes three questions: Can the source be pointed to a specific page? Is there a second source? Is the source itself within the public scope?
When it comes to self-checking, every fact should pass three questions: Can the source be pointed to a specific page? Is there a second source that matches for the same matter? Is the source itself within the public scope? If any of the three questions cannot be answered, the record should either be flagged as unresolved or be ruled out outright—if the source is a non-public channel, it is not written into the dossier, with no exceptions.
04 From zero to a usable list
▲ Figure 5 | Illustrative structure of a fact list: places that don't match get their own column, marking where judgment still cannot be made
In practice, the order is roughly as follows.
First decide what to verify. Don't rush to search for people; first write down the fields needed this time: organization and position, projects participated in, achievement records, same-name distinctions. Once the field list is clear, the search has boundaries. MuchSci's dossier structure is organized along the same lines: fields first, search second.
Search separately by field. Go item by item, and note the source as you search, to avoid changing the criteria while searching—criteria drift is the most common way this kind of organizing fails.
Keep a source for every item. Record it while searching; don't wait to fill it in later. Sources filled in later usually cannot be completed.
Leave unresolved items in a separate column. Items that don't match, have missing numbers, or are written inconsistently should not be deleted; mark them as they are. In a list, the "things that don't match" column is often more valuable than the "verified as consistent" column—it marks where judgment still cannot be made.
▲ Figure 6 | Four steps completed: define fields, search separately, keep sources, flag doubts
In this workflow, MuchSci handles consolidation and retrieval: archiving public records scattered in various places by field and linking each item's source back to the original text, so that the list can be located item by item when questioned. What to search and how to judge still remains up to you.
05 Boundaries
① This article discusses methods for organizing public materials and does not constitute any promise regarding due diligence outcomes. No matter how standardized the organization is, it cannot guarantee completeness, let alone the outcome of the collaboration itself.
② This article does not target or evaluate any real person or organization. The list structure and examples in the text are illustrative only and do not constitute investigation conclusions about any individual, nor an endorsement of anyone.
③ All operations are conducted only within the scope of public materials. Information from non-public channels, unverified secondhand accounts, and content involving personal privacy are not included in the scope of organization; when materials are used for formal purposes, the current regulations of the relevant institution and parties shall prevail.