Resources

Research guides built around evidence.

Practical guidance for company research, source review, private and public markets, Capital AI, screening, market maps and team workflows. The goal is to make research faster without lowering the standard for provenance, freshness or professional judgment.

Visit Help Center Browse FAQ

Research library

Make the method as clear as the conclusion.

Each guide explains a research problem, the operating principle behind it and the evidence behavior that keeps the work reviewable by someone other than the original analyst.

Research method

Build a company view from evidence, not from empty fields

A strong company view separates supported facts from unavailable information and keeps the reporting period, source and freshness visible. This is especially important for private companies, where disclosure can be uneven and a blank field should never be interpreted as zero.

Start with identity and source coverage

Confirm that the entity is resolved correctly and identify which filings, market providers, private-company sources or owned documents are actually connected. Coverage determines which questions the profile can answer reliably.

Keep missing values explicit

Do not replace an unavailable value with an estimate merely to complete a table. If an estimate is used as analysis, label it as analysis and keep it separate from sourced company facts.

Source review

Read a financial change with the reporting period attached

A number can change because the business changed, a new filing arrived, a period was restated or the normalization logic changed. Review the period and source before deciding that a changed value represents a new operating development.

Compare the same basis

Check currency, fiscal period, units and whether the figures are reported or derived before comparing one observation with another. Small metadata differences can create large apparent changes in a financial screen.

Reopen material evidence

When a number drives a valuation, recommendation or diligence conclusion, return to the source document or supported evidence. The research interface should shorten that path rather than replace the verification step.

Private markets

Research funding and transactions without filling the gaps

Private-market research often contains incomplete disclosure around terms, ownership and valuation. Use supported participants, dates, transaction types and amounts where available, while leaving undisclosed terms unavailable instead of creating a false sense of precision.

Separate event from interpretation

A financing or acquisition can be a verified event while the strategic rationale, ownership outcome or implied valuation remains uncertain. Keep those different levels of confidence visible in the research record.

Trace participants carefully

Investor and fund names can be ambiguous across vehicles or affiliated entities. Resolve the participating entity where the source supports it and avoid collapsing distinct funds or organizations merely because their names are similar.

Public markets

Use market data with the right freshness label

A price observation is only as useful as the user’s understanding of when it was captured and what rights support it. Real-time, delayed and end-of-day data should remain distinguishable throughout charts, tables, alerts and downstream exports.

Preserve freshness downstream

If a quote is exported through an API or integration, carry freshness metadata with it where supported. A downstream dashboard should not accidentally turn a delayed observation into something that appears real time.

Do not confuse recency with significance

The newest market observation may be operationally current without being economically meaningful. Research should connect price changes with filings, earnings, news or other source context before a conclusion is drawn.

Capital AI

Use AI to find evidence faster without lowering the standard

AI can help synthesize connected records, summarize source material and identify where to investigate next. It becomes risky when a fluent response is allowed to stand in for evidence that the connected data does not actually contain.

Ask questions that can be checked

Questions tied to companies, documents, periods, transactions or explicit comparisons are easier to review than broad prompts with no defined evidence boundary. Narrow questions also make unsupported assumptions easier to detect.

Treat citations as a path, not decoration

A citation is useful when it lets the researcher inspect the relevant source and understand what portion of the answer it supports. If the source does not support the claim, the answer should be narrowed or rejected.

Screening

Create screens another researcher can reproduce

A defensible screen uses criteria that correspond to supported fields and records why those criteria matter to the research question. Reproducibility makes it easier to understand why a company entered or left the universe after underlying data changed.

Use supported fields

A sophisticated filter is not useful if the data universe does not reliably cover the field being screened. Confirm coverage before using a criterion as a hard inclusion or exclusion rule.

Save the research logic

Record important thresholds, date assumptions and the reason each filter was chosen. A future teammate should be able to rerun the screen without reverse-engineering the original analyst’s intent.

Market maps

Build market maps from a defined universe

A market map should begin with an explicit research universe rather than a desire to fill every visual category. Companies and relationships belong on the map when screening logic or source evidence supports their inclusion.

Define categories before polishing the visual

Decide which business models, customer segments, geographies or other dimensions matter to the thesis first. Visual organization becomes more useful when the categories correspond to the decision the map is meant to support.

Keep the table behind the graph

Relationship maps are easier to audit when the underlying entities and edges can also be reviewed in a structured list. Researchers should not have to trust a visual node or link that cannot be traced back to the data behind it.

Team workflow

Create an evidence handoff another analyst can pick up

Shared research becomes useful when a teammate can see the question, universe, source set, assumptions and unresolved issues without restarting the work. Saving only a polished conclusion loses the operating context that makes the research defensible.

Keep open questions visible

A research handoff should identify what is known, what is unsupported and what still needs validation. That prevents the next analyst from treating an unfinished assumption as a confirmed fact.

Separate source facts from team notes

Internal interpretation can be valuable, but it should remain distinguishable from information obtained from filings, providers or other sources. Clear separation makes later review faster and reduces accidental citation of an internal assumption as external evidence.

Continue

Use Help when you need product guidance and Methodology when you need the evidence rules.

Open Help Center Read methodology