A strong legal answer can fail long before the drafting stage. The usual cause is not a lack of authority, but a research process that began with a broad search term, followed an attractive headnote, and never tested whether the proposition actually decides the issue. This guide to legal research workflow sets out a more disciplined approach for Hong Kong matters: define the legal question, find the right primary materials, verify their force and currency, then turn them into advice that another lawyer can audit.

Start with the question, not the database

A research request often arrives in factual language: can the employer dismiss this employee, is the clause enforceable, or does the regulator have this power? These are not yet research questions. Before searching, separate the facts that may affect the result from the legal elements that must be proved, construed or challenged.

Write a short issue statement that identifies the parties, the remedy or outcome sought, the governing area of law, and the decisive factual uncertainty. For example, instead of searching “unfair dismissal Hong Kong”, frame the question more precisely: whether the statutory protection applies to this category of employee, what threshold conditions apply, and which facts satisfy them.

This step prevents a common efficiency problem: collecting many cases that use the same vocabulary but answer a different question. It also exposes when the assignment requires several workstreams. A contractual claim, a statutory claim and a regulatory consequence may arise from the same events, but they should not be researched as one undifferentiated issue.

Build an issue map

For each issue, record the rule to be established, the material facts, potential exceptions, the relevant date, and the authority required. The date deserves particular care. Hong Kong legislation may have been amended, commenced in stages, or applied differently at the time of the events. A correct reading of the current text can still be wrong advice about a past transaction.

Also record assumptions. If the client has not confirmed whether a notice was served, whether consent was informed, or where a loss occurred, state that explicitly. Research can then proceed on alternative factual bases rather than quietly treating an assumption as evidence.

A guide to legal research workflow: move from source to proposition

The most reliable workflow follows the hierarchy of authority. Start with legislation and binding case law, then use secondary sources to orient the analysis or identify further materials. Commentary can be useful, particularly in an unfamiliar area, but it should not become the final support for a proposition that the statute or judgment can establish directly.

Begin with the relevant ordinance, subsidiary legislation and any applicable rules, codes or notices. Read more than the section that appears in the search result. Definitions, application provisions, transitional arrangements, exceptions, cross-references and schedules often change the effect of the operative wording. Identify the version in force on the material date, not merely the version displayed by default.

Then search for judicial treatment. Use a combination of precise legal concepts, statutory section numbers, distinctive factual patterns and the remedy sought. Exact keywords still have a place when the statutory language is central. They are less effective where the issue is expressed through varying language, such as whether conduct amounts to a repudiatory breach or whether a duty of care is owed in a particular relationship.

Semantic search is valuable here because it can retrieve cases that reason about the same legal proposition without repeating the chosen phrase. On Common Laws.ai, researchers can search Hong Kong case law by legal meaning, review AI-generated summaries and move quickly to extracted passages. That accelerates triage, but it does not remove the need to read the relevant reasoning in context.

Treat search results as leads, not answers

A case may look highly relevant for several reasons that do not make it controlling. It may concern a similar fact pattern while applying a different statutory regime. It may state a broad principle in passing. It may be a first-instance decision later qualified by an appellate court. Or it may cite a proposition without deciding it.

For every promising judgment, ask four questions. What issue was before the court? What facts mattered to the result? What rule was necessary to the decision? What did the court actually do with that rule? The answer should be tied to paragraph references, not a general impression of the judgment.

This distinction between ratio and useful but non-binding observation is especially important when preparing advice that may be challenged. A concise, accurate proposition with a pinpoint reference is stronger than a lengthy string citation that leaves the reader to infer its relevance.

Check authority, treatment and currency

Research is not complete when a case is found. Its status must be assessed. Consider the court level, whether the point was directly decided, later judicial treatment, and whether legislation has displaced or altered the analysis. A Court of Final Appeal decision will ordinarily carry greater weight than a first-instance decision, but even senior authority must be read for the issue it resolved.

Citation checking should be active rather than ceremonial. Follow key authorities cited in the leading case, then look forward for later decisions that apply, distinguish, doubt or explain it. Where a judgment describes an earlier case in broad terms, retrieve the earlier case itself before relying on that characterisation. Courts sometimes compress an authority for the purpose of a different issue.

Legislation requires the same discipline. Confirm commencement dates, amendments and repeals, and examine transitional and saving provisions where events span a change in law. A point-in-time legislative reference is not an administrative extra. It may determine the entire answer.

There is a trade-off between speed and exhaustiveness. For a short internal query, a targeted review of the statute, leading authorities and recent treatment may be proportionate. For an opinion, contentious correspondence, a transaction with significant exposure or litigation preparation, broaden the search and record the scope. The standard is not to find every case ever reported. It is to make a defensible judgement that the research is sufficient for the risk and purpose of the work.

Convert reading into a usable research record

Do not leave the analysis buried in browser tabs, copied extracts or a folder of judgment PDFs. As each authority is reviewed, create a compact record: full citation, court and date, proposition supported, key paragraph, material facts, status or treatment, and any limitation on use.

This makes the final drafting faster and improves quality control. It also reveals gaps early. If several authorities support a general principle but none addresses the statutory exception at the centre of the facts, the research is incomplete even if the case list is impressive.

A useful working structure is issue-led rather than source-led. Under each issue, place the governing text first, followed by the binding authority, persuasive authority where relevant, application to the known facts, and unresolved points. This mirrors the structure of sound advice and avoids writing a chronology of everything read.

Extract the decisive passages

A headnote, summary or search snippet is a navigation aid. The operative value lies in the judgment’s reasoning. Read the paragraphs before and after any promising extract to understand qualifications, factual context and whether the court adopted or merely recorded an argument.

When extracting a passage, capture enough surrounding text to preserve its meaning. Then write, in your own words, the proposition it supports. If you cannot state that proposition clearly and narrowly, the authority may not yet be understood well enough to cite.

Draft for verification, not display

The final product should let a supervising solicitor, client or colleague test the reasoning quickly. Lead with the conclusion that the current materials support, qualify it where facts or law remain uncertain, and show the shortest route from rule to application.

Avoid overstating predictive confidence. Legal research often produces a range rather than a binary answer, particularly where the test is fact-sensitive or the authorities point in different directions. Say what fact would strengthen or weaken the position. Identify what document, witness evidence or further instruction would resolve the uncertainty.

Before sending, run a final audit. Check each citation against the proposition beside it, every legislative reference against the relevant historical version, and every statement of law against the highest applicable authority. Remove authorities that add volume but no analytical value. The result should be concise enough to use under pressure and complete enough to withstand scrutiny.

The practical advantage of a disciplined workflow is not simply faster searching. It is the ability to know why a source matters, whether it remains good law, and exactly how far it takes the analysis. That is the standard that turns research into dependable legal work.