There is a particular silence that follows a badly structured research memo. The senior reads the first page, does not find an answer, skims the second, sees a list of cases, puts it down, and asks the question again verbally. Everything in the memo may be correct. None of it was usable, because it was organised around the researcher's journey rather than around the reader's decision. The difference between a memo that gets used and one that gets re-asked is almost entirely structural, and the structure is learnable in an afternoon.
The other half of the difference is honesty about confidence. A memo that presents an arguable position and a settled one in the same confident register forces the reader to redo the research to find out which is which. A memo that says plainly that a point is settled, that another is arguable, and that a third has contrary authority which has not been tested in this High Court, saves the reader the entire exercise. Stating confidence honestly is what turns a pile of search results into professional work product.
The test a memo has to pass
A senior should be able to read the first paragraph, know the answer, know how much weight it will bear, and know what is still open. Everything after that first paragraph exists so the answer can be checked, not so the answer can be discovered.
The Question Presented
A memo begins with the question it answers, stated in one or two sentences, narrowly enough to be answerable. This sounds trivial and it is the step most often botched, in two opposite directions. Too wide, and the question cannot be answered at all: whether a writ petition is maintainable, whether the contract is enforceable, whether the suit is barred. Too narrow, and the memo answers a question that has stopped being the client's problem.
The corrective is to build the legally material facts into the question. Not the whole narrative, and not a substitute for the facts section, but the two or three facts that would change the answer if they changed. A question that reads well has the shape: whether a party in this position, having done this thing, at this time, can obtain this relief in this forum. The first thing a senior should be able to do with it is correct it, and a question specific enough to be corrected is the point. If the question could be pasted into any other file in the office, it is too abstract to be useful.
The Short Answer, in the First Paragraph
The answer goes at the top. Answer first, qualification second. Not because the reader is impatient, though they are, but because everything that follows is easier to read when the reader knows what it is building towards. The discipline is to put the operative word early: yes, no, probably, probably not, it depends on a fact we do not have. Then the qualification, in the same paragraph, in a sentence beginning with the condition. The common failure is a short answer that spends four sentences describing the state of the law and never says what should be done. That is not a short answer. It is an abstract of the analysis moved to the front.
If the reader has to reach page three to find out what you think, you have not written a memo. You have written a report on your afternoon.
The short answer also states the governing authority, with the court, the bench strength where it matters, and the date. Not a citation string in the middle of a sentence, but enough to place it: a three judge bench of the Supreme Court in a given year, a Division Bench of your own High Court, a line of Single Judge decisions with no appellate consideration. Bench strength and date are the two facts that determine how much weight the authority will bear, and they belong in the first paragraph rather than buried in a footnote.
The Authority Table
After the short answer, before the analysis, put the authorities in a table. This is the single highest value formatting decision in a research memo, and it is the one juniors are least often taught. A table forces you to state what each case is for, and it lets the reader scan the field in fifteen seconds rather than reconstruct it from prose.
Five columns, and each of them is load bearing:
- Authority. Case name and citation, or the statutory provision. One row per authority, not one row per proposition.
- Court and bench. The court, and the bench strength where that decides anything. A five judge bench and a Single Judge are different instruments and the table should show it at a glance.
- Date. The date of the decision. Precedent research is a chronological exercise and a table sorted by date shows the direction of travel that prose conceals.
- Proposition. What this authority establishes, in one sentence, in your words, tied to your question. Not a summary of the case. The one thing you are using it for.
- How it has been treated since. Followed, distinguished, doubted, referred, overruled, or not considered. This column is what stops a memo from resting on an authority that has quietly been overtaken.
The last column is the one that separates a research memo from a citation list. It is also the column that reveals whether the research was actually done. A row that says nothing in it is a row where the forward check has not been run, and a senior will read the blank exactly that way.
Analysis Organised by Element, Not by Chronology
The analysis is where most memos lose their reader, and the cause is almost always the organising principle. A memo organised by the order in which you found things is a diary. A memo organised case by case, summarising each in turn, forces the reader to hold five summaries in their head and do the synthesis themselves, which is the work they asked you to do.
Organise by legal element. Break the question into what must be established and give each element its own heading. Under each: the rule, the authority for the rule, how the client's facts sit against it, and the counter. Four movements, in that order, repeated. It is monotonous to write and extremely fast to read, and it lets the reader go straight to the element they are worried about, which is what they will do anyway.
The Contrary Authority Section, Which Must Exist
Every memo needs a section headed contrary authority, or the equivalent, and it must be honest. This is not a matter of professional virtue alone, though it is also that. It is a matter of self interest with a short time horizon: the senior will find the contrary case, or the opponent will, and the person who did not mention it will be assumed either not to have found it or to have hidden it. Both conclusions are worse than the case itself.
The section should do three things. State the contrary authority accurately, at its strongest rather than in a weakened paraphrase. Say why it may not govern: distinguishable on facts, from another High Court and therefore persuasive only, obiter, per incuriam, or superseded by amendment. And say honestly how good that answer is. A distinguishing argument that depends on a fact the client may not be able to prove is not a strong answer, and describing it as one is the kind of thing that is remembered.
The omission that ends trust
A memo that does not mention the contrary case is not a neutral document with a gap. It is a document that told the reader there was nothing there. Once a senior has been surprised in court by an authority that should have been in your memo, every later memo you write is read with suspicion, and that is a slow thing to undo.
Confidence Markers That Mean Something
A closed vocabulary, used the same way every time
Use a fixed vocabulary for confidence, use it consistently, and make sure each word commits you to something specific. Ad hoc hedging language, it appears that, it is submitted that, it may be argued, communicates nothing because it means whatever the reader wants it to mean. A small closed set of markers means the reader learns your scale and can rely on it.
| Marker | What it commits you to | What would change it |
|---|---|---|
| Settled | There is binding authority directly on the point, and no live challenge to it. You would advise on this without qualification. | A pending reference to a larger bench, or a statutory amendment to the provision construed. |
| Well supported | A consistent line of authority supports the position, though the binding authority is not squarely on these facts. | A decision distinguishing the line on facts closer to ours, or a contrary view from a higher bench. |
| Arguable | There is a respectable case to be made, and a bench could reasonably go either way. Do not plan on winning it. | Appellate consideration of the point, or authority from the High Court where we are filing. |
| Unsettled | No authority resolves the point in this jurisdiction, or the authority conflicts and no court has reconciled it. | A judgment of the Supreme Court, or a Full Bench reference answering the question. |
| Contrary authority exists | There is authority against the position which must be dealt with. The memo identifies it and assesses the answer to it. | Whether the contrary authority binds this bench, and whether the distinguishing point is factually available. |
| Not researched | This question was outside the scope of the instruction and I have not looked. Nothing in this memo bears on it. | An instruction to research it. This marker is a request, not an apology. |
The third column is the part that is usually missing and the part that carries the most value. A confidence level with no statement of what would change it is an opinion. A confidence level with a stated trigger is a research plan the senior can act on, immediately, by asking for the one thing that would move the assessment.
Open Items, What Was Not Searched, and the Date
The most professionally valuable paragraph in a research memo is the one that says what has not been done. It is also the paragraph juniors are most reluctant to write, because it reads like an admission. It is the opposite: a memo without it silently claims completeness, and a claim of completeness is the most dangerous sentence in the document.
Say what was outside the scope. Say which courts you searched and which you did not. Say if you searched only reported decisions, or only a date range, or only in English where regional language judgments may exist. Say which factual assumption the analysis rests on. Say what document you asked for and did not receive.
Date the research, always
Good law is a point in time statement, not a property a judgment carries. Put the date the research was completed on the face of the memo. It costs nothing, it tells the next reader, who may well be you in four months, exactly what window was checked, and it is the difference between a memo that ages gracefully and one that quietly becomes wrong.
Finally, an appendix with the actual paragraphs relied on. Copy the operative passages, with paragraph numbers, so a reader can verify the proposition without reopening the file. This does two things: it lets the senior check you in ninety seconds instead of an hour, and it forces you to have actually read the paragraph, because you cannot paste what you have not opened.
What Makes a Memo Unreadable
The chronological narrative
A memo that recounts the research process in order, what was searched, what came up, what was then tried. Interesting to the writer, useless to the reader. Nobody wants the route. They want the destination and enough of the route to check it.
The string citation without propositions
Six cases listed with no statement of what each is for. This looks like thoroughness and functions as an instruction to the reader to go and do the work. If you cannot say in one sentence why a case is in the memo, it should not be in the memo.
Case by case summaries
A section per authority rather than a section per issue. The reader is handed the raw material and left to synthesise. Synthesis is the job, and a memo that skips it has skipped the only part that could not have been done by a search engine.
Hedging that hides the answer
Language so qualified that no position is taken. It feels safe and it is not, because the reader has to guess your view and will guess wrong. An honest arguable is far safer than a fog of it may be submitted.
The answer to a question nobody asked
A memo that reframes the instruction into something the researcher found easier or more interesting. If the question as asked cannot be answered, say so and say why, at the top. Do not quietly substitute a different question.
Formatting that fights the reader
No headings, no table, dense unbroken paragraphs, citations inline in the middle of sentences. The content may be excellent. It will not be read under deadline, which for practical purposes is the same as not existing.
The Drafting Sequence
The order below is deliberately not the order the memo is read in. Writing the short answer first, before the analysis, is what forces you to have a position; writing it last is how memos end up without one.
Restate the question in writing and get it confirmed
Write the question as you understand it and send it back to whoever instructed you. Five minutes here prevents an afternoon answering the wrong question. If the instruction contained an assumption you are unsure of, name it now rather than building the memo on it silently.
Break the question into elements before searching
List what has to be established for the answer to be yes. That list becomes the analysis structure and the search plan at the same time. Researching before this decomposition is how you end up with a folder of interesting cases and no organising principle.
Search, and record propositions as you go, not case names
For every authority you keep, write the one sentence proposition and the paragraph number immediately. A note that carries the case name only is a note you will have to redo. This is also where the appendix gets built, at no extra cost.
Run the forward check on every authority you intend to rely on
Before an authority enters the table, find out how it has been treated since: followed, distinguished, doubted, referred, overruled, or superseded by amendment. An authority that fails this check should come out now, not at the hearing.
Build the authority table and sort it by date
Populate all five columns. Reading your own table in date order usually tells you something the prose did not: that the line has narrowed, that the recent decisions all come from one court, or that nothing has considered the point for a decade.
Write the short answer, and commit to it
Write the answer before the analysis. If you cannot write it, you do not yet have one, and that is information: either more research is needed or the honest answer is unsettled. Assign the confidence marker now, while the research is fresh.
Write the analysis element by element, then the contrary section
Rule, authority, application, counter, for each element. Then state the opposing case at its strongest before you answer it. If the answer feels thin as you write it, adjust the confidence marker rather than strengthening the language.
Write the open items, date the memo, attach the appendix
State what was not searched and what assumption the analysis rests on. Date it. Paste in the paragraphs relied on. Then read from the top and check that the first paragraph alone answers the question, because that is the only part guaranteed to be read.
Using an AI Draft as a Starting Point
A generated draft can be a reasonable scaffold: it will produce a structure, a plausible decomposition into elements, and a set of candidate propositions to test. Used that way it saves an hour of staring at a blank page. Used any other way it produces a document that reads like a memo and has not been researched.
The rule is unconditional. Every citation in a generated draft must be verified against the judgment text before it stays in the memo. Not against another summary, not against a database entry, against the judgment. Confirm that the case exists, that the parties and the date are right, that the paragraph says what the draft attributes to it, and that the proposition is the ratio rather than an observation. A generated draft is also prone to a subtler failure: an accurate proposition attached to the wrong authority, which reads perfectly and cannot be caught by anything except opening the case.
Where a Research Platform Helps, and Where It Does Not
A platform helps with two columns of the authority table and nothing else. CourtMesh runs one search across the Supreme Court, all 25 High Courts, District Courts and Tribunals, sourced from official government portals, with roughly 310 million records keyword-searchable and structured filters for court, year, judge, case type and date range. That is how you assemble the field and fill in the court, bench and date columns quickly.
For the treatment column, citation relationships typed as followed, distinguished, overruled and referred are available across the analysed subset, roughly 2 million records semantically searchable. That is a fast first pass on the forward check, and only a first pass. Deep analysis covers a subset rather than the whole corpus, so an absence of flagged treatment is not evidence that a judgment is clean, and nothing in a citation graph will tell you the section construed has since been amended.
What no tool supplies
The proposition column is yours. So is the confidence marker, the honest assessment of the contrary authority, and the paragraph on what was not searched. Those are the parts of the memo that carry professional judgment, and they are the parts a senior is actually reading for. Every result links to the judgment text, and the record of the issuing court remains the authoritative one: confirm the citation, the date and the paragraph against it before the memo goes out.
Write the answer first
A research memo is judged on whether the first paragraph answers the question and whether the reader can trust the confidence you attached to it. Build the field with one search across the Supreme Court, all 25 High Courts, District Courts and Tribunals from official government portals, filter by court, year, judge, case type and date range, and run the forward check on every authority before it enters your table. Then do the part no tool does: state the proposition, mark the confidence, name the contrary case, and say what you did not search.
Explore CourtMesh
