Expert call notes in due diligence and what to record

Expert call notes are unusual among research documents in that the rules about what they may contain are stricter than the rules about how they are stored. Most teams have this the wrong way round. There is a shared drive, a naming convention and a retention policy, and there is no settled answer to the question of what an analyst should do when a former employee starts describing something they clearly should not be describing.

That question comes up more often than the storage question, and it comes up live, in the middle of a call, with a clock running.

What belongs in expert call notes

Anything the expert can properly discuss: industry structure, how a process works, what buyers weigh when choosing a vendor, how a market has changed, what a role involves day to day. What does not belong is anything that would be material and nonpublic about a specific issuer, along with anything covered by the expert's own confidentiality obligations to a current or former employer. The CFA Institute's Code of Ethics and Standards of Professional Conduct sets out Standard II(A) on material nonpublic information, and the prohibition it describes is not limited to insiders. It applies to anyone who comes into possession of the information.

The practical consequence for note taking is that some things said on a call should not be written down at all, and the analyst has to recognise them while listening.

None of what follows is legal or compliance advice. Every firm has its own policy, and where this article and that policy disagree, the policy wins.

Why the notes are worse than the calls

An expert call runs an hour. The analyst is running the conversation, following up, watching the time, and deciding in real time whether the last answer strayed somewhere it should not have gone. Writing is the fourth or fifth priority and it shows.

What tends to survive is the parts that confirmed what the analyst already thought. That is not dishonesty, it is attention. A remark that fits the thesis gets written down because it registers as significant. A remark that cuts against it registers as noise from someone who may not be well informed, and it goes unrecorded.

Then the notes go into a memo, the memo goes to an investment committee, and the committee is looking at a document that has been filtered twice by the same person without anyone intending it.

The second problem is volume. A diligence process might involve fifteen expert calls plus management meetings plus channel checks. No individual remembers fifteen hours of conversation, and the comparison that matters is across calls rather than within any one of them. Four experts independently describing the same constraint is a finding. One expert describing it is a data point.

Recording, where it is permitted

Some firms record expert calls as a matter of policy. Others prohibit it. Many expert network agreements address it directly, and several require a chaperone on the line instead.

Where recording is permitted, the sequence is not negotiable. The expert is told before the call begins, the agreement is on the record, and the purpose and retention period are stated. An expert who declines is not recorded, and that call proceeds on notes alone.

Where it is prohibited, the technique that remains available is recording your own summary immediately after the call ends. Nobody else's voice is captured, no agreement is breached, and the recall is far better at minute two than at hour six. Five minutes of dictation after each call produces a materially better record than a set of bullets typed during it.

That is worth saying clearly, because teams often treat the recording question as binary and conclude that if they cannot record the expert there is nothing to be done.

Compliance obligations do not go away because the record is better

A more complete record cuts both ways, and presenting transcription as a pure gain would be dishonest.

If a call strayed into territory it should not have, a transcript preserves that fact with a timestamp. Some analysts read this as a reason not to transcribe. The better reading is that it is a reason to end calls properly when they stray, escalate to compliance the same day, and follow the firm's process, which is what should happen regardless of what was written down.

Retention policies apply to transcripts exactly as they apply to notes, and in some firms transcripts fall into a category with a different retention period. That is a question to settle with compliance before building a workflow, not after.

The cross-border case

Diligence on a business with operations in several countries produces calls in several languages, and this is where record quality degrades fastest.

A call conducted in Portuguese with a former distribution manager in São Paulo, summarised in English by whoever was on the line, arrives at the investment committee as three bullets. The committee cannot assess the reliability of the summary because nobody on it speaks Portuguese, and the analyst who wrote it may have been operating in their second or third language too.

The workable arrangement is the same one that applies to any research artefact crossing a language boundary. Keep a transcript in the language of the call, checked by someone who speaks it, and translate that document separately rather than merging the two steps. A team handling Portuguese-language calls can work in Vomo's Portuguese edition, which keeps the correction pass in the hands of the person best placed to do it, with the English version produced afterwards as a second document.

Two documents instead of one. The cost is small and the thing it prevents is a committee treating a compressed second-hand summary as though it were testimony.

Reading fifteen calls as one body of evidence

The step most diligence processes skip is the one where the calls get read against each other rather than one at a time.

Done manually this is expensive, which is why it gets skipped. The analyst would have to hold fifteen conversations in mind simultaneously and notice that expert four and expert eleven described the same procurement cycle differently. In practice they notice it only if the discrepancy happens to be striking, and only if both calls were recent.

Done against a set of transcripts it becomes an ordinary search task. Which experts mentioned the renewal cycle. Which of them gave a number. Did anybody contradict the number that ended up in the model.

The last question is the valuable one and it is almost never asked, because until the transcripts exist there is no way to ask it. A figure enters a model from one call, gets treated as an input, and is never tested against the other fourteen conversations that might have qualified it.

The right hand column is not a claim that better records produce better decisions on their own. Committees can ignore evidence in any format. It does change what an analyst is able to answer when asked, and it removes the answer that damages credibility most, which is that the question cannot be checked.

What actually improves the notes

Three habits do most of the work, and none of them require a tool.

State at the start of the recording, or at the top of the notes, which company and which workstream the call belongs to. Diligence processes generate a lot of files and they all look alike three weeks later.

Write down the disagreements. If an expert says something that contradicts the thesis, that sentence has more information in it than the four that supported it, and it is the one most likely to be omitted by accident.

Separate what the expert asserted from what the analyst inferred. These blur quickly in shorthand notes, and by the time the memo is written the inference has quietly acquired the authority of a source.

A fourth habit belongs to the person running the process rather than the person on the call. Record how the expert came to know what they are describing. A former employee who left the business in 2019 and a current distributor competing for the same shelf space are both useful, and they are useful in different ways, and neither of those ways survives being reduced to the word expert in a footnote. Where the basis of knowledge is captured at the time, a reader two years later can weigh the testimony. Where it is not, every source in the file carries the same undifferentiated authority, which is the same as none.

Where this leaves the memo

The memo still gets written by a person, and it should. Nobody wants an investment committee reading fifteen raw transcripts.

The difference is that when a committee member asks whether any of the experts raised a concern about customer concentration, the answer takes a minute rather than depending on what the analyst happens to remember. And when the position is reviewed two years later by people who were not involved, the underlying record is still readable.