Earnings call preparation, explained.

What earnings call preparation is: three weeks of work for about 40 minutes of unscripted questions

From quiet period to transcript: one quarter's call in five steps

1. The quarter closesMany companies start a quiet period, limiting talk with investors until results are out. In the example, it starts on October 1.
2. Questions get predicted and answers writtenNotes, past calls and peer news become a ranked list, and each likely question gets a short spoken answer from the public record. This step is what this guide is about.
3. Finance and legal signEvery number ties to the final release draft, and every sentence that goes past the record is decided by a person. In the example, the disclosure committee, the group that considers what must be disclosed and when, meets on October 22.
4. Executives rehearseThe CEO, CFO and head of science practice answers aloud, against the clock, with follow-up questions. In the example, on October 20 and 26.
5. The release, then the callThe release goes out, the call opens to everyone, and what was prepared meets what is asked. In the example, 7:00 am and 8:30 am Eastern on October 29.

The company, people and dates in this example are fictional.

Why an open call is itself a lawful disclosure, and three more surprises about the Q&A

Q&A book, material, tie-out, non-GAAP: four words this work runs on

Fourteen terms are defined in the one-page summary.

How long earnings calls last, who speaks and how many analyst questions to expect

In shortExpect about an hour, with roughly two-thirds of it spent on questions. The best published count comes from 20 large Dutch companies: 64 minutes, nine analysts and about 19 questions. Older survey data from the US points the same way, with most calls running 46 to 60 minutes.

Published data on call structure is thin, so keep in mind how small these samples are. Researchers who timed the results calls of the 20 non-financial companies in Amsterdam's AEX index found an average of 64 minutes, from 35 to 95. The presentation took 23 minutes and the Q&A 41.7

A 2014 National Investor Relations Institute survey, as summarized by Davis Polk, found calls of 46 to 60 minutes at 68% of companies, with larger companies running longer.11 We found no newer public US figure, since the institute's later reports are for its members.

The same study counted an average of nine analysts on a call, from four to eighteen, and 19.3 questions, from 10 to 33. So 41 minutes of Q&A across 19 questions leaves about two minutes for each question, answer and follow-up, by our arithmetic. That is why a rehearsal card, the short prompt an executive practices from, holds a single opening line and a few numbers. Sell-side analysts ask most of the questions: 83% in the 2014 US survey.11

What analysts ask about on a results call, and what they get back

Of the 385 questions in that study, 41% were about financial items such as revenue, margins and cash flow, and the rest covered operations, strategy and governance.7 A larger study of conference call transcripts found that about 11% of analyst questions draw a non-answer, a rate that is stable over time and similar across industries. Harder, more negative and more detailed questions draw more of them, and questions about proprietary information do too.12

A prepared non-answer is often the right answer. "We will give 2027 guidance in January, as we always do" tells the analyst nothing new and sounds like a policy. The skill lies in having that sentence ready, with the words the company has used before, so nobody improvises a number under pressure.

Who gets called on matters as well. Management picks from the queue. One study found that analysts with more favorable recommendations are more likely to get a question in.13 Another found that firms favoring bullish analysts tend to underperform later, with more negative surprises and restatements.14 A team preparing for a call should plan for the skeptics, and should call on them.

What Regulation FD, Regulation G and the safe harbor ask of an earnings call, as of November 2026

In shortRegulation FD bars telling market professionals something material before telling everyone. An open call satisfies it, so the rule's force falls mostly on private conversations. Three other rules govern what you say on the call itself: Form 8-K's conditions for the call, Regulation G for adjusted figures, and the safe harbor for statements about the future.

Regulation FD, in force since October 23, 2000, says that when a company or someone acting for it discloses material nonpublic information to a broker, an adviser, a fund or a holder likely to trade on it, the company must make that information public. The disclosure must be simultaneous if it was intentional, and prompt if it was not.15

"Promptly" means as soon as reasonably practicable, and no later than the later of 24 hours or the next opening of trading on the New York Stock Exchange.16 So a slip noticed at 5:00 pm on a Friday could run to the Monday open.

The rule excludes people who owe the company a duty of trust, such as its lawyers and bankers, and people who expressly agree to keep the information confidential. It does not apply to foreign private issuers, non-US companies that hold that SEC status.15,16 Public disclosure can be a Form 8-K, the report a company files with the SEC when something significant happens, or any other method "reasonably designed to provide broad, non-exclusionary distribution".16

Form 8-K Item 2.02, Regulation G and the oral safe harbor on a live call

Three staff positions are worth knowing by heart. Confirming a prior forecast privately can itself be material, depending on whether it adds anything and how late in the quarter it comes.17 Going through an analyst's model, the spreadsheet forecast behind their estimates, is allowed for public facts and immaterial detail, but "may not" be a vehicle for passing along material nonpublic information.17 And the adopting release warns that a company "cannot render material information immaterial simply by breaking it into ostensibly non-material pieces".2

The quiet period before earnings: a custom each company defines

A quiet period is custom too. It is the stretch before results when a company limits its contact with investors. ICR, an IR adviser, notes that there "isn't a standard length" and that each company sets its own policy.26 Whatever you pick, write it down and keep to it.

Seven Regulation FD cases, two of them at drug companies, and what each one teaches an IR team

In shortThe usual pattern is a private conversation, often near the end of a quarter and often about earnings, and in several cases a share price move follows. Penalties run from $50,000 for one person to $6.25 million for one company. Stand-alone cases are few, and we found none in 2025 or 2026. The drug company cases are about tone and about FDA.

These are settled matters, generally without admitting or denying the findings. A law firm's review of the AT&T case recommends written protocols, advance review of topics for one-on-one calls, and a process for unintentional slips.30

What may feed a question and what may feed an answer: analyst notes, transcripts, filings and release drafts

In shortA Q&A book reads two kinds of material and keeps them apart. Analyst notes, peer news and investor feedback tell you what will be asked. Filings, past calls, investor day decks and this quarter's release and remarks tell you what you may say. Mixing the two is how a broker's estimate ends up in an executive's mouth.

Think of it as two shelves. The first shelf predicts: a preview from a covering analyst often says outright what they will ask, and past transcripts show who keeps returning to what. The second shelf answers: anything already public, such as the annual report (Form 10-K), quarterly reports (Form 10-Q) and press releases, plus what will become public on results morning. A draft release sits on the second shelf with a date on it. It stays confidential until it goes out, but the answer will be spoken after that moment.

Sell-side notes carry a second constraint. They are usually licensed through a research aggregator, a service that collects brokers' research, under terms your team should read before copying them into any tool. They also carry the analyst's own estimates, and a company repeating an estimate may seem to endorse it. Using a note to learn the question and never to supply the answer solves both.

Public, public on results day, licensed, internal: four labels every source in the book needs

How to rank analyst questions: probability against stakes, worked through on a rival's Phase 3 news

In shortA long list helps nobody. Put each question on two scales: how likely it is to be asked, and how much damage a poor answer does. Prepare hardest where both are high. Show what pushed each probability up, so a person can argue with it, and move the question when news changes the answer.

Probability comes from signals. An analyst writing "we will ask" is a strong one. A question asked on each of the last three calls is another. So are a peer's news and what investors told you in meetings. No standard formula exists, and no public study we found tests one for this purpose, so treat any percentage as a structured judgment. Its value lies in showing the reasoning.

Stakes come from what is exposed. Is there a public target at risk? Would a precise answer amount to new guidance? Is there a transaction, a legal matter or an undisclosed date nearby? A routine question on what drove the quarter's growth may be near certain and low in stakes. A question about a deal may be less likely and very high.

A worked example: what one rival's release does to two questions

At Northwind Bio, the question on oral competition sits at 44% and medium stakes on Tuesday. On Wednesday at 7:58 am, a watched peer reports Phase 3 data for a once-daily pill. Two things change. An analyst had written that if the data looked competitive, he would ask whether the $20 billion 2030 target still holds, so the likely wording changes and a public target is now on the line. Stakes go to high and probability to 82%.

The same release names pricing policy as the rival's main 2027 uncertainty, so the question on 2027 gross-to-net, the gap between a drug's list price and what the maker keeps after rebates, moves from 71% to 76%.

The split of signals shows why. Before the news, 62% of the oral question's probability came from analyst notes and none from peer news. After it, peer news accounts for 56%. In points, that is 0.56 × 82, or about 46 of the 82, against about 25 from notes. The team can see that the jump rests on one event, and it can ask whether the event deserves that much weight.

Keep score after each call. Count how many of the questions asked were in your book, and how many of those you ranked highly. Those two numbers tell you whether your ranking is getting better, and they are easy to get from the transcript. We found no public benchmark for either, so your first quarter is your baseline.

Writing an answer from the public record: precision creep, stale numbers and non-GAAP traps

In shortTwo errors do most of the harm. Precision creep is a sentence one step sharper than anything said before: a month for a half-year, a floor for a curve. A stale number is one that survived a change in the release draft. Both are invisible unless you compare each sentence with its source.

Precision creep: when "first half of 2027" turns into "March 2027"

Writers want to be helpful, and a good writer knows internal plans. So drafts drift. Analysts also ask for exactly that extra step: "could it be the first quarter?" The Q&A book has to hold the line the company chose, because a month reveals the plan, narrows the market's expectations and can later be shown to be wrong.

Sentences from the fictional example.

The arithmetic of a stale growth rate: why "up 26%" on $5.31 billion cannot be right

In the example, a draft answer says Veltrimab sales were "$5.31 billion in the quarter, up 26%". Release draft v7 says up 24.1%. A rebate adjustment arrived between drafts. The sentence mixes two drafts, and you can prove it. If $5.31 billion is 24.1% growth, last year's quarter was $5.31B ÷ 1.241, or about $4.28 billion. Growth of 26% on that base would be $4.28B × 1.26, or about $5.39 billion. That is $80 million of sales the release does not report.

Fictional example. The derived figures are our arithmetic on the sample's numbers.

A tie-out catches this when it treats a percentage as a number with a base, and when it runs again after every new draft. In the example the book holds 188 numbers across 41 answers. Each new release draft puts every one of them back in doubt until it is checked.

Adjusted earnings in pharma: the acquired R&D rule that changed in 2022

Drug companies buy pipeline assets, and they used to strip those costs from adjusted earnings. In April 2022, according to BDO, SEC staff said they would object to non-GAAP measures that exclude acquired in-process research and development (IPR&D) in an asset acquisition, or upfront and milestone payments under R&D arrangements, as "normal, recurring operating expenses for pharmaceutical companies".19 Bristol-Myers Squibb, for one, changed its non-GAAP results that quarter and cited comments from the SEC.31

So an answer on adjusted earnings per share has a specific trap: a deal's upfront payment now sits inside the figure, and analysts will ask you to explain it. Regulation G's bar on misleading measures applies to the discussion around the number as well.10

Why biotech and pharma earnings calls are harder: trial timing, FDA letters, Medicare negotiation, tariffs

In shortDrug companies depend on events that go one way or the other, such as a trial result, and on government decisions. Analysts press for trial dates, for how a meeting with FDA went, and for what a negotiated price will do. Newer topics have joined them: complete response letters, which FDA has published since 2025, and tariffs and most-favored-nation pricing deals. Each topic has its own careful wording.

That calendar explains why the example's card says "Do not estimate the 2028 impact". On October 29, the company may know more than the public does, and CMS must publish any agreed price by November 30. A careful answer repeats the filing, says when the price will be published, and stops.

Nine ways earnings call answers go wrong, from a number off the last draft to a call-back nobody logged

In shortMost errors come from documents that went out of date, or from the time after the call when people relax. Sort them by how often they happen and how much harm they do, and the highest priorities are the ones a person cannot see by reading the draft alone: precision creep, stale numbers and private follow-ups.

1. A stale number. The answer predates the final release draft, and nobody re-checked percentages.
2. Precision creep. A sentence is sharper than the record, because the writer knew the internal plan.
3. An answer copied forward. Last quarter's words now conflict with this quarter's facts.
4. A rival's news handled by email. The question that became urgent last week has no rehearsed answer.
5. Quoting the Street. An answer repeats an analyst's estimate, or a note's text reaches an executive's card.
6. An adjective about FDA. "Very productive" can be material, as TherapeuticsMD shows, and FDA's letter may now be published.4,22
7. An unprepared call-back. After the call, analysts ring IR for "color". These are private conversations, where Regulation FD's limits bite. Staff need a rule for what may be said and a log of it.17,30
8. A slip nobody escalates. The clock for a non-intentional disclosure starts when a senior official learns of it, so a person has to know whom to tell.16
9. Reading the card aloud. Answers that follow a script closely can read badly to the market, so prepare facts and lines, and leave the sentences to the speaker.6,7

What a bad answer can cost a life sciences company: lawsuits, enforcement, share price and a leaked release

In shortDirect penalties under Regulation FD are modest. The larger exposure is a securities suit, and life sciences companies drew 47 of 207 federal filings in 2025. Plaintiffs' firms now scan calls with software. Draft results are also valuable inside information, and thieves have stolen them before.

Dechert advises that disclosure plans cover what executives say on analyst calls, and it notes that plaintiffs' firms use AI tools to scan earnings calls, so companies should audit their own statements.36 An answer about FDA timing is the kind of sentence that gets compared with later events.

Many of these suits do not survive: of 14 regulatory-hurdle decisions Dechert counted in 2025, 10 were dismissed in full. Defending one is still costly. We found no public figure for the cost tied to a single answer on a call.

The call also moves markets on its own. Researchers report that both the presentation and the Q&A give investors new information.7 The SEC noted in 2026 that markets generally react when information is first disclosed, and less when it is repeated.1 What your executive says in that hour is the first telling for anything the release left out.

The book itself is a target. It holds draft results, internal thinking and legal judgments. In the 2015 case, the SEC alleged hackers took earnings announcements from newswire services over about five years and fed them to traders.37 Any system holding release drafts needs access limits to match.

AI for investor relations: where it already helps with earnings prep and where a person must decide

In shortAdoption rose fast. Surveys report anywhere from 42% to more than half of IR teams using AI. Teams mainly use it for drafting and research, and simulating analyst questions is the use practitioners back most in one survey. The limits: models misread financial documents, materiality stays a person's judgment, and a tool that reads draft results holds inside information.

What regulators have said about AI in earnings preparation: very little so far

We found no SEC rule written for AI in investor communications. The existing duties apply whoever or whatever drafts: antifraud rules, Regulation FD, disclosure controls.

The life sciences rules you may hear about were written for other work, in our reading. FDA's January 2025 draft guidance concerns AI that supports regulatory decisions on drugs.41 Part 11, FDA's rule on electronic records, is tied to records required by its other rules, known as predicate rules.42

Draft EU GMP Annex 22, part of the EU's good manufacturing practice rules, concerns critical manufacturing uses, and it remains a draft: EudraLex Volume 4, which holds those rules, listed no Annex 22 on November 3, 2026.43,44 The EU AI Act's high-risk dates moved to December 2, 2027 and August 2, 2028.45 Ask your own counsel to confirm how each applies to you.

Three cautions follow from the evidence. FinanceBench is from 2023 and models have improved, yet it shows why every figure needs a check that does not depend on the model.40 As one practitioner in the Gladstone write-up put it, AI can "check the work product, not generate it", and the same write-up recommends "human approval before materials reach the market".8 And an AI tool that reads draft results is holding inside information, so where it runs and who can reach it are disclosure-control questions.

Semiannual reporting, the EU's new inside information rules and foreign issuers' earnings calls

In shortQuarterly filing may become optional in the US, though a proposal is all that exists so far. Calls were never required, and companies would still choose whether to hold them. Outside the US, Regulation FD gives way to other regimes. The EU changed how it treats intermediate steps from June 5, 2026, with clinical trials and marketing authorizations among its listed examples.

Semiannual reporting. On May 5, 2026, the SEC proposed letting companies report twice a year in place of three Forms 10-Q: a new Form 10-S for the first half, then the Form 10-K, with the choice made by a box on the Form 10-K. Comments closed on July 6, and we found no final rule as of November 3, 2026.1

Apart from technical edits to Item 2.02, the proposal makes no general change to the rules on earnings releases or guidance, and it notes that many companies hold calls. In our reading, fewer scheduled calls would pack more topics into each one and raise the stakes of any contact with analysts in between.

Foreign private issuers. Regulation FD excludes them.16 Several global drug companies report under this status, and they answer to home rules, plus US antifraud rules where their shares trade.

The EU. Under the Market Abuse Regulation, the EU's rulebook on inside information, as amended by the Listing Act, companies need not announce intermediate steps in a drawn-out process from June 5, 2026, and only the final event is disclosed. Those steps can still be inside information, so confidentiality and insider lists continue. A company may delay only if the withheld fact does not contradict its latest public statement on the matter.46

Commission Delegated Regulation (EU) 2026/789, dated April 8, 2026, was published in the Official Journal on July 16, 2026, and applies from the third day after that. It lists typical final events, including the conclusion of clinical trials and authorization to commercialize medicines, and forms of communication that count for the contradiction test, including publicly accessible pre-close calls (calls held ahead of results), roadshows and webinars.47

Japan. A Fair Disclosure Rule has applied since April 1, 2018.48 Other markets have their own rules, so a global company's book needs a note on who may say what, and where.

How IR teams prepare today: the controller, deputy counsel, CFO office and the hand-offs in between

In shortFive roles carry the work over three to five weeks, passing Word files and spreadsheets between them. The most time goes to drafting answers and reading. In the example's step estimate, a large company spends about 150 hours a quarter. Companies do it in house, with an adviser's help, or increasingly with software.

Counsel's part is well established: in the 2014 survey, in-house counsel reviewed the script at 71% of companies and helped develop it at 37%.11 The quoted lines are the example's fictional characters.

Earnings call rehearsal and mock Q&A in one adviser's five-week plan

ICR's five-week plan sets themes and compiles Q&A topics four weeks out, reviews Q&A with bullet-point responses in the final week, and holds a final mock Q&A a day before the call.49 BNY's practice note adds that pre-recorded remarks leave leaders better prepared for the live Q&A.50 The three IR officers in the Dutch study all described a briefing on anticipated questions.7

In house, with an adviser or with software: how teams build a Q&A book

Northwind Bio's Q3 2026 call: a peer's Phase 3 news at 7:58 am, a matrix re-ranked, 41 answers signed in order

In shortThis is the example from the top of the page, inside the product. Six AI agents read the notes, calls and news, and rank 41 predicted questions. Lena Ortiz, the VP of investor relations, re-ranks when a rival reports. The controller, deputy general counsel and disclosure committee then sign in that order, and every flag is decided by a person.

1

Every likely analyst question on one matrix of probability against stakes

Lena opens the question matrix on Wednesday, October 7, 22 days before the call. The 16 questions above 20% sit in a grid: probability across the top, stakes down the side. The top-right cell reads "Prepare hardest". Each chip's dot gives its status: disclosure flag, drafted, in review or approved.

She picks the question on the Talvoren 2027 floor, the level where that medicine's falling sales stop, at 88%. The panel shows what each signal adds: analyst notes 44%, past calls 41%, investor feedback 15%. Below that sit the evidence, the likely askers and the draft answer. Rachel Kim of Harbor & Pike wrote that she expects "the Q&A to focus on where Talvoren bottoms in 2027".

2

A peer's Phase 3 news arrives, and one click re-ranks the questions

At 7:58 am, a banner reports that Tarnwell Pharma, one of three watched peers, has published oral Phase 3 results in plaque psoriasis. It says which questions may move before anyone asks. Lena clicks "Re-rank with this news".

The steps run where she can see them. The Research Reader reads the release, the Question Forecaster matches it to all 41 questions, the Answer Drafter re-drafts from the disclosure library, and the Disclosure Guard checks the result. The banner gives 18.6 seconds for the run in this sample. The last step waits for her.

3

Oral competition jumps from 44% to 82% and moves to "prepare hardest"

The chip slides across the grid and leaves a marker: "Q-16 was here · 44%". Its stakes go from medium to high, because the $20 billion 2030 target has been on the record since investor day. The 2027 gross-to-net question rises from 71% to 76%, since the peer named pricing policy as its main uncertainty.

Peer news now supplies 56% of the oral question's probability. The panel says when it moved, from what, and why.

4

Three sentences for the CEO, each cited to a public passage

The question is reworded the way Victor Salas of Thornlea Capital Markets is likely to put it: does a once-daily pill change your view of Veltrimab above $20 billion in 2030? The answer for Dana Okafor, the CEO, comes only from the investor day presentation and prepared remarks. The 2030 outlook already assumes oral competitors from 2028, Veltrimab leads new-to-brand prescriptions in psoriasis, and sales above $20 billion in 2030 are still expected.

The checks read "Every sentence cited", "Approved sources only", "Numbers tie to release v7", "Nothing new or material" and "Forward-looking covered". The rehearsal card adds the line not to cross. Lena approves it.

5

"March 2027" is flagged: a month is sharper than "the first half of 2027"

Anika Bose of Tidemere Equity Research will press for the date of the VLT-302 study results. The draft for Dr. Maya Chen, the chief scientific officer, says "We now expect topline results in March 2027." The Disclosure Guard flags it as new material information, shows the press release of September 14 with "the first half of 2027" marked, and proposes that wording with its citation.

Lena can use the public wording, or keep it and ask legal. The agent never clears its own flag. On an open, announced call the month would reach everyone at once, which usually satisfies Regulation FD's selective disclosure rule. The problem left over is a new date that nobody decided to announce. Your counsel will want to make that decision on purpose, and keep the words from reappearing in a private call-back.

6

Rehearsal in the analyst's voice, timed to 30 seconds

For Rehearsal 1 on October 20, the Rehearsal Coach asks the question as Rachel Kim would and times Omar Haddad, the CFO, against a 30-second target. Then it throws the follow-up: "So is it fair to model a 2027 decline similar to 2026?"

The card gives the opening line, three numbers ($1.12B, −41.3%, about 45% US erosion), the bridge, a sentence that moves on to Veltrimab's growth, and the line not to cross: "Do not give a 2027 Talvoren number or a floor." While a flag is open, the screen warns him to practice the public wording.

7

188 numbers tied to release draft v7, and one that was not

Lena submits the book. Tomás Reyes, the corporate controller, runs the tie-out across 41 answers and 188 numbers. 187 tie. One answer still says Veltrimab was "up 26%", from draft v6, before a rebate adjustment. Draft v7 says 24.1%.

He applies the v7 number and signs. Adjusted earnings per share, operating margin and R&D are checked against the reconciliation tables in the same step.

8

No legal signature while an answer carries wording that is not public

Priya Ranvelle, the deputy general counsel, cannot sign with three flags open: a 2027 Talvoren decline and floor, "late-stage talks for an obesity asset", and "March 2027". The screen shows each next to the public wording.

She uses the public wording for all three and signs. The original sentences stay on the audit trail, in case anyone later asks what was considered and why it changed.

9

Locked for the call: version 1.0, 41 questions, no flags open

The disclosure committee, chaired by Omar Haddad, meets on October 22 in the plan. Twelve answers still lack IR approval. Each has been cited and passed through the Disclosure Guard, finance and legal. The committee approves them as they stand, and it could send them back. The book locks as one record: 41 questions, 41 approved, no disclosure flags open, version 1.0.

The rehearsal cards go to the CEO, CFO and chief scientific officer. Any change from here needs the committee.

The design goals for a live deployment are 46 hours of IR preparation a quarter against about 150 by hand (the solution's own step estimate), 89% of questions asked on the last call already in the book, and nine flags caught before legal review with none reaching the call. These are targets, and the sample dashboard shows them on fictional quarters. Measure your own.

The ranked question list, the Street, the source passage, the dashboard and your disclosure settings

In shortThe walkthrough above followed one quarter's main path. These screens show the rest: signals and the run-up to the call, the ranked table, the moved question in full, a source document, covering analysts and their estimates, the rehearsal plan, the committee's last check, the trail, the dashboard and the settings that hold your policy.

Six agents read, rank and check; the IR lead, controller, counsel and disclosure committee decide

In shortEach agent has one job and limits it does not cross. Agents read notes and draft from approved documents, and none can clear a flag, sign or lock. People approve each answer, tie out and sign, decide every flag and lock the book, in an order you set.

The six earnings call agents, and what each must never do

Who signs a Q&A book: the order from investor relations to the disclosure committee

Maker and checker stay apart: agents draft and check, and people approve and sign. Every re-rank, edit, flag decision and signature is on the audit trail with who and when. Draft results stay on your company's platform. These are controls that support your own disclosure controls and procedures, and the legal judgment stays yours. Botminds is SOC 2 Type II audited and manages information security to ISO 27001.

Lena, Omar, Tomás, Priya and Dana: five people who carry one results call

From sell-side notes and draft releases to a signed book: what the solution reads, writes and connects to

In shortIt reads what IR already reads, keeps sources that predict apart from sources that answer, and produces a ranked matrix, cited drafts, flags with public wording, rehearsal cards and a signed book. It connects to your research feed, transcript provider and IR drive, and re-checks answers whenever a new draft arrives.

Research feeds, transcripts and the IR drive: systems an earnings prep rollout connects to

The solution scales across products, analysts and quarters, inside the roles and access rules you set, with a full audit trail and export. The settings hold your order of signatures, the switchable rules, the probability that puts a question in the book, and how news is handled. Before you connect a research feed, check your license with the aggregator.

What hand-built earnings prep costs you: IR hours, stale numbers and the question nobody rehearsed

In shortThe hours are the part you can count, and they are modest for one company. The larger exposure is one unplanned sentence, a number from the wrong draft, or a surprise on the question that a rival's news made urgent. Published penalties frame the risk without telling you your odds, so start by measuring your own last call.

Scoring your last four calls: a one-week baseline for your Q&A book

The surprise on the call, the expectation you set by accident, and other costs a return model misses

What automating the Q&A book returns: IR hours, executive rehearsal time and fewer late rewrites against the tool's cost

In shortPut in your own numbers. Every default is an assumption to replace. For one company, hours alone roughly cover a modest yearly cost, so the case rests on better prediction and fewer unplanned sentences. A small company with few analysts, one product and a skilled adviser may get more from writing down its method.

IR hours, rehearsal time, tie-out and prediction: where earnings prep can pay back

Calls a year, hours per call and executive time: your earnings prep numbers

On these defaults, four calls release $68,480 of time against a $60,000 cost, so about 3.5 calls of value cover it, which the calculator rounds to 4. Change the cost to $120,000 and time alone no longer pays. Then the decision rests on the risk you wish to reduce, and that is worth saying out loud in your business case.

How ready your IR team is: a disclosure library, a written flag policy, a scored last call and a clear signer

In shortPick the answer closest to your team today. Eight questions score 0 to 2, and the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

A low score means gathering the disclosure library and scoring old calls first

What to ask earnings call software about citations, draft results, flags and analysts' licensed research

In shortAsk every vendor the same questions, us included. The answers show where each sentence comes from, what happens when a draft changes, who decides a flag, where your draft results go, and how you would score the tool on your own past calls.

We found no published independent test of any vendor's accuracy at predicting analyst questions as of November 2026. A replay on your own past call is the fairest comparison, and it is easy to do honestly, because the transcript already holds the answer key.

Replay your last earnings call using only what was public before it, then run one quarter in parallel

In shortScore your last four calls, gather the library, replay one call blind and run one live quarter beside your usual book. None of these steps commits you to a vendor. The scoring and the library improve your next call either way.

1. Score your last four calls. Mark each question asked as in the book, ranked high, or a surprise.
2. Gather the library, with public status and date on each document, and write down your policy on guidance and call-backs.
3. Replay one past call blind, using only documents dated before it. Score prediction against the transcript and flags against what you know now.
4. Run one live quarter in parallel, with your usual book still going to the CFO, and compare the two at the lock.
5. Decide with five numbers: questions in the book, questions in the top tier, flags caught early, numbers caught by the tie-out, and hours.

Questions in the book, top-tier hits and flags: how to score an earnings prep pilot

Earnings call preparation questions answered: Q&A books, Regulation FD, quiet periods, safe harbor and AI

In shortQuestions IR teams and buyers search for, each answered on its own.