Investor communications testing: the equity story explained

An earnings script can make sense inside the company while leaving the questions investors care about unanswered. Our hyperscaler client expected attention to settle on margins. Revenue was about to double year on year, while rising memory costs were forcing the margin guide down, according to the Artificial Societies earnings-call case study. The team had prepared remarks and a guidance range, but selective-disclosure constraints prevented it from testing those materials with the audience concerned.

Internal reviews and advisory consultations draw on a handful of perspectives. Simulation let the hyperscaler team examine the story with audiences built around the stock. We would judge an earnings rehearsal by whether it challenges the team’s assumptions about the questions, before judging the polish of the script. In this engagement, the difference between the expected margin discussion and the simulated focus on demand and financing was the useful finding.

The evidence for the method is our 86% distribution accuracy against a 91% human ceiling across 1,000 surveys, explained in full on the accuracy page. These are survey benchmarks; the earnings-call evidence is the comparison between anticipated questions and the discussion that followed.

An equity story test starts with the audiences whose interpretation you need to understand. For the hyperscaler, we built standing societies around the stock and matched each persona to a real individual through public data and social activity. The design covered people asking questions on the call and people interpreting the announcement afterwards. It did not involve showing draft material to those real individuals.

Source: Artificial Societies, hyperscaler earnings-call case study. The study reports 2,400+ AI personas in total.

We began with the question each persona in Sell-Side Analysts was likely to raise. We then tested the guidance and alternative prepared remarks with Buy-Side Investors, before examining how the announcement would travel through the surrounding audiences and the press. That sequence tested the script choice alongside the questions and coverage the team needed to prepare for.

The test produced nearly 125,000 individual responses, according to the Artificial Societies earnings-call case study. We captured results at aggregate and persona level, so the team could examine differences by audience, question and script version. For your equity story, that means a draft can be assessed against the concerns it provokes, alongside the overall response it receives. A preferred script remains a preference among the versions tested.

The hyperscaler’s pre-call briefing put demand durability and financing commitments in front of a team preparing for margin questions. Simulated analysts gave those topics greater attention than the margin reset, and the published case study reports that the live questions followed that pattern. The team also had likely questions phrased for the individual analysts expected on the call. Those outputs gave preparation a concrete object: the question to answer, with the analyst’s wording attached.

The Financial Media society extended the rehearsal beyond the call. It anticipated growth as the headline, margins as a cost story and financing as the risk drawing attention. The case study reports that this matched the following morning’s coverage. Script testing also identified the prepared remarks preferred by the simulated investors. We treat those findings as evidence for choosing and rehearsing communications; they do not establish what a security will do.

The value of testing competing narratives also appears in our separate Teneo case study. In late 2025, we tested six technology narratives for a major US company with Teneo. That engagement concerned a technology strategy, so the earnings-call account remains the relevant evidence for investor preparation.

An equity story is the capital-markets narrative behind an initial public offering (IPO), an earnings message, a strategic pivot or a restructuring. Testing asks how audiences interpret that narrative: their conviction, the risks they perceive and the questions they want answered. Our message-testing page covers the wider narrative-testing method; an investor engagement gives it a specific audience and decision.

Earnings and guidance are the worked applications here. The hyperscaler tested the guidance range alongside alternative remarks, then examined likely questions and press treatment. For a guidance change, we would keep the proposed numbers and their explanation together in the test. Assessing the explanation alone would leave out part of the material the investor has to interpret.

An IPO roadshow or a mergers and acquisitions (M&A) announcement offers a different decision to test. For an IPO, the proposed application is the roadshow narrative; for M&A, it is how investor segments interpret the announcement. These are possible applications, separate from the delivered earnings-call engagement. We would scope them around the proposed narrative and the investor audience, with no promise about pricing, valuation or trading. For the broader stakeholder questions around a transaction or strategic change, see testing corporate strategy announcements.

Selective disclosure was the hyperscaler team’s stated constraint. The operational answer was to test with simulated audiences: the earnings-call case study records that drafts, numbers and scripts remained inside our private, secured environment. That is a claim about how we handled the engagement. Decisions about an issuer’s disclosure obligations remain outside the simulation’s scope.

For US context, the Securities and Exchange Commission adopted Regulation FD in its Selective Disclosure and Insider Trading release of 10 August 2000 (opens in a new tab). The general rule on selective disclosure (opens in a new tab) addresses material nonpublic information disclosed by an issuer, or someone acting on its behalf, to specified recipients. It calls for simultaneous public disclosure when disclosure is intentional and prompt public disclosure when it is non-intentional.

The recipient categories include investment advisers and institutional investment managers, as well as securities holders in circumstances where trading on the information is reasonably foreseeable. Regulation FD has scope limits and exceptions, including foreign private issuers. We describe it as background to the US disclosure environment, without attributing that rule to this client. Simulation supplies a way to examine draft communications; it supplies no legal conclusion about an issuer’s obligations.

The hyperscaler earnings-call engagement took 48 hours from start to finish, according to the Artificial Societies earnings-call case study. The client received a pre-call briefing and access to our platform to continue testing. That duration belongs to the engagement described here. For planning your own work, we would agree the audience and the materials to test before setting a delivery schedule.

The standing societies also gave the client an audience it could return to. The case study says they were ready for the next quarter, while platform access let the team rerun scenarios as the current numbers firmed up. For a recurring results process, we would include that continuing use in the scope from the outset, alongside the immediate briefing. The decision is whether you need a rehearsal for the current script or an audience for successive rounds of preparation.

An equity story simulation establishes how the modelled audiences respond to the material you test. In the hyperscaler engagement, the useful evidence concerned likely analyst questions, script preference and press framing. You can inspect those outputs by audience and individual persona, then decide which concerns the prepared remarks need to address. We would preserve an uncomfortable question in the briefing even when the preferred script receives a favourable overall response.

The audience model depends on observations of real people. The hyperscaler societies used public data and social activity to match personas to individuals; where diverse observations do not exist, there is nothing to ground a persona in. A persuasive simulated answer cannot fill that gap. Equally, a script preference cannot establish a share-price outcome: the decision we support is what to say and how to prepare for the questions it invites.