In our previous earnings call work, simulated sell-side analysts captured aspects of the questions real analysts later asked. We ended that piece with an open question: could simulations capture what happens next, especially how information moves between audiences?
Results day doesn’t end with the earnings call itself. Analysts publish their reports, journalists write headlines emphasising certain developments, and the story that the wider audience ends up absorbing unfolds in the hours after the call.
Five hours before Nike reported on 1 October 2026, we simulated 27 “universes”, or possible sets of results. Then we asked 23 simulated analysts and 499 simulated journalists how they’d react in each universe.
We preregistered our predictions and evaluation plan before the earnings call. In the universe that ended up matching reality, we compared our predicted headlines to the actual headlines that came out over the next few days.
Why Second-Order Effects
Second-order effects are the consequences of people's responses to an initial event or decision. A company raises its prices, customers change what they buy, and competitors may respond with offers of their own. A product launch prompts reviews, which can influence later customers' expectations and choices. In each case, the first response changes the conditions in which others make decisions.
These effects are important because they can amplify, weaken, or redirect the impact of the original action. Understanding this process can help organisations anticipate consequences they would miss by looking only at the first response. But they are difficult to study in real life: people encounter different information, interpret it through their own priorities, and react to one another over time. We can only ever observe one version of events. We can’t see what would have happened if a different decision had been made, or how people would have responded in turn. Simulations let us explore these “what ifs” under controlled conditions, while comparisons with real events help us assess their accuracy.
As we discussed in our previous article, earnings calls provide a public record to learn from and a real event to test against. They also let us study how analysts’ assessments feed into wider coverage. For investor relations teams, the questions have practical implications. Which number will dominate the headlines? How will the press frame it? Would stronger margins or a better outlook change their story?
How We Set It Up
Nike provided a good setting for this type of test. Its fiscal Q1 2027 results (covering June to August 2026) were highly anticipated: Nike was in the middle of a turnaround, and the results would provide an update on revenue, gross margin, and outlook.
To test for second-order effects, we first narrowed the first-order possibility space to those three factors. Each result could fall in line with, above, or below consensus, giving 27 (3 × 3 × 3) combinations. Each combination formed a scenario briefing representing a different possible reality, or “universe”. For this test, everything else in the briefing, including China, Nike Direct, and inventory, was held at expected levels.
In each universe, our 23 simulated analysts wrote their reactions, which 499 simulated journalists then read before producing headlines. Each simulated analyst was modelled on a real analyst who had previously covered Nike earnings calls. Before receiving any scenario briefing, each analyst wrote their top question for management. They then responded to each scenario with a question, a research note, and a rating decision.
Each simulated journalist was modelled on a real reporter at Bloomberg, Reuters, The Wall Street Journal, the FT, CNBC, or another outlet. After reading the simulated analysts’ reactions, each journalist chose an angle, a focus, and a headline for their story.
A tree of 27 universes. From Nike Q1 FY27 results, 1 October 2026, revenue branches below, in line, or above; each branch splits by gross margin, below, in line, or above; and each of those by guidance, cut, held, or raised, into universes U01 to U27, each with 23 analysts. The path that came true runs through revenue in line, gross margin in line, and guidance cut, to U13. Each universe also holds 499 simulated financial journalists. 14,117 simulated interviews in total, all run before the release.
The simulations produced 621 analyst reactions and 13,473 headlines before Nike’s results were public.
What Actually Happened
Under our classification, Nike's results corresponded to Universe-13: revenue in line, gross margin in line, and outlook cut.
| Measure | Actual | Our scenario classification |
|---|---|---|
| Revenue | $11.21bn, down 4% | In line (consensus about $11.3bn) |
| Gross margin | 42.8%, up 0.6 points | In line with guidance |
| Outlook | Full-year revenue now down high single digits | Cut |
The call also included significant news beyond the three factors we were testing. Nike announced a new cost plan with further job cuts, and Greater China revenue fell 22%. Our scenario briefings had held these other developments at expected levels across all 27 universes, so they did not introduce additional variation into the simulated analysts’ and journalists’ responses.
Finding 1: Simulated Headlines Captured the Negative Tone and Focus on Outlook
In Universe-13, 469 of our 499 simulated journalists (94%) wrote a negative headline, and none wrote a positive one. We collected 11 real headlines published in the following days by Bloomberg, CNBC, Forbes, and Reuters. None of them were positive.
The outlook was the main theme. Most simulated headlines focused on Nike’s delayed recovery, an angle that also appeared in the real coverage:
| Simulated, before the call | Real, 1 October |
|---|---|
| “Nike's Long Road Gets Longer as Turnaround Timeline Slips Into Second Half” (Bloomberg journalist) | “Nike Warns Sales Slump Will Worsen This Fiscal Year” (Bloomberg) |
| “Nike Cuts Back-Half Outlook as Brand Cleanup Drags Down Sales Volume” (Reuters journalist) | “Nike plans more job cuts, geographic changes after forecasting sharp revenue drop” (Reuters) |
Finding 2: Margin and Outlook Had the Biggest Effect on Simulated Coverage
Our simulated journalists responded more strongly to changes in gross margin and outlook than to changes in revenue. Moving from a worse-than-expected result to a better-than-expected one reduced negative coverage by 46 percentage points for margin and 42 for outlook, compared with 21 for revenue.
Three grids, one per outlook, of the percentage of 499 simulated journalists writing a negative story; rows are revenue below, in line, and above, columns gross margin below, in line, and above. Outlook cut: revenue below, 98%, 95%, 72%; revenue in line, 98%, 94%, 66%; revenue above, 98%, 86%, 70%. Outlook kept: revenue below, 98%, 74%, 60%; revenue in line, 95%, 42%, 24%; revenue above, 72%, 33%, 17%. Outlook raised: revenue below, 87%, 49%, 32%; revenue in line, 74%, 38%, 17%; revenue above, 68%, 24%, 12%. Outlined: the version that came true on 1 October (revenue and margin in line, outlook cut), 94%.
Comparing universes lets us explore counterfactuals: how might the coverage have changed if Nike had reported different results? By changing one factor at a time, we can examine which factors have the greatest influence on simulated coverage.
We can start with Universe-13, which matched Nike’s actual results: outlook cut, revenue and gross margin in line, and 94% of simulated headlines negative. In Universe-22, only revenue changed, exceeding expectations and reducing negative coverage to 86%. In Universe-16, only gross margin changed from Universe-13, exceeding expectations and bringing negative coverage down to 66%. In this comparison, stronger margins did considerably more to soften simulated coverage than stronger revenue.
The size and wording of the surprises may explain some of this difference. Our revenue beat was about 3% above consensus. Our margin beat, on the other hand, was more than 150 basis points, well ahead of guidance, and the briefing described it in stronger terms. The larger response to margin could therefore partly reflect the scale of the improvement and how we presented it. A next step would be to make the surprises more comparable in scale and consistently described to test whether margin still has a larger effect.
Finding 3: Positive Results Drew Limited Enthusiasm
Even Nike’s strongest simulated results generated a cautious response. In Universe-27, where revenue, gross margin, and outlook all exceeded expectations, only 18% of headlines were positive. Most (70%) were neutral, with simulated journalists acknowledging the improvements while continuing to focus on risks.
Poor results produced a much more consistently negative reaction. In Universe-1, where all three measures fell below expectations, 98% of simulated headlines were negative.
This caution extended across universes. Fewer than 2% of all 13,473 simulated headlines were positive. Even Universe-14, where revenue and margin met expectations and the outlook was maintained, drew 42% negative coverage. Within the simulation, concerns about Nike’s recovery persisted even when results exceeded expectations.
Finding 4: Analyst Ratings Largely Held Steady, Even When Coverage Was Negative
While the tone of simulated press coverage varied widely across the 27 universes (12% to 98% negative headlines), simulated analysts rarely changed their ratings, making just 24 downgrades and no upgrades across 621 decisions.
Almost all the downgrades (23 of 24) occurred in universes where the outlook was cut. 11 of the 24 downgrades occurred in the worst-case universe (Universe-1), where all three measures fell below expectations.
For an investor relations team, this shows why ratings alone give an incomplete picture of the response to results. In Universe-13, no simulated analyst changed their rating, yet 94% of simulated headlines were negative. In reality, several firms, including UBS, Citi, and Morgan Stanley, maintained their ratings while cutting price targets, and press coverage focused on Nike’s worsening outlook, job cuts, and falling share price. Preparing for that coverage requires looking beyond the recommendation to the concerns expressed in analysts’ assessments.
The Misses
Giving simulated analysts the scenario matching Nike’s actual results did not improve their questions. Across the seven analysts who spoke on the real call, we found no statistically significant improvement over questions generated before the scenario briefing.
Missing information may help explain why. Analysts on the actual earnings call asked about the new cost plan, job cuts, and China developments that were absent from our scenarios. Due to the way we designed the experimental structure and the scenario briefings, the simulated questions focused more on the revenue, margin, and outlook figures we varied.
Our classification of the results also mattered. Revenue was about 1% below consensus, which we treated as in line. Several outlets called it a miss and gave the shortfall more attention than our simulated journalists did.
This builds on a finding from our earlier earnings call work: the briefing strongly influences what simulated audiences focus on. In that study, explicitly naming a controversy concentrated analysts’ questions around it. The Nike study highlights two further choices that deserve testing: which developments the briefing includes, and how it labels them. Leaving out restructuring news limited the questions our analysts could ask about it; describing revenue as “in line” may have reduced the attention paid to a shortfall that real journalists treated as a miss. Both studies show why briefing design needs to be evaluated alongside the personas themselves.
What This Means for Results Day
For investor relations and communications teams, comparing universes could help with three parts of preparation, by providing specific questions and headlines to prepare against:
Decide what needs explaining. Explore which results attract the most attention and which concerns persist even when performance improves.
Rehearse alternative outcomes. Change one result at a time to see how simulated questions and coverage respond.
Test the full announcement. Include developments such as restructuring and regional performance that could shape the reaction.
A Closer Look at the Research
Design
We simulated 27 scenarios for Nike's first quarter of fiscal 2027, varying revenue, gross margin, and outlook above, in line with, or below expectations. Each scenario had a short briefing. There were 23 simulated analysts covering Nike and 499 simulated journalists modelled on reporters at 10 outlets. Journalists received the simulated analysts' reactions before producing their coverage. The study ran on 1 October 2026, about five hours before the call.
Measures
The preregistered measures covered question-topic overlap, similarity of question wording, a blind judge's ability to match questions to analysts, research-note title similarity, and headline similarity for scenario 13 compared with the other 26 scenarios. Text embeddings were used to compare meaning. Analysis of tone and headline themes was exploratory.
Results against the preregistration
Headline similarity met the statistical threshold, with a negligible effect: 0.753 versus 0.750, p = 0.002 after correction for multiple tests. The tests of question topics, wording, and analyst matching did not support an improvement. Research-note titles could not be evaluated because no real titles were publicly available.
Limits
Only seven of the 23 analysts asked a question on the real call, leaving a small sample for the question tests. The 11 real headlines came from four outlets and were gathered from search results. Simulated responses depended on briefings that held China, Nike Direct, and inventory at expected levels. The scenarios also differed in the size and wording of their surprises. The study compared the resulting coverage with reality, but did not separately establish how much the analysts' reactions influenced the journalists.