The setup
Earnings week is a narrative contest with a scoreboard. The numbers are fixed by the time the release is drafted; what remains undecided is which story the numbers get told inside — yours, or the one a short-seller published a month ago and has been patiently seeding since.
This scenario simulates the contest. The setup gives neither side a knockout: the beat supports management, the retention dip supports the short thesis, and every audience must decide which fact is the signal and which is the noise. The cast covers the whole information food chain — analysts who move institutions, media that moves retail, retail that increasingly moves itself, and an IR function that has one prepared explanation and a choice about how much to disclose. The horizon is one trading week.
The strategic question the report answers is not “will the stock go up” — MiroFish is exploratory decision support, not a price oracle, and this library never pretends otherwise. The question is which explanation survives five days of adversarial retelling, and what disclosure, offered when, would have strengthened it. That is a communications decision, it is in management’s control, and it is exactly the kind of many-voices problem a multi-agent simulation is for.
The prompt
Copy this into MiroFish as your scenario question, swapping the specifics for your own:
A mid-cap software company reports earnings Thursday: revenue slightly beats, but net retention dips for the second straight quarter. A short-seller published a thesis last month claiming retention is the crack in the story. Simulate the five trading days after the print across: sell-side analysts, financial media, the short-seller and allies, retail investors on social platforms, and the company's IR account. Management frames the retention dip as 'one large legacy customer, now lapped.' Tell me which narrative owns each audience by Friday, where the frames collide, and whether proactively disclosing the customer's name changes the outcome.
Seed files that help
- The draft earnings release and prepared remarks — the exact sentence explaining the retention dip is what every agent will parse.
- The short thesis itself (or a faithful summary), because the simulation needs the strongest version of the opposing case.
- Transcripts or notes from the last two earnings calls, which teach the analyst agents what management has previously promised.
What to look for in the report
- Frame adoption by audience: analysts, media, and retail rarely settle on the same story — the report shows which audience the short thesis captures first.
- The collision round: the moment the 'one customer' explanation meets the 'second consecutive dip' pattern; whichever side wins that exchange tends to own the week.
- Evidence hunger: what the agents keep asking for that management hasn't offered — usually the disclosure decision the real IR team should be debating.
- Retail decoupling: whether retail sentiment detaches from the analyst consensus, which changes how much the narrative matters at all.