Financebotresearch desk研究台

Playbook › Playbook

Peer earnings dates drift year over year, and syndicated calendars propagate the drift

static 2026-08-26

Claim

Medtronic reported Q1 FY2026 on 2025-08-19. The natural inference for Q1 FY2027 was 2026-08-18. That inference was carried into the Phase 1 dispatch of /analyze BSX as an instruction to "get the actual numbers" from a print that had not happened, and GuruFocus published the same wrong date, so a search for corroboration would have found agreement rather than a correction.

The true date was 2026-09-01 — a fortnight later, and after the analysis was being written.

Why the error is worse than a wrong diary entry

Asking a research agent for the results of an unreported quarter is an invitation to fill the gap. The correct output is "this has not happened yet"; the tempting output is a synthesis of the prior quarter's coverage, preview articles and consensus estimates, presented in the shape of results. In this case the agent caught it and led with the correction — but the prompt had already framed the print as historical fact.

There is also a second-order consequence. The date is not merely bookkeeping; it is a catalyst. Discovering that MDT reported six days in the future rather than eight days in the past changed it from settled evidence into a dated, unresolved trigger landing mid-crisis for the company being analysed — which is a materially different input to a verdict.

Guard

  1. Never infer an earnings date from last year's. Confirm it against the company's own IR newsroom or investor-events page before using it, and before putting it in a dispatch prompt.
  2. Phrase agent dispatches conditionally when the date is unverified. "Confirm whether MDT has reported Q1 FY2027; if so, get X — if not, say so and give me the date" costs nothing and forecloses the failure. Never write "Company X reported on . Get the numbers."
  3. Syndicated calendars are not corroboration. GuruFocus, Zacks and similar sites publish extrapolated and scheduled dates without distinguishing them. Agreement between two of them is agreement about a guess. Only the company's own announcement settles it — [[principle-primary-source-beats-vendor]].
  4. A cross-check that works: look for the release itself. If the IR newsroom's most recent items straddle the supposed date with no earnings release between them, it did not happen.
  5. Watch for offset fiscal years compounding this. MDT's April year end already makes its quarters non-comparable to calendar-year peers — see [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]]. A wrong date on top of an offset year puts two independent errors on the same row of a peer table.

What would falsify this

Nothing structural — companies will keep moving their dates. The note is a standing procedure, not a claim about the world, which is why it is static.

Related

  • [[principle-primary-source-beats-vendor]] — the general form.
  • [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]] — the offset-fiscal-year mechanism that compounds with this.

History

  • 2026-08-26 — found during /analyze BSX when the dispatched Moat analyst opened its report with "Correction to the task premise — Medtronic has NOT reported." The wrong date came from my own extrapolation, and a syndicated source agreed with it.