An earnings report is a company's periodic disclosure of revenue, earnings per share and related financials for a fiscal quarter or year, released first as a press release and filed with the SEC shortly afterwards on Form 10-Q or Form 10-K. Everything downstream of that moment, the consensus surprise, the index aggregate, the sector weight that shifts on the next rebalance, is derived from it. Most published coverage of earnings stops at the calendar entry. This article picks up where the calendar stops.
What is an earnings report?
An earnings report is a company's periodic disclosure of revenue, earnings per share and related financials for a fiscal quarter or year, issued as a press release and filed with the SEC on Form 10-Q for a quarter or Form 10-K for a fiscal year. The press release comes first and is unaudited. The filing comes later and is the authoritative version.
The gap between those two documents is where most data problems begin. A press release typically leads with adjusted or non-GAAP earnings per share, excluding items the company considers non-recurring: stock-based compensation, acquisition amortisation, restructuring charges, impairments. The 10-Q presents GAAP figures with the non-GAAP reconciliation in a separate table, and the two numbers can differ by a wide margin for companies with heavy equity compensation.
The number the market trades at 4:01 PM and the number that eventually lands in an index earnings calculation are therefore not always the same number. Consensus estimates are usually compiled on an adjusted basis, so the "beat" or "miss" headline measures adjusted EPS against an adjusted estimate. Index-level earnings aggregates may use GAAP, operating earnings, or a vendor-specific normalisation, and the choice changes the result materially.
A fourth-quarter release adds another wrinkle: it often contains full-year figures that are preliminary, with the audited versions arriving weeks later in the 10-K. Anyone keying a process to the press release is keying it to a provisional number.
What the earnings calendars actually give you, and what they leave out
Earnings calendars give you scheduling: ticker, company name, date, announce time, consensus EPS estimate, and after the fact the reported figure and surprise percentage. That is genuinely useful, and the free ones do it well. What no calendar gives you is the link from the reported number to anything at portfolio or index level.
Reading Yahoo Finance's earnings calendar on 8 October 2026, the page lists each row with symbol, company, earnings date including announce time, EPS estimate, reported EPS and surprise percentage; the Alphabet row showed an October 28, 2026 date at 4 PM EDT with a 3.03 EPS estimate. That is the standard shape of the format across free providers.
Wall Street Horizon, according to its own site as read on 8 October 2026, publishes corporate event data covering earnings dates, dividends, stock splits, investor conferences and mergers and acquisitions for institutional traders. The positioning is event timing accuracy rather than financial statement depth, which is a different and legitimate product.
What none of these publish, by design:
- Revision history on the estimate. A 3.03 consensus that was 3.28 six weeks ago tells a different story than a 3.03 that has not moved. The calendar shows the current value only.
- Fiscal-period alignment. A calendar row labels the quarter by the company's own fiscal convention, so two rows on the same page can refer to entirely different three-month windows.
- GAAP versus adjusted basis. The estimate basis is rarely stated on the row.
- Any index or fund linkage. No calendar tells you what weight that company carries in the indexes and ETFs that hold it on the day it reports.
| Earnings calendar | Holdings-level data | |
|---|---|---|
| Primary object | Scheduled event | Position in an index or fund |
| Update trigger | Company announcement or estimate | Daily pricing and rebalance |
| Identifier depth | Ticker, company name | Ticker, FIGI, sector, weight |
| History | Past reported EPS and surprise | Point-in-time holdings and weights |
| Answers | When does this company report | What does this report move, and by how much |
The two are complements, not substitutes. A workflow that tracks earnings without tracking weights is measuring a signal with no transmission mechanism attached.
When do the big names report, and why the date keeps moving
Most large caps report within a predictable window after their fiscal quarter closes, but the quarter itself closes on the company's calendar, not yours. Announce dates published more than two or three weeks ahead are usually estimated from prior-year patterns rather than confirmed by the company, and they move.
Two conventions dominate US large-cap scheduling. Technology and most post-close reporters release after the 4 PM ET equity close, usually at 4:05 or 4:30 PM, with the call thirty to sixty minutes later. Banks, industrials and consumer staples more often release pre-open, between 6:30 and 8:00 AM ET. The distinction matters for anyone timing a data pull, because a pre-open reporter's number is already in the market before the session opens and a post-close reporter's number is not reflected in that day's closing holdings file.
The practical rule: a date sourced from a calendar two weeks out is a scheduling input for resource planning. It is not a data point you build a dated process around until the company's investor relations page confirms it.
Nvidia earnings report timing and the January fiscal year
Nvidia operates on a fiscal year ending in late January, which means its fiscal quarters run roughly a month ahead of the calendar. Nvidia's fiscal third quarter ends in late October and is reported in November, so the "Q3" in an Nvidia headline is not the calendar third quarter, and fiscal year labels run one year ahead of the calendar year for most of the year.
This causes concrete errors in aggregation. Line Nvidia's fiscal Q3 up against a December-year-end peer's calendar Q3 and you are comparing a period ending in late October with one ending 30 September. For a semiconductor cycle that turns in weeks, a month of offset is not a rounding issue.
AMD earnings report and the semiconductor reporting cluster
AMD runs on a calendar fiscal year ending in late December, reporting its quarters a few weeks after each calendar quarter closes. That puts AMD, Intel and Texas Instruments in roughly the same reporting window, several weeks ahead of Nvidia on each cycle.
The clustering has a measurable effect on sector-level data. When a large fraction of a sector reports inside a ten-day window, the sector's trailing earnings aggregate moves in a step rather than smoothly, and any index P/E computed during that window is mid-transition. If you are producing sector comparisons, note which constituents have rolled and which have not. The classification layer behind that question is covered in our ETF List by Sector: 2026 Classification Guide.
Apple, Tesla and Micron, three fiscal calendars, three alignment problems
Apple's fiscal year ends in late September, so its fiscal Q1 is the December quarter and its fiscal Q4 report lands in late October or early November covering the September quarter. Tesla runs on a standard December calendar year. Micron's fiscal year ends around the start of September, making it one of the earliest reporters in each cycle and a frequently used read-through on memory pricing.
Put those three in one spreadsheet keyed on "Q3" and you have three different periods: Micron's fiscal Q3 ends in late May, Tesla's Q3 ends 30 September, and Apple's fiscal Q3 ends in late June. Any aggregate built on the label rather than the period end date is wrong, and it will be wrong quietly. The fix is to key every earnings record on the fiscal period end date and carry the label only as metadata.
Do stocks go up or down after an earnings report?
Direction after an earnings report is not predictable from the surprise alone, because the price move prices the gap between the result and investor positioning, not the result against published consensus. A company can beat consensus and fall, which happens routinely when buy-side expectations ran above the published number or when guidance disappoints.
What is more tractable is dispersion. Options AI's weekly earnings page, as read on 8 October 2026, pairs each reporting name with a realised one-day move figure, which is the options market's own estimate of the magnitude of the move rather than its sign. Treating an implied move as a dispersion estimate and not a directional forecast is the correct reading.
The institutional consequence is not stock selection. It is that dispersion around a reporting cluster raises index-level realised volatility for a few sessions, widens the spread between cap-weighted and equal-weighted returns, and can push tracking behaviour in concentrated sector ETFs away from its usual range. If a single holding carries double-digit weight in a sector fund, that fund's one-day behaviour on its report date is substantially a single-name event.
That weight is the variable you need, and it is the one earnings data does not contain.
From EPS to index level: how one report reaches an index earnings number
An index earnings number is built by normalising each company's reported EPS, multiplying it into the index through the company's share count or weight, and rolling the trailing four-quarter sum forward as each new report arrives. There are at least four decision points in that chain, and two vendors making different choices will publish different numbers for the same index on the same day.
The chain runs in this order:
- The company reports. Press release adjusted EPS first, GAAP in the filing days or weeks later.
- The vendor normalises. Choose GAAP, operating, or an adjusted basis; decide how to treat discontinued operations, one-time tax items and impairments.
- The figure is scaled into the index. For a divisor-based index this uses share counts and the index divisor; for a weight-based calculation it uses each holding's weight on the computation date.
- The trailing twelve months roll. The newest quarter replaces the one from four quarters ago, which can move the aggregate sharply if the dropped quarter contained a large charge.
- Forward estimates replace actuals in any forward P/E or forward earnings yield, which introduces the estimate vendor's own consensus methodology on top of everything above.
This is why an index P/E quoted in two places on the same morning can differ by more than a point without either being wrong. One may use GAAP trailing earnings, another operating earnings; one may use as-reported share counts, another float-adjusted weights; one may include negative-earnings constituents in the aggregate, another may handle them differently. We work through the aggregation arithmetic in Earnings EPS: Filings to Index Level and the specific divisor and basis questions in S&P 500 PE Ratio: How It's Computed.
The lesson for anyone publishing a valuation figure: state the basis. An index P/E without a stated earnings basis and date is not a reproducible number.
Why holdings data is the missing half of earnings analysis
The weight attached to each reported EPS comes from the holdings file, not from the earnings calendar. Index weights move every day with price and change discretely on rebalance and reconstitution dates, so an earnings aggregate computed with last quarter's weights is wrong by construction before you start.
The size of the error scales with concentration. In a cap-weighted US large-cap index where the top ten names carry a substantial share of the total, a quarter of price drift between rebalances meaningfully changes how much each reported EPS contributes to the aggregate. For sector funds the effect is sharper still.
AmericanETP's position in this chain is the holdings side. Holdings files for 3878 US and global indexes and US-traded ETFs are updated twice daily, with the primary update at 6pm EST and a secondary update around noon EST, delivered as CSV with FTP access available. Bloomberg FIGIs are included in the holdings files, which matters for joining earnings records keyed on one identifier to position records keyed on another; the Bloomberg Extensions page covers that identifier layer.
For historical work the archive is the operative feature. Holdings datafiles go back to July 2009, which means you can reconstruct what an index or ETF actually held, and at what weight, on any past earnings date rather than applying today's composition to a 2014 quarter. Sample files are available on the Downloads page, and membership changes between dates are tracked in the Constituent Change Report.
Restatements, revisions and the data problems that show up weeks later
Earnings calendars record the reported number once and never revisit it. Financial statements, however, change after publication, and several categories of change can alter a historical quarter months after the fact.
The main ones:
- Press release versus filing. The adjusted EPS in the 8-K press release and the GAAP EPS in the 10-Q are different figures covering the same quarter. Which one your database stored determines what your backtest tested.
- Restatements. When a company restates a prior period, the historical quarter changes. A time series refreshed from today's vendor snapshot silently inherits the corrected figure, which the market never saw at the time.
- Segment reclassification. Reorganising reporting segments restates prior-period segment revenue and operating income without changing the consolidated total, which breaks any analysis keyed on segment lines.
- Share count changes. A large buyback raises EPS without raising earnings. In an index aggregate built on share counts, the buyback changes the index-level figure through the denominator rather than the numerator, and the two effects need to be separated before anyone calls it earnings growth.
The practical consequence is that any backtest keyed on as-reported numbers requires point-in-time data on both sides: point-in-time earnings as they stood on the date, and point-in-time holdings as the index was actually composed on that date. Combining a restated earnings series with today's index membership produces a result that could not have been obtained by anyone in real time, and it almost always looks better than reality.
Where earnings data and holdings data come from: vendor landscape in 2026
No single vendor covers the full chain from announce date through filing to index weight, so most institutional setups assemble two or three sources. The sensible way to compare them is by what each is built for, not by feature counts that go stale.
Corporate event specialists. Wall Street Horizon, per its own site as read on 8 October 2026, publishes corporate event data covering earnings, dividends, stock splits, investor conferences and mergers and acquisitions for institutional traders. If your problem is knowing accurately and early when an event will occur, that is the category built for it.
Free broad-audience calendars. Yahoo Finance and CNBC serve a wide retail and professional audience with scheduled dates, estimates and surprise percentages at no cost, and they do that job well. They are not built to be a system of record, and treating a free calendar as one is a sourcing decision rather than a vendor failure.
Platform vendors. FactSet, LSEG (Refinitiv), Morningstar and S&P Global Market Intelligence position themselves as broad multi-asset data and analytics platforms rather than single-dataset providers. The structural difference that does not go stale is scope: a platform is sold on the premise of covering many datasets and workflows in one environment, where a holdings-only provider solves one layer and expects you to join it to the rest. That breadth is a genuine advantage for a team that wants fewer vendors and one place to work, and it is the main reason those platforms command what they do. We have not verified the current scope of any specific earnings, estimates or filings module at those vendors against their own documentation for this article, so confirm capabilities directly with each one rather than from secondary write-ups. We cover one of them in detail in our Refinitiv Review 2026, the cost structure in Refinitiv Pricing: What It Costs in 2026, and the scope difference in AmericanETP vs Refinitiv (LSEG) 2026.
AmericanETP. Holdings and identifiers as CSV, twice-daily updates, FTP available, archives back to July 2009, at $1500 per year or $150 per month for an individual and $2500 per year or $250 per month firm-wide. To be explicit about the boundary: AmericanETP does not publish an earnings calendar and does not supply consensus estimates. The product solves the weight and composition half of the problem, and is priced to sit alongside whatever earnings source you already use rather than to replace it.
The company has been doing this since the 1990s under its original name, MasterDATA, which was changed to AmericanETP to avoid confusion with IBM's use of the term. That predates most current ETF data competitors, which is relevant mainly because it is why the archive goes back as far as it does.
Building an earnings workflow that survives audit
A defensible earnings workflow uses four distinct sources and keeps them distinct: a calendar for dates, SEC filings for authoritative figures, holdings data for weights, and an archive for point-in-time reconstruction. The failure mode is letting one source answer a question it was not built to answer.
Assembled in order, it looks like this. Use the calendar to schedule, and confirm dates against company investor relations before anything dated depends on them. Pull the press release figure on the night for real-time use, then overwrite it with the filing figure and keep both, flagged by basis and source document. Join to holdings on a stable identifier rather than ticker, since tickers change and are reused, which is the case for carrying FIGIs through the join.
File-level hygiene is unglamorous and it is where audits break. Every stored record needs the fiscal period end date, not just the quarter label; the basis, GAAP or adjusted; the source document type; and the retrieval timestamp. Holdings records need the as-of date of the file they came from, not the date you loaded them.
Before publishing any composite or fund-level earnings figure, reconcile it against the underlying holdings: sum the weights, confirm they total what they should, check that every holding with a reported quarter has been rolled and every one that has not reported is flagged. If a sector earnings yield moves more than a point in a day, the usual cause is a weight file that updated while the earnings series did not. If you want to check a specific index or fund against sample files before committing to a process, contact AmericanETP.
FAQ
What time will Google report earnings?
Alphabet, Google's parent company, has historically reported after the US market close. Yahoo Finance's earnings calendar, as read on 8 October 2026, showed an Alphabet row with an October 28, 2026 date at 4 PM EDT and a 3.03 EPS estimate. Calendar dates more than two weeks out are often estimated rather than confirmed, so verify against Alphabet's own investor relations page before relying on it.
What is an earnings call?
An earnings call is the conference call a company holds after releasing its quarterly results, in which management presents the quarter and takes analyst questions. It typically begins thirty to sixty minutes after a post-close press release. For companies that give forward guidance verbally rather than in the release, the call often moves the stock more than the printed numbers do.
Which companies have reported good earnings today?
A "good" result is conventionally one where reported EPS exceeds the consensus estimate, which free calendars display as a surprise percentage after the fact. That metric has two limits worth knowing: the comparison is usually adjusted EPS against an adjusted estimate, not GAAP, and a beat frequently coincides with a price decline when guidance or buy-side expectations ran ahead of published consensus.
Do stocks go up or down after an earnings report?
Direction cannot be predicted from the earnings surprise alone, because the move prices the result against investor positioning rather than against published consensus. Options markets price an expected magnitude of movement without a direction; Options AI's weekly earnings page, as read on 8 October 2026, pairs each reporting name with a realised one-day move figure. The reliable statement is that dispersion rises around earnings dates, not that prices rise.
When does Nvidia report earnings?
Nvidia operates on a fiscal year ending in late January, so its reporting calendar runs roughly a month behind the calendar quarters: fiscal Q3 ends in late October and is reported in November. Nvidia reports after the close, with the call following shortly afterwards. Because the fiscal labels run ahead of the calendar year, always key Nvidia records on the fiscal period end date rather than the quarter label.
What is the difference between GAAP and adjusted EPS in an earnings report?
GAAP EPS follows US Generally Accepted Accounting Principles and appears in the Form 10-Q or 10-K filing. Adjusted or non-GAAP EPS is the company's own figure excluding items it considers non-recurring, commonly stock-based compensation, acquisition amortisation and restructuring charges, and it usually leads the press release. Consensus estimates are typically compiled on the adjusted basis, so headline beats and misses measure adjusted against adjusted while index earnings aggregates may use a different basis entirely.
How do earnings reports affect ETF holdings?
An earnings report does not directly change what an ETF holds. It changes the holding's price, which changes its weight in a cap-weighted fund the same day, and it changes the fundamentals that drive eligibility and weighting at the next scheduled rebalance. Capturing both effects requires holdings data that updates daily rather than quarterly, which is why weight files and earnings records belong in the same workflow.
AmericanETP provides daily updated holdings, sector weights, fundamentals and Bloomberg FIGIs for 3878 US and global indexes and US-traded ETFs, as CSV, with archives back to July 2009. See the data and sample files at americanetp.com.