Company Earnings Results 2019–2026

Key findings

1. 225,779 earnings results for 33,900 companies, reported in the news between January 2019 and June 2026. 60,163 are full-year results, 142,897 quarterly and 22,719 half-year.
2. 35,956 results say how they compared with analyst expectations: 62% beat, 27% missed and 11% were in line.
3. Coverage is led by United States (79,283 results), India (45,650) and Canada (20,556), with 152 countries in all.
4. Median reported full-year revenue growth (results with US$100M+ revenue): fiscal 2019 +7.0%, fiscal 2020 +3.7%, fiscal 2021 +16.3%, fiscal 2022 +17.0%, fiscal 2023 +10.0%, fiscal 2024 +7.0%, fiscal 2025 +7.8%.
5. 28% of quarterly results with a net-income figure were losses, from 20% (2026) to 37% (2020) by year.
6. 83% of results carry a revenue or profit figure that was found again in the reporting sentence; 3.7% had a figure blanked because it failed the check.

Results by report year and period type

Full data, by the year the result was first reported.

Beat, miss or in line by report year

Beat and miss rates by sector (top 30 sectors by results with a verdict)

Sectors with at least 100 results that state a verdict.

Largest companies by latest full-year revenue (top 50, fiscal 2024 or later)

Each company's most recent full-year result whose revenue figure was confirmed in the reporting sentence. Figures are as reported; some articles quote group or gross revenue.

Fastest full-year revenue growth (top 50, prior-year revenue ≥ US$1B, fiscal 2023 or later)

Computed from two consecutive full-year results of the same company in the same currency, kept only when it agrees (within 3 points) with the growth the later article reported. Jumps can include acquisitions.

Results by country (top 30)

Beat rate shown where at least 30 results state a verdict.

Full-year medians by sector (top 40 sectors)

Full-year results with revenue; sectors with at least 50.

Column dictionary

23 columns in the sample CSV.

Method

1. Extraction. Radar's pipeline reads each earnings article and extracts structured fields, including the source sentence.
2. Scope. Results linked to a company website and first reported between 2019-01-01 and 2026-06-30. Sentences that only preview or forecast a result (for example "is expected to report", "analysts expect") with no reported figure are dropped, as are records without any revenue, net-income or EPS figure.
3. Periods. Period labels are normalised to FY, H1, H2 and Q1–Q4. When the fiscal year is not stated it is read from the period text, else from the report date (a full-year, Q4 or H2 result reported in January–June belongs to the previous year).
4. Deduplication. All reports of the same company, fiscal year and period are one row. The row keeps the record with the widest coverage, then the most figures, then the earliest report. report_date is the earliest report; news_mentions counts the reporting articles.
5. Amount check. Revenue and net income are compared with every money figure in the source sentence (tolerance 0.6–1.7×). A figure is blanked when the sentence talks about that metric and gives figures with a scale word ("million", "crore") but none matches, or matches only after a power-of-ten shift (crore/lakh or million/billion slips); when it exceeds a ceiling (revenue US$700B a year, US$400B a half-year, US$200B a quarter); when net income is more than five times revenue or less than a millionth of it; or when a very large revenue (over US$20B a year, US$10B a half-year, US$5B a quarter) is not corroborated by another result of the same company within 0.4–2.5×. EPS above 100,000 and growth above 1,000% are blanked.
6. Outlooks and market figures. Sentences about outlook, guidance, consensus estimates, annualised run-rates, market size or combined industry revenue are dropped even when they contain a reported figure.
7. Currency. A "$" figure for a company in Canada, Australia, New Zealand, Singapore, Hong Kong, Chile, Mexico, Colombia, Argentina or Taiwan is read in the local currency unless the sentence says US dollars; rupee figures of Pakistani, Sri Lankan and Nepali companies use the local rupee; crore/lakh figures of Indian companies are INR; a "USD" label with no dollar in the sentence for a non-US company is cleared. US$ columns use one fixed rate per currency (approximate 2024 averages); currencies without a stable rate get no US$ figure.
8. Normalisation. Country names are unified (USA / US / United States → United States). Company names and sectors come from Fliar's company index where the website is known, otherwise from the article.

9. News columns. The excerpt is the earliest reporting sentence that names no data provider. Source and URL come from that article when it can be retrieved.

10. Sample. Each companie key is ranked by sha1(key). Keys are taken in rank order until the file holds max(25%, 5,000) rows (capped at 100,000).

11. Profile links. Links are built from Fliar's company index and checked against it. A row whose company has no public profile has an empty link.

Limitations

• Derived from news: results that were never reported are missing, and coverage is stronger for large, listed and English-reported companies.
• Figures are as reported. Some articles quote adjusted or segment figures, or profit before tax where net income was expected.
• Company-to-website links are automatic; a small share of results may be attached to the wrong company website.
• US$ conversions use one fixed rate per currency, so they are for comparison, not exact.
• A few records still mix currencies inside one result (for example net income in local currency labelled US$).
• beat_or_miss is only present when the article said so.
• news_source and news_url are blank when the article could not be retrieved at release time (some article-store nodes were offline).
• 2026 covers January to June only.

Cite this dataset

@misc{fliarbi_company_earnings_results_2026,
  author    = {{Fliar BI}},
  title     = {{Company Earnings Results 2019–2026}},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.23432150},
  url       = {https://doi.org/10.5281/zenodo.23432150},
  note      = {Sample release. Licence CC BY-NC 4.0}
}