Search earnings calls

Search English earnings calls to create a dataset of measures.

Run a search

Start with a draft query and the date range you want to study. For a complete walkthrough, use Create your first dataset.

1. Open New search.
2. Enter words or phrases in Query, for example supply chain OR logistics. Set Start date and End date.
3. Check the selected measures under Metrics, and optionally enter a Search name. Leave filters and scoring at their defaults unless you need the settings below.
4. Select Search corporate text. This creates a dataset and counts as one search against your plan limits.

Need help choosing terms? Use the Keyword Tool. For operators and phrases, see query syntax; wildcards depend on the selected engine.

Review and export

When the search completes, start on the dataset page:

• Inspect the text. In Evidence, review matched sentences and switch to Speaker context when you need the surrounding text. Start with Exposure; inspect matches explains what to look for.
• Check what drives the results. Review the time series and company or sector breakdowns. Under What this measure matches, choose Show terms to inspect the query terms. Investigate unexpected concentrations before relying on the measure.
• Export when ready. Open Downloads for the data files and query/settings JSON. Follow Export data to choose the files and save your research record.

The main CSV has one row per earnings call, identified by earningscallID, including calls with zero matches. Its columns depend on the selected measures.

Adjust settings when needed

Use these panels when your research question needs a different measure or a restricted sample.

Individual companies cannot be selected before a search. Filter the export or use the Snippet Tool's company filters afterward.

Scoring options

With the new engine, Options contains two controls for risk and sentiment scoring. They change scoring, not which calls match your query:

• Text used for scoring: “Matched sentence” or “Words around each match”. Selecting “Words around each match” lets you set Words in the window from 5 to 100 words total, including the matched term or phrase (default 22, with presets 11, 22, and 44; a window of 22 covers ±10 words around a two-word phrase).
• How relevant words are counted: With standard search, choose “Whether any appear” or “How many appear”. Updated search uses “Whether any appear”; “How many appear” is unavailable.

For the sentence-count examples, keep Matched sentence and Whether any appear, the defaults. See how scoring changes the counts before choosing another mode. Save the engine and settings with your research record.

Engine-dependent features

Where Use new search engine is shown, turning it off selects the previous engine. Availability depends on your account. With updated search, marked Beta next to the page title, new searches always use the new engine and there is no switch.

• Wildcards (* and ?): With standard search, these require the previous engine; otherwise, list the word or phrase variants explicitly. Updated search supports words ending in *, but not ?.
• Counts for each term also require the previous engine. Under Options, select Download counts for each search term to add the Keyword breakdown file.
• Combining or comparing datasets requires both to use the same engine. See dataset compatibility.

Changing engines can clear incompatible options. Check the settings again before running the search.

If the query is too broad

If the platform rejects the query as matching too much of the corpus, use more specific phrases or exclusions. See query troubleshooting, or contact support if the broad query is essential to your project.