Deep Research Agents

Generates a comprehensive summary of earnings results for a company, analyzing financial performance, key metrics, and management commentary from earnings reports and conference calls.

Lists the forward-looking commitments still in force, and how each one got there. This is the live guide, not a score.

The agent reads filings and earnings releases over a lookback window (two years by default), extracts each piece of guidance and each expected event, and threads restatements of the same target into one history. For each open commitment it narrates when it was first given, whether management later raised, narrowed, or lowered it, and where it stands now. Metric targets whose period has not ended, and events whose window has not closed. Outcomes are not scored.

Scope a run to a single earnings period to see only the guidance that print issued or updated. In that mode each thread is flagged new (the period introduced it) or updated (the period restated a commitment that already had history). That is not a score of the print. For the commitments that print put to the test, use Guidance Due at Earnings. Leave the period unset to see the full set of open commitments.

Every thread links back to the filings it was built from. For the scored multi-year record, use Guidance Track Record. For how guidance is extracted, standardized, and threaded, see Guidance Scorecard.

Scores closed guidance against what actually happened, over a long window (five years by default). This is the historical record, not the live guide.

It takes the commitments that have since closed, finds what was reported, and grades each one. Metric targets get beat, met, or missed against the final guided figure. Events get occurred, delayed, pending, or cancelled. Pending means the window is still open at the as-of date; it is listed so the table is complete, and it does not count against accuracy. A target with no comparable actual is left unscored rather than guessed. Verdicts are kept only when a reported actual backs them.

The run then rates the company on two scores from 1 to 5:

• Accuracy: how closely outcomes matched the guidance, weighted toward the targets that mattered most.
• Discipline: how consistent and reconcilable the company's guidance practice has been over time.

An overall rationale cites the threads that drove the scores. For commitments still in force, use Current Guidance. For scoring tied to one reporting date, use Guidance Due at Earnings. For the methodology behind the verdicts and scores, see Guidance Scorecard.

Scores the commitments that came due at one earnings release. Prior targets this print put to the test, not the guidance this print issued.

Select a company and an earnings period (defaults to the most recent reported). The agent finds the commitments that resolve at that release: metric guidance whose guided period matches what the company is now reporting, and expected events whose window closes on or before the release. It evaluates each as of the earnings release date, so the verdict reflects what was known then rather than today. A fourth-quarter release tests the full-year guidance it reports against.

Each tested metric gets beat, met, or missed against the final guided figure. Each event gets occurred, delayed, pending, or cancelled. A verdict is kept only when a reported actual backs it. The run also returns accuracy and discipline scores for the guidance this release resolved.

For the new guide issued at that print, run Current Guidance scoped to the period. For the multi-year record, use Guidance Track Record. For the methodology behind the verdicts and scores, see Guidance Scorecard.

The Tariff Impact Tracker estimates how the April 2025 U.S. reciprocal tariffs, and the Section 232 and IEEPA tariffs that came with them, affect a single company. Select a company and the agent reads its filings and earnings calls, supplemented by web search, to find where management has discussed tariff exposure and what it has done about it.

The output separates two impacts and reports each as a money range:

• Revenue impact: lost sales and demand destruction attributable to tariffs, both already observed and forward-looking.
• Cost impact: the net hit to costs and profitability the company actually bears after its own pricing actions, cost reductions, and supply-chain adjustments.

Behind those estimates is a table of every tariff figure the company disclosed, each with its period, the date management stated it, and its basis: gross (before mitigation), net (after it), or a refund of tariffs already paid. Guidance appears each time it was given, so you can see an outlook being raised or cut, and actuals sit apart from expectations. A refund is shown as a one-time benefit rather than an improvement in the underlying business.

Alongside the numbers, the agent narrates the qualitative picture: pricing strategy, supply-chain changes, and capital-allocation decisions management has tied to tariffs. Every figure and claim cites the filing or call it came from, and up to three verbatim quotes are attributed to their speaker and source. When a company has no meaningful tariff exposure, the agent says so rather than manufacturing an impact.

The estimates are model-derived from what the company has disclosed, not read from a tariff schedule. They reflect the net effect management reports bearing, so a company that has fully passed tariffs through to price shows a smaller cost impact than its gross exposure would suggest. Where a company gives only a gross figure with no offset, the agent shows it as gross rather than presenting it as net.

The Common Ground Update reads an earnings release against everything the company has told the market before and reports only what is new. Pick a company, and optionally an earnings period; it defaults to the most recent one. The agent then compares the release with the company's common ground, the body of facts it has built up across past filings, releases and calls.

Most of an earnings release repeats what the market already knows. The same strategy slides, the same cost program, the usual quarter-on-quarter numbers. All of that gets filtered out. What remains falls into a few groups: statements that contradict or walk back earlier ones, plans and commitments appearing for the first time, material revisions to figures or timelines, and familiar plans whose timing moved.

The output starts with a short overview of the biggest changes, then lists the new facts by topic (strategy, guidance, capital allocation and so on). Each fact is grounded in the source documents of that release. If a release contains nothing new, the agent says so instead of dressing up routine results as news.

The first run on a company has to work through its document history, so it can take a few minutes. Later runs reuse that work and finish much faster.