Start with what matters to the market

Investors can read the same earnings report very differently. The key is not the number of metrics, but which one is driving the market's view of the stock.

Earnings Analysis starts with what the market expected before the release. It then shows whether the results confirmed that view, challenged it, or left the debate unresolved.

Separate the quarter from the outlook

Current-quarter profit is backward-looking. Orders, inventory, deferred revenue, guidance, and management tone often say more about what comes next. The analysis separates durable growth from one-off gains, cost cuts, and carefully managed expectations.

• Map the disagreementPinpoint the core tension the market cared about going in — and why bulls and bears walked away from the same report with opposite conclusions.
• Assess growth qualitySeparate business expansion and margin improvement from one-time gains and accounting effects, so headline results do not obscure the underlying trend.
• Set clear checkpointsTie a hold, wait, or trim decision to metrics the next report can verify instead of relying on instinct.

Build a trackable earnings map

1. 01

   One-line verdict

   It opens with the report's main storyline: strong results but weak guidance, or numbers that look good but don't hold up on quality.

2. 02

   The numbers that matter

   The handful of metrics driving the market's view, how they compared with expectations, and why the difference matters.

3. 03

   Trend signals

   Comparing across recent quarters: is growth accelerating, slowing, stabilizing — or at an inflection point.

4. 04

   Decision scenarios

   For each situation—holding, watching, or concerned—it identifies the earnings metrics that could confirm or invalidate the thesis.

When to use it

Use it after a company reports, when the stock moves sharply after hours, or when investors disagree about the same set of results. It connects the numbers, valuation, and price reaction in one research view.

It won't hand you a price target or bury you in models. It makes the key changes clear and gives you concrete signals to test your thesis over time.

Start with original documents

For a reproducible company example, choose one reporting quarter from Microsoft Investor Relations and collect its earnings release, financial statements and outlook. Use the annual reports archive for definitions and accounting context. These are source entry points, not evidence that a particular investment conclusion is correct.

Save the exact document URL, reporting period, publication date, page or table reference, currency and units. Keep management guidance separate from reported results. If you use market consensus, record its provider and observation time; if it is unavailable, do not claim a “beat” or “miss.”

Example: turn one earnings question into a reusable workflow

This is a worked protocol, not a completed product benchmark or a recommendation to buy a security. It deliberately contains no invented financial figures or model scores.

1. Freeze the inputs. Use the same quarter, original documents, question and research cutoff in each run. Record the model or product version, run date and whether web access is enabled.
2. Ask for an evidence table. “Using only these documents, compare revenue, operating income and operating cash flow with the same period a year earlier. For each value, give its source, page or table, period, currency and units. Distinguish facts, management guidance and your interpretation. Mark missing evidence as unavailable.”
3. Check the output manually. Recalculate changes from the source figures. Check whether cash flow covers the quarter or year to date, whether definitions changed, and whether each citation actually supports the claim. Log unsupported statements and missing values rather than filling them in.
4. Keep a research checkpoint. Record the current evidence, one alternative explanation, what would invalidate the interpretation and the next disclosure to review. Repeat the same worksheet after the next release and compare what changed.

How to report a real test

Publish the input document links, exact prompt, settings, raw output and corrected output. Report the number of checked numerical claims and the number supported by the cited documents, missing-source cases, calculation errors and manual corrections. Define each denominator; record timing and cost only when measured. One run on one company cannot establish general accuracy or investment returns.

For the wider framework and limitations, see our research methodology, data sources and investment research disclaimer. Keep the worksheet and source copies with your research so another reader can reproduce the checks.