Guide
Earnings Breakdown: How to Read an Earnings Report With AI
AI can compress a long filing in seconds. It can also invent a confident EPS figure, blur reported and adjusted results, or compare the quarter with the wrong estimate. This guide gives you a six-layer earnings read and five checks to run before you trust any AI or human take.
Free plan includes 3 AI research analyses per week. No card required.
Want the current numbers? Check the live Market Snapshot — and get the verified Earnings Radar every Friday in the free weekly newsletter.
Week-of-earnings checklist
Use this during the 48 hours around a print. Save the inputs before the release so the result cannot quietly rewrite what you expected.
- Know the consensus EPS and revenue you are comparing against, including the estimate’s as-of date.
- Separate numbers in the earnings press release from figures in the filed 10-Q or 10-K.
- Flag restructuring, acquisition, tax, impairment, and other one-time items before celebrating a beat.
- Note the share count and whether the company is discussing basic or diluted shares.
- After the print, define the volatility window and write down what new evidence would change your thesis.
The six layers
Read in order. Each layer narrows the gap between a polished summary and the underlying evidence. The prompts help AI organize the work; the verification step keeps you in control of the facts.
- 1
Headline beat or miss
Start by defining the comparison. Reported revenue or EPS can beat the latest consensus while still falling from the prior year, and an estimate revised down a week earlier creates a different signal from a beat against a stable estimate. State the period, currency, accounting basis, and comparison date before interpreting the result.
- Ask the AI…
- “List reported revenue and EPS, the consensus figures, the absolute and percentage surprise, the comparison period, and the as-of date for the consensus. Label reported and adjusted EPS separately.”
- Verify yourself…
- Match revenue and net income to the earnings release and income statement. Check the provider’s estimate timestamp and confirm whether its EPS consensus is GAAP or adjusted.
- 2
Revenue quality
A higher top line is not automatically better revenue. Look at organic growth, acquisitions, currency effects, customer or product mix, and whether demand was pulled forward. Then compare management’s current language with its previous guidance: maintained numbers paired with softer wording can still matter.
- Ask the AI…
- “Break revenue growth into reported, organic, acquisition, currency, segment, and geographic effects where disclosed. Quote the exact guidance language and identify what changed from the previous quarter.”
- Verify yourself…
- Check the revenue line, segment note, and management discussion in the filing. Reconcile any non-GAAP organic-growth figure to the company’s published table rather than accepting an AI calculation without inputs.
- 3
Margins and costs
Margins show whether growth is becoming more or less valuable. Compare gross and operating margin with both the prior quarter and prior-year period. Separate lasting changes—pricing, product mix, labor, logistics, or scale—from temporary items such as restructuring or an unusual legal charge.
- Ask the AI…
- “Show gross and operating margin on reported and adjusted bases, calculate the year-over-year change in percentage points, and classify each major cost driver as structural, cyclical, or one-time with supporting evidence.”
- Verify yourself…
- Recalculate margins from the income statement. Read the non-GAAP reconciliation and footnotes to see exactly what was excluded and whether the same ‘one-time’ cost has appeared before.
- 4
Cash and balance sheet
Accounting profit matters more when it becomes cash. Compare operating cash flow and free cash flow with net income, while accounting for working-capital timing and capital spending. Then review cash, debt, maturities, buybacks, and stock-based compensation so an EPS improvement is not mistaken for a stronger balance sheet.
- Ask the AI…
- “Explain the gap between net income and operating cash flow, calculate free cash flow using a stated definition, summarize debt and liquidity, and show whether buybacks actually reduced diluted shares after stock compensation.”
- Verify yourself…
- Use the cash-flow statement, balance sheet, debt footnote, and diluted weighted-average share count. Confirm the free-cash-flow definition because companies and providers may calculate it differently.
- 5
Forward look
The market often reacts more to what comes next than to the completed quarter. Record the guidance range, midpoint, covered period, and assumptions. Listen for changes in confidence, demand visibility, pricing, backlog, or spending—and note important questions management avoids rather than filling the silence with an invented answer.
- Ask the AI…
- “Compare new guidance with prior guidance and current consensus. List the assumptions management stated, identify unanswered questions, and distinguish direct quotations from your inference.”
- Verify yourself…
- Check the release, prepared remarks, and transcript. Compare ranges on the same accounting and currency basis, and do not treat an analyst paraphrase as management guidance.
- 6
Thesis test
Finish with your case, not the model’s conclusion. A useful earnings read asks whether the new evidence confirms, weakens, or invalidates the reasons you were following the company. Price movement is not the thesis: the test is whether the operating facts crossed a threshold you defined beforehand.
- Ask the AI…
- “Map each new fact to my written thesis. Classify it as confirming, weakening, invalidating, or unresolved, explain why, and name the next observable evidence that would change the classification.”
- Verify yourself…
- Compare the answer with your pre-earnings notes. Make sure the AI did not introduce a new assumption, hide contrary evidence, or rewrite your threshold after seeing the result.
The five sanity checks
Fluency is not verification. Run these checks after the summary is complete and before using it to update your research notes.
1.Every figure has a source and an as-of date
A number without provenance is a lead, not evidence. Require the filing line, company table, or named provider plus the period and retrieval date. This is especially important for consensus estimates, which change as analysts update models.
2.Reported and adjusted are never blended
GAAP results and management-adjusted results answer different questions. Keep both, label both, and show the reconciliation. Never compare reported EPS with adjusted consensus or use an adjusted margin beside reported costs without saying so.
3.Per-share math matches the current share count
Splits, buybacks, issuance, and stock compensation can make per-share trends misleading. Check diluted weighted-average shares for the period and confirm that historical comparisons were adjusted for any split.
4.A beat uses a stated consensus, not vibes
‘Better than expected’ is meaningless unless expected by whom and when is clear. Record the provider, analyst count where available, estimate basis, and as-of date. A beat against one source may be a miss against another.
5.The bear case is specific enough to falsify
A token risk paragraph does not test a conclusion. Ask for a measurable condition—a margin threshold, demand indicator, cash-conversion level, or guidance change—that would prove the optimistic interpretation wrong.
Mini example (hypothetical — not advice)
Imagine a fictional company called Northstar Components. It reports quarterly revenue of 520 million hypothetical dollars and adjusted EPS of 1.08. A data provider’s consensus, captured the morning before the release, was 505 million and adjusted EPS of 1.02. These numbers are invented only to demonstrate the process; they do not refer to a real company or security.
Layer 1 — headline: Northstar is 15 million above that revenue consensus and 0.06 above adjusted EPS consensus. That is not enough to call the quarter strong. The comparison must say “adjusted,” name the provider and as-of time, and separately show reported EPS from the filing. If reported EPS were 0.74, the gap would need an explanation rather than being blended into the headline.
Layer 2 — revenue quality: suppose revenue grew 8%, but an acquisition added six percentage points and the legacy business grew 2%. Ask the AI to show that bridge and quote management’s guidance language. Then verify the acquisition contribution in the release or filing instead of repeating the 8% headline as if it were all organic demand.
Checks 1 and 2: attach a source and as-of date to each figure, then keep reported and adjusted EPS in separate columns. If either check fails, the summary is unfinished. The exercise produces a cleaner description of the quarter, not a buy or sell conclusion.
How Ticker-Terminal fits
Earnings Intelligence brings verified earnings dates, reminders, and simplified revenue and EPS breakdowns into the same research workflow. It helps you organize the print quickly without treating an AI summary as the source of truth.
The free plan includes 3 AI research analyses per week with no card required. Run the same ticker free, then re-check the two filing numbers the conclusion depends on most. That final step is deliberate: the tool should shorten the reading, not replace your verification.
Related reading
Learn
Free investing education from behaviour and cycles to AI research.
How to research a stock with AI
Put this earnings read inside a complete seven-step research workflow.
AI investing: where it helps and misleads
Why verification matters more than a fluent answer.
Best AI for stock research
Compare research tools by data quality, transparency, and repeatability.
Read the next print with a verification habit
Get 3 AI research analyses per week on the free plan. No card required. Use the six layers, check the source numbers, and keep the decision yours.
Start freeFree guide series
Get the “how to research a stock” checklist
Two emails. The 7-step checklist now, then the next guide in the series — starting with Which AI is best for stock research.
- The 7-step “how to research a stock” checklist, sent immediately
- The next recommended guide in the series, two days later
- Occasional new guides only — no daily market noise, unsubscribe in one click