AGENT SKILL

summarizing-earnings-calls

Transforms earnings call transcripts into structured summaries with financial guidance, KPI changes, management sentiment, and analyst Q&A highlights. Use when processing quarterly earnings calls, preparing investment research, or tracking company performance narratives.

License 许可
Apache-2.0
最近更新
2026-11-09
Stars
★ 0
Forks
0

01 前言 Front Matter

name
summarizing-earnings-calls
tags
[ "summarization", "equity-research", "investment", "research" ]
source
https://github.com/CaseMark/skills
license
Apache-2.0
language
en
metadata
{ "author": "casemark", "skill_modes": [ "Summarization" ], "document_types": [ "Summary Report" ], "practice_areas": [ "Equity Research", "Investment Management" ] }
description
Transforms earnings call transcripts into structured summaries with financial guidance, KPI changes, management sentiment, and analyst Q&A highlights. Use when processing quarterly earnings calls, preparing investment research, or tracking company performance narratives.
imported_at
2026-09-20
source_path
skills/finance/summarizing-earnings-calls/SKILL.md
source_repository
CaseMark/skills

02 资源文件 1

03 SKILL.md 正文

Summarizing Earnings Calls

When To Use

  • Processing a quarterly or annual earnings call transcript into a structured research summary
  • Preparing investment memos where beat/miss framing and guidance changes are the primary deliverable
  • Tracking management narrative shifts across consecutive quarters for a coverage universe
  • Building a comparable summary format across multiple tickers for portfolio-level review

Inputs To Gather

Collect before drafting. Do not proceed without the required items.

Field Required Purpose
Ticker / Company name Yes Determines industry KPI set and peer context
Quarter & fiscal year Yes Anchors YoY and QoQ comparisons
Consensus estimates (Revenue, EPS minimum) Yes Enables beat/miss framing — omit beat/miss language entirely if unavailable
Prior-quarter guidance (if any) Yes Quantifies guidance revisions
User's coverage context Recommended "SaaS infrastructure analyst" vs. "generalist" shifts emphasis
Specific focus areas Recommended e.g., "margin trajectory," "China exposure," "AI capex"
Prior-quarter summary Optional Enables language-delta and KPI-trend analysis
Industry vertical Auto-infer Drives KPI extraction set (see Industry KPI Matrix below)

Workflow

1. Structural Parse

Segment the transcript into canonical sections:

  1. Operator / Safe Harbor — confirm disclaimer exists; do not summarize it.
  2. CEO Prepared Remarks — strategic narrative, high-level results, thematic emphasis.
  3. CFO Prepared Remarks — revenue, margins, EPS, cash flow, balance sheet, guidance.
  4. Other Executive Remarks — segment heads, CTO on product, etc.
  5. Analyst Q&A — treat each question/answer pair as a discrete unit.

If section headers are missing, infer boundaries from speaker changes and flag with [INFERRED SECTION BREAK].

2. Headline Numbers Extraction

Build the Financial Snapshot Table with these columns: Metric | Reported | Consensus | Beat/Miss | YoY Δ | QoQ Δ. Include at minimum: Revenue, Adj. EPS, Gross Margin, Operating Margin, FCF.

Rules:

  • Always state GAAP vs Non-GAAP basis. If management leads with Non-GAAP, include both. Flag large adjustments (SBC > 10% of revenue, restructuring, one-time items).
  • Use basis points (bps) for margin deltas, percentages for revenue/EPS deltas.
  • Write "Not disclosed" for undisclosed metrics — never infer values.

3. Guidance Analysis

Build the Guidance Table with columns: Metric | New Guidance | Prior Guidance | Δ | Consensus | vs. Street.

Classify each metric as: Raised / Maintained / Lowered / Narrowed / Initiated / Withdrawn.

Quantify magnitude at midpoint: "$150M raise at midpoint, +2.1% vs prior guide" — not just "guidance was raised."

4. Segment & KPI Deep Dive

Extract segment-level disclosures and map to industry-relevant KPIs.

Industry KPI Matrix

Vertical Key KPIs
SaaS / Cloud ARR, NRR/NDR, RPO/cRPO, CAC payback, gross retention, logo adds, DBE
Retail / Consumer Same-store sales, comp growth, inventory turns, store count Δ, e-commerce mix
Banks / Financials NIM, provision for credit losses, CET1 ratio, loan growth, deposit costs, NCO rate
Industrials Book-to-bill, backlog, organic growth, price vs volume, input cost trends
Pharma / Biotech Pipeline milestones, Rx volume, pricing/reimbursement, patent cliffs, R&D % of revenue
Semiconductors Design wins, wafer utilization, ASP trends, inventory days, lead times
Media / Streaming Subscriber adds/churn, ARPU, content spend, ad revenue mix

For each reported segment, produce a mini-table: Segment | Revenue | YoY Δ | Margin | Commentary (use direct management language in commentary).

5. Management Sentiment & Tone

Score on a 5-point scale across four dimensions — overall confidence, demand environment, margin outlook, competitive positioning — each supported by a direct quote.

  • Hedging language flags: Track frequency of "we believe," "we expect," "cautiously optimistic." Compare to prior quarter if available.
  • Tone shifts: Explicitly call out language-strength changes vs. prior quarter (e.g., "shifted from 'confident' to 'cautiously optimistic' on enterprise demand").
  • Forward-looking statements: Note material claims but flag as subject to safe-harbor provisions. Do not treat as commitments.

6. Analyst Q&A Extraction

Structure each exchange with columns: # | Analyst (Firm) | Topic | Key Takeaway | New Info? | Evasion Flag.

  • New Information: Flag any Q&A disclosure absent from prepared remarks — this is where alpha lives.
  • Evasion Detection: Flag non-answers, redirects, or vague language. Evasion on a topic is itself a signal.
  • Follow-up patterns: Note when multiple analysts pressed on the same topic — indicates Street-level concern.

7. Capital Allocation & Balance Sheet

Tabulate: share repurchases, dividends, net debt/EBITDA, M&A commentary, and CapEx — each with current quarter, prior quarter, and commentary.

8. Risks & Watchlist

Compile from three sources:

  • Management-disclosed: Macro, regulatory, supply chain, FX, competitive.
  • Q&A-implied: Topics where multiple analysts pushed or management evaded.
  • Data-inferred: Decelerating growth, margin compression, rising DSO.

Tag each item: [HIGH] [MEDIUM] [LOW].

Output

Assemble the final summary in this order:

  1. Header: Ticker, Company, Quarter, Date, Call Duration
  2. TL;DR (3–5 bullets): Most investable takeaways — what changed, what surprised
  3. Financial Snapshot Table (Step 2)
  4. Guidance Table (Step 3)
  5. Segment & KPI Detail (Step 4)
  6. Management Tone (Step 5)
  7. Analyst Q&A Highlights (Step 6)
  8. Capital Allocation (Step 7)
  9. Risks & Watchlist (Step 8)
  10. Appendix: GAAP/Non-GAAP reconciliation notes, definitions of non-standard KPIs

After presenting the draft, ask the user to verify: completeness of metrics, emphasis balance across sections, accuracy of consensus/guidance figures (highest error-risk fields), sentiment scoring alignment, and formatting for their distribution workflow.

Quality Checks

# Check
1 Every reported metric cites GAAP or Non-GAAP basis
2 Beat/miss framing uses user-provided consensus — never fabricated estimates
3 Guidance deltas quantified at midpoint, not just directional
4 No forward-looking statement presented as a commitment or fact
5 Analyst names and firms attributed where identifiable
6 Evasion flags supported by specific non-answer quotes
7 YoY and QoQ deltas use consistent basis (same quarter, same metric definition)
8 Industry-specific KPIs match the company's actual vertical
9 Risks section includes at least one item from each source (disclosed, Q&A-implied, data-inferred)
10 [VERIFY] tags attached to any data point sourced from inference rather than explicit transcript text
11 No material Q&A information omitted from summary
12 Compliance disclaimer present if output will be distributed externally
13 No editorializing without quoted evidence ("management sounded defensive" requires a supporting quote)
14 Escalate to human review when transcript is partial, has audio-to-text artifacts, or uses unusual reporting frameworks

Reference Files

File Description
references/EARNINGS-SUMMARY-TEMPLATE.md Output structure template for earnings call summaries with all nine sections and appendix

04 同仓库更多 Skills