earnings-preview

name: earnings-preview description: Build pre-earnings analysis with estimate models, scenario frameworks, and key metrics to watch. Use before a company reports quarterly earnings to prepare positioning notes, set up bull/bear scenarios, and identify what will move the stock. Triggers on "earnings preview", "what to watch for [company] earnings", "pre-earnings", "earnings setup", or "preview Q[X] for [company]".

Earnings Preview

Language: English-only by default (this is a fast-turnaround, monitoring-style note, matching the project's tracking-skills English-default rule). Produce bilingual / Simplified-Chinese output only on explicit request (in Chinese, bilingual, --lang zh).

Workflow

Step 1: Gather Context

• Identify the company and reporting quarter
• Pull consensus estimates via web search (revenue, EPS, key segment metrics)
• Find the earnings date and time (pre-market vs. after-hours)
• Review the company's prior quarter earnings call for any guidance or commentary

Step 2: Key Metrics Framework

Build a "what to watch" framework specific to the company:

Financial Metrics:

• Revenue vs. consensus (total and by segment)
• EPS vs. consensus
• Margins (gross, operating, net) — expanding or contracting?
• Free cash flow
• Forward guidance vs. consensus

Operational Metrics (sector-specific):

• Tech/SaaS: ARR, net retention, RPO, customer count
• Retail: Same-store sales, traffic, basket size
• Industrials: Backlog, book-to-bill, price vs. volume
• Financials: NIM, credit quality, loan growth, fee income
• Healthcare: Scripts, patient volumes, pipeline updates

Step 3: Scenario Analysis

Build 3 scenarios with stock price implications:

For each scenario:

• What would need to happen operationally
• What management commentary would signal this
• Historical context — how has the stock moved on similar prints?

Step 4: Setup & Positioning

Cover YTD / relative performance, valuation vs. the name's own 3–5yr history (forward P/E, EV/EBIT, PEG), crowdedness / short-interest / factor exposure where available, and an explicit statement of the market bar — what the buy-side already expects (the whisper) vs. published sell-side consensus. Close with the trade: what you would buy or sell into the print. (Sell-side analogs in the Learning section below.)

Step 5: Catalyst Checklist

Identify the 3-5 things that will determine the stock's reaction:

1. [Metric] vs. [consensus/whisper number] — why it matters
2. [Guidance item] — what the buy-side expects to hear
3. [Narrative shift] — any strategic changes, M&A, restructuring

Step 6: Output

A 1,000–2,500-word (2–4 page) into-the-print note. Required skeleton (the canonical layout per the Learning section below):

• Header Data Block (rating | 12m PT | last price | implied move) + multi-year estimate strip
• Bolded 3–5 sentence thesis — the call AND the setup
• House vs. Consensus vs. Guide table with vs Cons delta column
• Estimate Changes (Prev / Cur / % revision) table
• Bull/base/bear scenario table with quantified reaction bands
• Setup & Positioning (Step 4)
• Sell-side view evolution (卖方观点演变) — required when ≥2 zsxq broker previews cover the name (who-expects-what table + into-the-print revisions; see the Learning section)
• Upside / Downside risk lists + catalyst checklist

A literal one-page quick preview is an explicit opt-in mode only (user says "quick preview" / "one-pager") — never the default.

File location (MUST): save to reports/earnings/<TICKER>_Q<X>FY<YY>_preview_<YYYY-MM-DD>.md (English ticker first per the project filename rule; the _preview marker distinguishes it from post-print earnings-analysis updates in the same bucket). Commit + push in the same task (Conventional Commit, e.g. feat(reports/earnings): ...). The post-print scorecard (loop closure via earnings-analysis) reads this file back — an unsaved preview cannot be graded.

Step 7: Verify & log

Before saving/committing:

1. HTTP-check every URL with a real-browser User-Agent — 200 OK only; drop or replace failures per the link-validation rule.
2. Spot-check 3–5 numbers (consensus EPS/revenue, implied move, historical reaction stats) — each must string-match the URL cited in the same paragraph.
3. Confirm every house figure is paired with its consensus / prior / guide counterpart.
4. Conditional — when ≥2 zsxq broker previews covered the name: confirm the sell-side view-evolution block landed (every institute view dated + /zsxq/pdf/ cited, into-the-print revisions arrow'd with triggers, no blended consensus across disagreeing notes).
5. Append <details><summary>Verification log — YYYY-MM-DD</summary>...</details> listing each check. Spec: reference/citations.md.

Further viewing — explainer videos (optional, but default to including)

When this preview turns on something a reader would struggle to picture from prose alone — the product or segment in focus for the upcoming print when its mechanics matter to the setup (a new chip / device architecture, a manufacturing or scientific process, the unit economics of a subscription or marketplace model, a market-structure concept that drives the KPI you're watching) — attach 1–3 short explainer videos (YouTube and/or Bilibili) so the reader can see it, not just read about it. Default to including them on any topic; omit only when the preview is purely numeric with nothing worth visualizing.

Videos are a teaching aid, NOT a citation — they live in their own slot, never enter the citation chain, and never carry a number.

• Where: a **Further viewing** bullet list at the end of the section the concept lives in, or a single 📺 note beside the hard concept.
• Durable sources only: the company's own product / IR / engineering channel, an OEM or reputable teardown / cutaway channel, or a well-known explainer channel — not a low-view re-upload that will be deleted or is clearly pirated.
• Validate before committing — 200 OK only. YouTube / Bilibili return 403 to bare urllib, so HTTP-check each URL with a real-browser User-Agent; drop dead / private / region-gated links (a 404 link is worse than none). Flag Bilibili that may need login/VPN outside CN: (Bilibili — may require login/VPN outside CN).
• Label honestly: [<what it shows> — <why it helps>](URL). No statistic, price target, share figure, or growth rate is ever attributed to a video (a video can't be string-matched against its source).

Learning from sell-side institutional research

The patterns below are distilled from how Goldman Sachs, Morgan Stanley, UBS, J.P. Morgan, Bernstein, Deutsche Bank, Citi, BofA and Jefferies build into-the-print previews (Jefferies "[Ticker] [Quarter] Preview", MS "Into the Print", UBS "[Ticker] Preview · Evidence Lab inside", JPM "Key Changes", Bernstein "[Ticker] [Quarter] preview"). The single load-bearing differentiator vs. a generic preview is the quantified gap — house vs. consensus vs. guide, shown line-by-line with a delta column. Everything else hangs off that gap. Apply these on top of the workflow above; preserve every existing rule (citations, numerical-accuracy, chart source annotations, language defaults, file naming).

Lead with a Header Data Block, not prose (Step 6 / new lead element). Mirror Jefferies/JPM/MS/UBS/Bernstein: open with a compact key-value block — Rating | 12m Price Target (current vs. PRIOR) with % to PT | last price (as-of date) | 52-week high–low | market cap | ADV — then a multi-year FY estimate strip (Revenue / EBIT or EBITDA / Adj Net Profit / EPS for 2 actual + 2 estimate years). State the valuation method on the PT line (e.g. "34x next-FY EPS", "SOTP", "DCF WACC 12%"). Keep the standing Rating/PT/upside line near the top so the reader sees the call without scrolling. The % to PT is computed against the dated last price above (today), which satisfies the report-date-price rule for your own PT — but any borrowed PT you quote (consensus mean target, a named broker's PT) must show the price on its note's date + the upside it fixed (consensus $130 vs $96 @ 2026-05-20 → +35%), not just vs today's last price.

Open with a bolded 3–5 sentence thesis (the call AND the setup in one breath). Jefferies/MS lead with a verdict like "meets reduced rev/GM cons but misses EBIT by ~35%; cons still too high, too early to bottom-fish, results May 26." State the call, the gap to consensus, and the trade into the print up front — do not bury it in a generic output list.

House vs. Consensus vs. Guide table with an explicit vs Cons delta column (Step 1 + Step 6). This replaces the plain consensus table. Carry the delta on every line — revenue (total + by segment), gross margin (total + by segment), EBIT, EPS — plus a "vs guide range (low/mid/high)" note since institutions position forecasts against the company's own guide, not just consensus (Jefferies "JEFe vs Cons", JPM revenue "versus consensus"). Consensus must be cited to a deep URL the agent actually fetched and that string-matches the figure — e.g. the Yahoo Finance analysis page, a stockanalysis.com forecast page, or a dated zsxq broker PDF labelled Analyst view: (cite via the /zsxq/pdf/<file_id>/<name> direct-download route). NEVER attribute a number to Visible Alpha / Bloomberg / FactSet unless quoting a secondary source that itself prints it (use a source-chain label). Always date-stamp the consensus snapshot; each house number must be paired with its consensus/prior/guide counterpart per the numerical-accuracy rule.

Sell-side view evolution (卖方观点演变) — mandatory when ≥2 zsxq broker previews cover the name (Step 4 + Step 6). Pre-print disagreement IS the setup signal — never blend the notes into a fake consensus. Render a who-expects-what table next to Setup & Positioning: | Institute | Date | Rating / PT | Estimate vs Cons | Swing factor they flag / what would prove them right | — every row dated (the filename's -YYMMDD suffix is the authoritative pub date; sanity-check against create_time) and cited via the /zsxq/pdf/<file_id>/<name> direct-download route. Call out into-the-print revisions explicitly — an institute that updated numbers/PT within the last month is a signal; give the stated trigger (channel checks, order data, a peer's print) and render it as a dated arrow (UBS $120 (26-05) → $135 (26-06, post-peer beat)); a 2026-03 view and a 2026-06 view from the same institute are two different views, not duplicates. Detect revisions and PT dispersion (min/median/max, spread %) mechanically BEFORE re-reading any PDF — STRICTLY read-only pre-pass: /opt/anaconda3/bin/python3 with sqlite3.connect('file:db/stock_price_target.db?mode=ro', uri=True), SELECT research_institute, rating, price_target, target_currency, report_date, report_file_id, upside_pct FROM price_targets WHERE company_ticker=? ORDER BY research_institute, report_date (writes stay exclusively with scripts/persist_pts.py). Where institutes read the same datapoint oppositely (opposite ratings, PTs >20% apart), keep both rows and let the bull/bear scenario table say whose evidence resolves it.

Add an Estimate Changes (Prev vs. Cur, % revision) table (Step 1 + Step 6). A preview usually moves numbers — show before/after. Mirror JPM "Key Changes (Prev/Cur/%)" and Jefferies "Change to JEFe". Place it as a sibling to the House-vs-Cons table.

Decompose any guide-change call into named $ and % contributions (Step 2). Bernstein-Adobe pattern: "Q1 beat +$0.16, buyback +$0–3.50, M&A ~$0." Each input must trace to a source containing that number; label the summed figure as a derived calc per the numerical-accuracy rule.

Quantify the Stock Reaction column — make it a band, not an adjective (Step 3). Tie each scenario row to (a) the options-implied move computed from the at-the-money straddle and (b) the historical average 1-day post-print move over the last 4–8 quarters, and anchor each row to the specific KPI threshold that triggers it. Templates: Bernstein Akeso ("HR 0.70–0.72 → +10–20%; HR > 0.75 → −10–20%"), UBS Canada Goose ("balanced upside/downside skew, historical ±11.7%"). Replace "stock could move" with a percentage band tied to a named scenario.

Setup & Positioning section (Step 4 in the workflow above). Cover YTD / relative performance, valuation vs. the name's own 3–5yr history (forward P/E, EV/EBIT, PEG), crowdedness / short-interest / factor exposure where available (JPM Style/Quant %-rank panel; UBS "crowdedness below sector average"), and an explicit statement of the market bar — what the buy-side already expects (the whisper) vs. published sell-side consensus, as a third column alongside house and consensus. Frame the deliverable as "what we would buy or sell into the print." This upgrades the thin one-line "Trading setup."

Attempt named alt-data / channel checks before generic web search (Step 1). Cite deep URLs: app/web panels (Sensor Tower, SimilarWeb), card-spend trackers, sector trackers (STR RevPar, Planespotters deliveries, semis: DRAM/NAND contract prices, WFE $), and the prior-quarter call transcript for guidance language. Broker proprietary labs (UBS Evidence Lab, Deutsche Bank DBDig survey, GS surveys) are the analog edge — when their reads come from the local db/zsxq.db library, label them *Analyst view:* and never blend them into a primary-filing citation.

Sector / multi-name catalyst-preview mode (second operating mode). When previewing a coverage universe rather than a single name, produce a dated catalyst/event grid that rates each upcoming catalyst by importance (High / Very important) and expected surprise direction, plus a per-name beat / in-line / miss call with vs-cons deltas (MS "Catalyst Preview: What's Ahead?", GS sector "1Q results preview", UBS sector "Earnings Preview"). Respect the skill's English-default convention for this tracking-style output. Boundary with [[catalyst-calendar]]: catalyst-calendar owns the calendar itself (dated grids across event types over a universe); this mode exists only for per-name into-the-print calls (beat / in-line / miss with vs-cons deltas and scenario bands) on names reporting in the window. Route pure scheduling requests to catalyst-calendar, and accept its handoffs for the names worth a call.

Split Risks into Upside vs. Downside bullet blocks (Step 5). GS Moderna / MS Broadcom / Bernstein convention: two explicit lists rather than one catalyst checklist, each risk tied to its KPI and the scenario it would push the print toward.

Title encodes the call and the hook (Step 6). Not just ticker/quarter — e.g. "Memory Cost Pressure Not Yet Peaked; Cons Still Too High", "It's A Tricky One", "We See a Balanced Upside/Downside Skew", "Roadmap to Profitability". Numbers are always paired — every house figure shown next to consensus, prior, or guide ("PT raised 475→490", "EPS $3.05 vs cons $2.90"); a standalone number reads as incomplete. Date-stamp every estimate and consensus snapshot.

Build in loop closure (Important Notes). Institutional previews are accountable — banks publish a post-print scorecard (JPM "As We Previewed, [Ticker] Delivered…") grading the call against the actual print. After results, chain to the earnings-analysis skill as the natural follow-up so the preview's calls get graded.

Primary-source-first & development-over-time rule (MANDATORY)

The user's standing preference for every report-producing skill: reference the 10-K / 10-Q / original investor-relations materials as much as possible, cite them at page level, and present the material so the reader can see the company's development over time — what's new this period.

1. Source-preference order for any company fact. (1) The company's own filings — 10-K / 10-Q / 8-K / DEF 14A / 20-F / 6-K / S-1 on EDGAR, or the non-US equivalent (年度报告 via cninfo, HKEX annual report, 有価証券報告書, 사업보고서); (2) original IR materials — earnings press release, earnings / investor-day deck, call transcript, shareholder letter; (3) third-party industry research; (4) news. Business sections especially run on the 10-K. For business fundamentals — what the company does, segment structure, products and how they make money, customers and concentration, competition, manufacturing / supply chain, IP, regulation, headcount — the 10-K is the default first-stop source (Item 1 Business, Item 1A Risk Factors, Item 7 MD&A, each cited with page), refreshed by the latest 10-Q for in-year changes; non-US equivalents use the annual report's business chapter (年度报告 经营情况讨论与分析, 有価証券報告書 事業の状況). Never cite a news rewrite for a fact that lives in a filing or an IR original — chase the original. Sell-side / zsxq broker notes are NOT displaced by this rule: they remain the separate *Analyst view:* layer (with their own page-level cites) and are never blended into the company-fact layer.

2. 10-K / 10-Q / annual-report citations must carry page numbers. Format: [NVDA FY2025 10-K, p. 42 — Segment results](https://www.sec.gov/...). When the EDGAR HTML doc makes the print page hard to pin down, give the Item + note/section heading instead (Item 2 MD&A — Data Center revenue, Note 17 — Segment Information) so the reader lands within one page-flip of the number. A bare [10-K](url) with no page/section locator fails the citation bar. The same locator discipline applies to prospectuses (page), IR decks (slide number), and non-US annual reports (第 N 页 / p. N).

3. Present development over time — "what's new". Do not render the company as a static snapshot. Wherever the output's structure allows, frame disclosures diachronically: trace the same line item across consecutive 10-Ks / 10-Qs (segment revenue & mix, risk factors added / dropped, customer-concentration %, capacity / capex, backlog, headcount, guidance language) and state explicitly what is NEW in the latest filing versus the prior one. Preferred presentations: an evolution table (FY23 → FY24 → FY25, each column cited to its own filing + page) and/or a short "What changed this period / 本期新变化" callout where the section covers a recurring disclosure.

4. English originals stay English — even in Chinese-language reports. When the original source is English (SEC filing, English IR deck / transcript / press release), cite and quote the English original directly; do not substitute a Chinese-media rewrite for language consistency. Symmetric with the existing original-language rule: the original's language always wins, whichever it is.

AI / Robotics / Semiconductor — detailed-narrative rule (MANDATORY)

When the subject of the output — the ticker, theme, sector, ETF holdings, deal, or any name that materially drives the analysis — sits in AI (foundation models, AI software/agents, AI infrastructure: datacenter compute, networking, power), robotics (humanoids, industrial automation, AMRs, actuators / reducers / sensors / end-effectors), or semiconductors (fabless, foundry, IDM, memory/HBM, equipment/WFE, materials, EDA/IP, advanced packaging), give those names a detailed narrative treatment, not summary bullets:

• Write full narrative prose for the sector-relevant sections — mechanism and causality ("X drives Y because Z"), not headline restating. Bullets may organize the prose but never replace it.
• Cover the sector-specific dimensions that apply:
  • Technology position & roadmap — process node / architecture / model-capability cadence vs named competitors (e.g., N2 vs 18A, HBM3E→HBM4, GB200→Rubin, Optimus gen-3 vs Figure 03).
  • Supply-chain position — key suppliers and customers up/down the chain, single-source chokepoints (TSMC/CoWoS, EUV, HBM), where pricing power sits, content-per-unit ($ per GPU / per robot / per vehicle).
  • AI demand linkage — the explicit path from AI capex to this name's P&L (orders → backlog → revenue recognition) with the actual disclosed numbers, never a generic "AI beneficiary" label.
  • Robotics linkage — design-win status, which platforms (Tesla Optimus, Figure, Unitree, domestic Chinese OEMs), volume and timeline realism vs the hype cycle.
  • Cycle context — where the semi / memory-pricing / AI-capex cycle stands right now and what that implies for forward estimates.
  • Geopolitics & export controls — US BIS rules, China localization, tariff exposure, entity-list status where relevant.
• Quantify the narrative. Each dimension covered should carry at least one sourced number (TAM, ASP, capacity, units, share). All figures obey the project's numerical-accuracy rule — every number traces to a URL or PDF page cited in the same paragraph.
• Engage the sell-side view. Where the zsxq library or other broker sources are in scope for this skill, the AI/robotics/semi narrative must engage the institute view (PTs, estimate revisions, cross-broker disagreement) rather than ignoring it.

This rule deepens the skill's existing output format — it never replaces or shortens the required structure. For subjects outside these sectors, the skill's baseline depth applies unchanged.

Important Notes

• Consensus estimates change — always note the source and date of estimates
• "Whisper numbers" from buy-side surveys are often more relevant than published consensus
• Historical earnings reactions help calibrate expectations (search for "[company] earnings reaction history")
• Options-implied move tells you what the market expects — compare to your scenarios