💡 왼쪽 원문을 읽으면서 오른쪽에 따라 써보세요. Tab 키로 힌트를 받을 수 있습니다.

원문 렌더가 준비되기 전까지 텍스트 가이드로 표시합니다.

Who this is for

This guide targets undergraduate seniors, masters students, and MBA candidates aiming for Equity Research Analyst (ERA) or Research Associate (RA) roles. We translate phrases like "industry analysis, company analysis, report writing" from a job description into concrete daily actions, documents, and meetings. Rather than "you write reports", we show how an out-of-consensus call gets backed by evidence and committed to paper. Even if you already have one or two years on the sell-side and are weighing a buy-side jump, the comparison frame here is useful.

In the securities industry, "writing" is far more tool-heavy, convention-bound, and metric-driven than outsiders assume. So this guide bundles tools, deliverables, evaluation metrics, and compensation into a single picture.

Why research is fundamentally a writing job

Research compresses into three verbs: read the company, build the model, write the conclusion. Modeling actually consumes less time than writing. A senior analyst produces roughly 80-120 notes plus 4-8 initiation reports per year — one document every 1.5 days on average. The model is just an instrument; the document is the unit of evaluation.

Reports are not summaries — they move capital. A single line from Morgan Stanley or a Goldman target price change can swing a stock 3-7%. That is why accuracy, tone, convention, and compliance disclosures all matter at the same time. Once you internalize that research is a writing job, your preparation direction becomes much clearer.

Sell-side vs Buy-side: two paths

Sell-side analysts at brokerage firms write research to sell to institutional investors. Mirae Asset, Korea Investment, Samsung, NH, Goldman Sachs, and Morgan Stanley belong here. Reports are public, the sales team carries them to clients, and evaluation runs through II Rankings, Greenwich Surveys, and StarMine.

Buy-side analysts at asset management firms write internal-only research. Mirae Asset Global Investments, Korea Investment Trust Management, Fidelity, Capital Group, and T. Rowe Price are examples. The report has an audience of one — a single portfolio manager. The evaluation metric is the fund's return. Sell-side rewards "read by many"; buy-side rewards "contributed to performance."

| Dimension | Sell-side | Buy-side |

| --- | --- | --- |

| Audience | Many institutional investors | 1-3 internal PMs |

| Output frequency | 2-3 notes per week | 4-8 memos per month |

| Report length | 5-30 pages | 2-5 page memos |

| Model sharing | External | Internal only |

| Evaluation | II Ranking, client votes | Fund return contribution |

| Comp structure | Base + volatile bonus | Base + fund-perf bonus |

| Work hours | 70-90 hrs in earnings season | 50-70 hrs |

| Media exposure | High (TV, interviews) | Almost none |

Most entrants start on the sell-side because there are more seats and faster training in modeling and writing. The classic path is to jump buy-side in years three to five.

A real day: 9am tape → 11am model update → 2pm conference call → 4pm note → 7pm review

A Seoul Yeouido sell-side analyst's day looks roughly like this.

- 07:00 Arrive, scan overnight US markets, FX, commodities, and sector news on the Bloomberg Terminal.

- 07:30 Morning meeting. Summarize yesterday's events on your covered names in three minutes for the sales force.

- 09:00 KOSPI opens. Tape comments, intraday price monitoring of covered names.

- 10:00-12:00 Model update. Drop yesterday's earnings into Excel and feed guidance into next-quarter estimates.

- 12:30 Lunch — usually a sandwich at your desk, or an IR meeting (IR officer visits to update over lunch).

- 14:00 Conference call. Sit through a 30-60 minute call on a covered name or global peer (e.g., TSMC, ASML). Record via Tegus.

- 15:30 Write the note. Distill guidance changes, product commentary, and competitive dynamics into 2-3 pages.

- 17:30 Submit for compliance review. Any target price change requires both the head of research and compliance to sign off.

- 19:00 Prep tomorrow's morning meeting comment, finalize the one-liner for the sales force.

- 20:00 Leave the office. In earnings season this slips to 22:00 often.

In earnings season (six weeks after quarter end), this pattern repeats weekly. Outside that, you have time for long sector reports or initiations.

Deliverables and modeling tool stack

The annual output of a research department splits into six categories.

- Initiation Report: starting coverage on a new name. 30-80 pages. Industry, company, model, valuation, risk — all of it. Four to eight per year.

- Update Note: reacts to earnings, news, or guidance changes. 2-5 pages. Highest frequency.

- Sector Report: a full industry view. 50-150 pages. Quarterly or semi-annual.

- Quarterly Preview/Review: before or after earnings. Consensus comparison is the heart.

- Morning Meeting Comment: the day's one-liner on covered names. The sales force takes it to clients.

- Custom Note: a one-on-one note for a specific client (e.g., Tiger Global). Heavy in II vote season.

Length, tone, and conclusion placement differ across deliverables. An initiation opens with "where is this industry in five years"; an update note opens with "what changed on yesterday's call."

| --- | --- | --- | --- |

For juniors, Bloomberg + FactSet + Excel are enough. As you become senior, Tegus and AlphaSense become indispensable.

Model types and sector applications

| Model | Best-fit industries | Key variables |

| --- | --- | --- |

| DCF (Discounted Cash Flow) | Stable cash flow, infrastructure, telecom | WACC, terminal growth |

| EV/EBITDA Multiple | Capital-intensive, cyclical | Forward EBITDA, multiple |

| P/E Multiple | General equities, large caps | Forward EPS, multiple |

| P/B (Price-to-Book) | Banks, insurance, asset-heavy | ROE, cost of capital |

| SOTP (Sum-of-the-Parts) | Holding cos, conglomerates | Per-segment value |

| Residual Income | Banks, financials | ROE, cost of equity, BV |

| Dividend Discount Model | Utilities, REITs | Dividend growth, discount rate |

| --- | --- | --- | --- |

Most reports show DCF + EV/EBITDA + P/E side by side and use an average to set the target price. A favorite interview question is "why this multiple?" Juniors typically cover 1-2 sectors; seniors go deep on 4-6 names.

A simple DCF model (Python)

def calculate_dcf(

initial_revenue: float,

revenue_growth: list,

ebitda_margin: list,

tax_rate: float,

capex_pct: float,

nwc_pct: float,

wacc: float,

terminal_growth: float,

shares_outstanding: float,

net_debt: float,

) -> dict:

"""Project FCF over forecast horizon, compute equity value per share."""

years = len(revenue_growth)

revenue = [initial_revenue]

for g in revenue_growth:

revenue.append(revenue[-1] * (1 + g))

revenue = revenue[1:]

ebitda = [r * m for r, m in zip(revenue, ebitda_margin)]

da = [r * 0.04 for r in revenue]

ebit = [e - d for e, d in zip(ebitda, da)]

nopat = [x * (1 - tax_rate) for x in ebit]

capex = [r * capex_pct for r in revenue]

change_nwc = [

(revenue[i] - revenue[i - 1]) * nwc_pct if i > 0 else revenue[i] * nwc_pct

for i in range(years)

]

fcf = [n + d - c - w for n, d, c, w in zip(nopat, da, capex, change_nwc)]

pv_fcf = sum(f / (1 + wacc) ** (i + 1) for i, f in enumerate(fcf))

terminal_value = fcf[-1] * (1 + terminal_growth) / (wacc - terminal_growth)

pv_terminal = terminal_value / (1 + wacc) ** years

enterprise_value = pv_fcf + pv_terminal

equity_value = enterprise_value - net_debt

price_per_share = equity_value / shares_outstanding

return {

"enterprise_value": enterprise_value,

"equity_value": equity_value,

"price_per_share": price_per_share,

"pv_fcf": pv_fcf,

"pv_terminal": pv_terminal,

}

Hypothetical semiconductor company example

result = calculate_dcf(

initial_revenue=50_000, # units: billion KRW

revenue_growth=[0.15, 0.12, 0.10, 0.08, 0.06],

ebitda_margin=[0.35, 0.36, 0.37, 0.37, 0.36],

tax_rate=0.22,

capex_pct=0.12,

nwc_pct=0.05,

wacc=0.085,

terminal_growth=0.025,

shares_outstanding=600_000_000,

net_debt=-5_000, # net cash

)

print(f"Equity value per share: {result['price_per_share']:,.0f} KRW")

A production model adds scenario analysis, sensitivity tables, and quarterly decomposition, but the skeleton is the same. When the interviewer asks you to whiteboard a DCF, you must draw these five steps (revenue → EBITDA → FCF → discount → per share) without hesitation.

II Rankings, Greenwich, StarMine evaluation

External evaluation of sell-side analysts runs through three surveys.

- Institutional Investor (II) Rankings: released around March every year. Global institutional investors vote for sell-side analysts. The number-one finisher gets the "Top Ranked Analyst" title and a direct boost to bonus.

- Greenwich Associates: strongest in Europe. Based on institutional PM interviews.

- Refinitiv StarMine: quantitative metrics (estimate accuracy, recommendation returns). An objective complement to human votes.

In Korea, Hankyung Best Analyst and Maeil Best Analyst carry equivalent weight to II. Winning Best Analyst typically lifts next-year compensation 20-40%. Breaking into the II Top 3 makes you the headhunter's first call and dramatically raises your buy-side jump leverage.

Korean sell-side houses

| House | Strong sectors | Notes |

| --- | --- | --- |

| Mirae Asset Securities | IT, semis, global macro | Strong overseas network |

| Korea Investment | Financials, autos | Many Hankyung Best wins |

| Samsung Securities | Semis, displays | Group company synergy |

| NH Investment | Chemicals, consumer | NH-group stability |

| KB Securities | Financials, real estate | Group IB linkage |

| Shinyoung Securities | Small-mid caps, value | Value-investing color |

| Kiwoom Securities | Internet, gaming | Retail data edge |

| Daishin Securities | Healthcare, biotech | Analytical depth |

| Meritz Securities | China exposure, macro | Macro-strong |

| Hana Securities | Financials, utilities | Stable coverage |

New RA hiring runs in the fall (seasonal) and spring (rolling). Each house typically takes 5-15 new RAs.

Japanese sell-side houses

| House | Strengths | Notes |

| --- | --- | --- |

| Nomura (野村證券) | Japan-wide, global IB | Largest Japanese house |

| Daiwa (大和証券) | Japan small-mid caps | Number two Japanese house |

| SMBC Nikko (SMBC日興) | Financials, megabanks | SMBC Group |

| Mizuho (みずほ証券) | Autos, industrials | Mizuho FG |

| SBI Securities (SBI 証券) | New names, fintech | Retail strength |

| Mitsubishi UFJ Morgan Stanley | Global collab | JV with MS |

Japanese sell-side research is Japanese + English by default. Recently, demand has grown for Korean and Chinese speakers to cover regional peer comparisons.

Global IB research houses

| House | HQ | Strengths |

| --- | --- | --- |

| Goldman Sachs | New York | Most sectors, M&A linkage |

| Morgan Stanley | New York | Tech, healthcare |

| JP Morgan | New York | Financials, FICC linkage |

| Bank of America | New York | Industrials, consumer |

| Citi | New York | Emerging markets, macro |

| Wells Fargo | San Francisco | Real estate, REITs |

| Jefferies | New York | Small-mid cap IPO linkage |

| UBS | Zurich | Luxury, global wealth |

| Barclays | London | European industrials |

| HSBC | London | Emerging Asia |

Global IB research desks sit in Hong Kong, Tokyo, and Seoul. The Korea-covering bench at a foreign bank typically numbers 5-15 analysts. The foreign analysts you see at Korean IR events are largely a known cast.

Buy-side research

| Firm | HQ | AUM | Notes |

| --- | --- | --- | --- |

On the buy-side a single analyst typically covers 10-30 names deeply and sends Buy/Hold/Sell memos to PMs. The biggest difference from sell-side: there is no external distribution, and your call performance flows directly into the fund's annual return.

Quant vs qualitative: the model is the tool, judgment is the substance

Juniors sometimes assume "great at Excel equals great at research." Half true. A model does not value a company; a model is a conditional statement that says "if my assumptions are right, value is this." The assumptions themselves (revenue growth, margin, WACC) are the substance.

Good analysts read the industry structure more than the model. Semi cycles run on roughly 18-month DRAM price waves. Autos run on SAAR (seasonally adjusted annual rate) and average selling price. In healthcare, one line of Phase 3 data swings market cap 50%. Without this structural sense, no matter how precise the DCF, your conclusion will look identical to consensus.

When an interviewer asks "why are you different from consensus," the answer has to live inside the industry assumptions, not inside the model.

AI impact in 2026: BloombergGPT, AlphaSense, Tegus AI, Hebbia

Between 2024 and 2026 the research department's AI adoption accelerated sharply. The key shifts:

- BloombergGPT: an LLM embedded in Bloomberg. News summaries, bond price commentary, automated IR document analysis.

- AlphaSense Generative Search: keyword search becomes natural-language Q&A. You can type "what did TSMC say about N3 on the Q1 2024 call?" verbatim.

- Tegus AI Summaries: auto-summarization of expert call transcripts. A senior PM reads a 30-minute interview as a five-minute digest.

- Hebbia: ingests 10-Ks, 10-Qs, and slides at once. Generates comparison tables automatically.

- Daloopa: extracts fundamentals automatically. Lands directly into Excel.

- AI transcript analysis: ChatGPT or Claude detects tone shifts (positive → cautious) in IR materials.

The result: junior RAs spend 70% less time on data wrangling. Seniors spend more time on differentiation — "why should this name be seen differently?" RAs whose role was pure data entry are likely to disappear inside five years. Those who craft industry hypotheses and write well survive.

A simple example of analyzing IR transcript tone with an AI-style helper

from collections import Counter

def analyze_call_tone(transcript: str) -> dict:

"""Very crude keyword-based tone analysis. Production uses LLMs."""

positive = ["strong", "robust", "exceed", "outperform", "accelerate", "raise"]

negative = ["weak", "decline", "miss", "slow", "cautious", "lower"]

words = transcript.lower().split()

counter = Counter(words)

pos_score = sum(counter[w] for w in positive)

neg_score = sum(counter[w] for w in negative)

net = pos_score - neg_score

if net > 5:

tone = "positive"

elif net < -5:

tone = "cautious"

else:

tone = "neutral"

return {

"positive_hits": pos_score,

"negative_hits": neg_score,

"net_score": net,

"tone": tone,

}

sample = "Q1 revenue exceeded guidance, margins remained robust. We raise FY guidance."

print(analyze_call_tone(sample))

In real practice, instead of keywords, you feed the full transcript to an LLM and ask "how has the tone shifted versus the same-quarter call last year?" That kind of comparison is becoming standard.

Writing the report and breaking consensus

A good report shows three things on the first page alone.

- Conclusion (Investment Thesis): one line. "TSMC: N3 yield stabilization supports a 15% upside to 2026 EPS vs consensus."

- Why: three points. (1) N3 yield reaches 90% (2) Apple M5 ramps in volume (3) NVIDIA AI demand persists.

- What Can Go Wrong: three risks. (1) Geopolitics (2) Samsung Foundry 2nm catch-up (3) Mac demand slowdown.

Juniors most often botch this structure. Junior reports usually open with one to two pages of "industry background" and bury the conclusion on page three. Seniors put the conclusion in the first line and use the body to reinforce it. The technique is called the Pyramid Principle and McKinsey-trained writers preach it constantly.

The market impact of a report is proportional to how far it stands from consensus. "Our EPS estimate is 8% above consensus" is a single line that signals you differ from the street average.

| Call type | Definition | Evaluation |

| --- | --- | --- |

| Above consensus + right | Above the street, actual above | Best call |

| Above consensus + wrong | Above the street, actual below | Biggest demerit |

| In-line consensus + right | At the street average, right | Mediocre |

| Below consensus + right | Below the street, actual below | Best call |

| Below consensus + wrong | Below the street, actual above | Big demerit |

An analyst who only puts out in-line calls will never win an II vote. The balance is: take differentiated risk, but anchor it in clear evidence.

A real good report (format and tone)

A typical sell-side update note's first page skeleton looks like this.

[Cover] TSMC (2330 TT) — Maintain Buy

Target Price: NT$1,200 (from NT$1,050, +14%)

Current Price: NT$1,050 (as of 2026-05-27)

Upside: 14%

[Executive Summary]

N3 yield reached 90% per supply chain checks, supporting 2026 EPS upgrade.

We raise FY26 EPS to NT$72 (from NT$65), 11% above consensus of NT$65.

Reiterate Buy with target NT$1,200 (22x FY26 PER, premium to 10-yr avg of 18x).

[Key Points]

1. N3 yield 90% — sustainable through 2H26 per fab visit.

2. Apple M5 ramp in Q3 — incremental NT$80B revenue.

3. NVIDIA Blackwell sustained demand through 2027.

[Risks]

1. Geopolitical (China-Taiwan tension).

2. Samsung Foundry 2nm catch-up.

3. Macro slowdown impacting Mac demand.

[Valuation]

Target NT$1,200 = 22x FY26 PER x NT$54 mid-cycle EPS.

DCF cross-check: NT$1,180 (WACC 8.5%, terminal 3%).

[Compliance disclosure]

Author owns no position in TSMC. Firm acted as JV banker for ...

That single page is effectively 90% of the report. The remaining 10-20 pages deep-dive the evidence behind it.

Comparable Analysis (Comps)

Hypothetical semiconductor peer set

comps_data = {

"ticker": ["TSMC", "Samsung Electronics", "SK Hynix", "Intel", "Micron"],

"market_cap_usd_bn": [600, 380, 80, 200, 110],

"forward_pe": [22.0, 12.5, 9.8, 18.5, 11.2],

"forward_ev_ebitda": [12.5, 6.8, 5.2, 10.5, 6.5],

"rev_growth_yoy": [0.18, 0.08, 0.25, 0.05, 0.30],

"ebitda_margin": [0.55, 0.25, 0.40, 0.30, 0.35],

}

comps = pd.DataFrame(comps_data)

median_pe = comps["forward_pe"].median()

median_ev_ebitda = comps["forward_ev_ebitda"].median()

print(f"Median Forward P/E: {median_pe:.1f}x")

print(f"Median Forward EV/EBITDA: {median_ev_ebitda:.1f}x")

Premium/discount calc against your name

target_ticker = "TSMC"

target_pe = comps.loc[comps["ticker"] == target_ticker, "forward_pe"].values[0]

premium = (target_pe / median_pe - 1) * 100

print(f"{target_ticker} P/E premium to peers: {premium:+.1f}%")

This table is one of the report's core visuals. The next paragraph's job is to answer "why does the name deserve that premium?" You rarely show only one multiple — usually P/E, EV/EBITDA, P/B, and EV/Sales appear side by side and the body justifies which one matters most.

Certifications: CFA, KFIA, CMA, CIIA

| --- | --- | --- | --- | --- |

| CMA (US) | IMA | ~ `$1,000` | ~ 50% | US management accounting |

After joining a Korean sell-side house, picking up the KFIA Investment Analyst within the first year is essentially the default. CFA takes longer but is near-required if you ever move to a global IB. Even Level 1 helps an RA application.

English, Korean, and Japanese language ability

At a Korean sell-side house, English is fine at the "reads IR materials" level. But a global IB's Korea desk (e.g., Goldman Seoul) writes everything in English. You cannot translate Korean to English — you must write in English directly. TOEFL 100+ or IELTS 7.0+ is roughly the floor.

If you want to also cover the Japanese market, Japanese becomes a weapon. 80% of Japanese IR materials are Japanese-first; English translations lag by a week. JLPT N2 or higher is a strong edge for Japanese peer analysis. With the recent surge of interest in the Japanese market, Korean sell-side houses now pay a premium for Japanese-capable analysts.

20 interview questions

The usual battery for an RA interview.

- 1. Pitch one stock you like in five minutes.

- 2. How did you estimate that name's EPS?

- 3. If your estimate differs from consensus, why?

- 4. How did you set WACC? What is the risk-free rate?

- 5. How do you justify your multiple (P/E, EV/EBITDA)?

- 6. Name three scenarios where your call would be wrong.

- 7. Which US-market mover had the biggest swing yesterday?

- 8. Whiteboard a DCF.

- 9. Which industry do you like most, and why?

- 10. What was your worst call? What did you learn?

- 11. How do you model a company under margin pressure?

- 12. Walk through how FX flows through the P&L.

- 13. How does valuation differ between cyclical and growth industries?

- 14. What does it take to break consensus?

- 15. Which IB do you like most, and why?

- 16. Why are you pursuing CFA Level 1?

- 17. How does AI change the research role?

- 18. Buy-side or sell-side, and why?

- 19. Pick one published research report you love.

- 20. Where do you see your career in five years?

Roughly 70% of the questions probe stock analysis and modeling, the other 30% probe character and motivation. Question one shows up in nearly every interview, so be ready with a name or two — model included — to defend cold.

Compensation and bonus

| --- | --- | --- | --- |

Numbers above are base plus bonus. A Best Analyst (II number one) gets bonus a tier above. Jumping to buy-side keeps base similar but adds fund-performance bonus, which can multiply 2-3x the sell-side level in good years.

| Item | Sell-side | Buy-side |

| --- | --- | --- |

| Junior comp | KRW 60-80M (Korea) | KRW 70-90M (Korea) |

| Senior comp | KRW 200-500M | KRW 300M-1B (good year) |

| Bonus volatility | Medium | Very high |

| Bonus as percent of base | 30-150% | 50-500% |

Career path

A typical sell-side career path:

- Year 1-2: Research Associate (RA). Senior support, model updates, data cleaning.

- Year 3-5: Associate. Start your own coverage (2-4 names). Speak at morning meeting.

- Year 5-8: VP. Cover the core names in your sector. Critical during II vote season.

- Year 8-12: Director / Executive Director. Sector head, plus global macro linkage.

- Year 12+: MD / Senior Analyst. Aim for II number one under your name.

- Optional: Head of Research, Strategist, or buy-side jump to PM and ultimately CIO.

The buy-side path differs. Typically you spend 3-5 years sell-side, join the buy-side as an Associate Analyst, become a Senior Analyst in 5-7 years, then assist a Portfolio Manager, and eventually run your own fund as PM.

90-day study routine

A 12-week routine for a new RA or interview candidate.

| Week | Task | Deliverable |

| --- | --- | --- |

| 1 | Pick a sector, gather consensus data | One-page sector summary |

| 2 | Read top three companies' annual reports | One-page business model writeup |

| 3 | Drop five-year income statements into Excel | Base estimation model |

| 4 | Estimate revenue (price, volume, mix) | Estimation model v1 |

| 5 | Estimate margins (cost structure analysis) | Estimation model v2 |

| 6 | Build the DCF model | DCF sheet |

| 7 | Multiples comparable analysis | Comps table |

| 8 | Target price + sensitivity | One-page summary |

| 9 | Compare your call to consensus | Above/Below decision |

| 10 | Write a 3-page initiation note | First note v1 |

| 11 | Self-review vs a senior report | Note v2 |

| 12 | Mock pitch in five minutes | Pitch script |

Finish this routine and you walk into the interview as a candidate who built a model, made a recommendation, and can defend it with three reasons. That is roughly the top 10% of new-RA applicants.

Failure modes and collaboration flow

The three biggest traps in a research career:

- Wrong call + refuses to admit: holding a Buy for over a year, downgrading to Hold only after a 30% drop. Clients leave.

- Cognitive dissonance: massaging assumptions until your model matches consensus. The whole point of the report disappears.

- Consensus-chasing: in-line calls for a year. No votes at evaluation time.

- IR dependence: modeling exactly what the company IR said. Your view vanishes.

- Model overfit: 30 variables you cannot follow next quarter yourself. Simplicity is a virtue.

- Micro vs macro confusion: stuffing macro variables into single-name analysis until stock-picking ability evaporates.

Research looks like a solo task from the outside, but in practice it is collaborative.

- Morning meeting (07:30): pass commentary to sales and trading.

- IR meetings (ad hoc): lunch or call with the company's IR officer.

- Conference calls (earnings season): 30-60 minute call with IR + CFO, one or two questions per analyst.

- Marketing trip (semi-annual): tour clients (institutions) with your report. Korea → Hong Kong, Singapore, London, New York.

- Sales roadshow: do client rounds with the sales team so they can carry your call.

- Compliance: every target-price change needs pre-approval. Without it, nothing publishes.

Good seniors openly acknowledge wrong calls and write the next note explaining which assumption broke. That transparency builds long-term reputation.

Macro monitoring (Python automation)

from datetime import datetime, timedelta

def daily_macro_dashboard():

"""Macro dashboard you run at 7am each morning."""

tickers = {

"S&P 500": "^GSPC",

"Nasdaq": "^IXIC",

"KOSPI": "^KS11",

"Nikkei": "^N225",

"USD/KRW": "KRW=X",

"USD/JPY": "JPY=X",

"WTI Crude": "CL=F",

"Gold": "GC=F",

"10Y Treasury": "^TNX",

}

end = datetime.today()

start = end - timedelta(days=5)

rows = []

for name, ticker in tickers.items():

data = yf.download(ticker, start=start, end=end, progress=False)

if data.empty:

continue

last = data["Close"].iloc[-1]

prev = data["Close"].iloc[-2]

change_pct = (last / prev - 1) * 100

rows.append({"name": name, "last": last, "change_pct": change_pct})

df = pd.DataFrame(rows)

df["arrow"] = df["change_pct"].apply(lambda x: "UP" if x > 0 else "DOWN")

return df

Drop this on the monitor next to your desk before the morning meeting.

print(daily_macro_dashboard())

Running this at 07:00 daily means 30 minutes before the morning meeting you already see the macro variables that hit your sector. That is one of the senior-vs-junior gaps. Juniors listen at the morning meeting; seniors arrive with a view.

Closing

A research analyst is "the person who reads the company and writes the conclusion." Writing is just the format used to deliver that conclusion to the market. The model, consensus, industry hypothesis, sales collaboration, compliance, and II vote all move together. If you start a month before your interview and convert the 28 sections of this guide into your own deliverables one by one, you can walk in and articulate "my model, my call, my risks" without hesitation. Today, pick one stock and start writing the first page of a three-page initiation note — conclusion in one line, three reasons, three risks.

필사 모드: Securities Research Analyst — The Writing Job: Sell-side, Buy-side, Modeling, and Report Craft (2026 Complete Guide)

Who this is for

Why research is fundamentally a writing job

Sell-side vs Buy-side: two paths

A real day: 9am tape → 11am model update → 2pm conference call → 4pm note → 7pm review

Deliverables and modeling tool stack

Model types and sector applications

A simple DCF model (Python)

Hypothetical semiconductor company example

II Rankings, Greenwich, StarMine evaluation

Korean sell-side houses

Japanese sell-side houses

Global IB research houses

Buy-side research

Quant vs qualitative: the model is the tool, judgment is the substance

AI impact in 2026: BloombergGPT, AlphaSense, Tegus AI, Hebbia

A simple example of analyzing IR transcript tone with an AI-style helper

Writing the report and breaking consensus

A real good report (format and tone)

Comparable Analysis (Comps)

Hypothetical semiconductor peer set

Premium/discount calc against your name

Certifications: CFA, KFIA, CMA, CIIA

English, Korean, and Japanese language ability

20 interview questions

Compensation and bonus

Career path

90-day study routine

Failure modes and collaboration flow

Macro monitoring (Python automation)

Drop this on the monitor next to your desk before the morning meeting.

Recommended reading, resources, and References

Closing