earnings-analysis

name

earnings-analysis

description

Drishti MCP workflow for high-quality Indian listed company earnings deep dives. Use when the user asks for an earnings deep dive, quarterly results review, latest results analysis, YoY/QoQ financial performance breakdown, margin analysis, revenue/EBITDA/PAT table, segment performance, concall-backed interpretation, post-result risks, or a plain-English explanation of reported earnings. Produces sourced, compact, decision-useful analysis using Drishti earnings filings, announcements, concalls, news, events, price/volume data, and company metadata. No buy/sell/hold recommendations.

Indian Equity Earnings Deep Dive

Create professional EARNINGS DEEP DIVE RESPONSES for Indian listed companies using Drishti MCP as the primary evidence source.

This skill is for agent responses, not long-form DOCX reports. The output should feel like a concise institutional research note: factual, sourced, easy to scan, and useful to an investor who wants to understand what changed in the latest reported quarter.

Key Characteristics

• Length: 800-1,800 words by default; shorter if the user asks for a brief summary.
• Tables: 2-5 compact markdown tables.
• Charts: Not required unless the user asks; use tables and sourced commentary first.
• Audience: Indian equity investors, analysts, or agents needing reliable earnings interpretation.
• Focus: Reported quarter, YoY/QoQ performance, drivers, management commentary, risks, and what to monitor next.
• Tone: Plain English, analytical, and source-grounded.
• Recommendation Policy: Do not provide buy, sell, hold, rating, target price, or personalized investment advice.

When To Use

Use when the user requests:

• "Do an earnings deep dive for [Company]"
• "Analyze [Company] Q1 FY27 results"
• "Explain [Company]'s latest quarterly results"
• "Give me revenue, EBITDA, PAT, margins, YoY and QoQ"
• "Summarize the concall after results"
• "What drove the result and what are the risks?"
• "How did the market react after earnings?"

Do not use if:

• The request is for a generic company profile with no earnings angle.
• The user asks for upcoming results only; use an upcoming earnings calendar workflow instead.
• The user asks for a full published research report, valuation model, rating, or target price.

Critical Requirements

1. Drishti MCP First

Use Drishti MCP as the system of record for Indian listed company facts.

The expected MCP namespace is mcp__Drishti. If Drishti tools are not already visible in the active tool list, discover them with tool search using a query such as Drishti MCP earnings filings concalls announcements, then proceed with the returned Drishti tools. Do not substitute non-Drishti data for Indian market facts unless Drishti is unavailable and the limitation is disclosed.

Before using Drishti data tools in a new session, call:

read_me(mode="data")

Then use Drishti tools for:

• Symbol resolution and metadata
• Earnings filings and structured result tables
• Announcements and exchange filings
• Concall transcripts, summaries, Q&A themes, and upcoming calls
• Result-date news and market context
• Price and volume reaction when relevant
• Corporate calendar events

Use web search only when Drishti cannot answer a required external context question. Label external-web evidence separately from Drishti evidence.

2. No Invented Data

Never invent:

• Revenue, EBITDA, PAT, EPS, margins, AUM, NIM, GNPA/NNPA, order book, volume, realizations, segment numbers, guidance, dates, or management commentary.
• YoY/QoQ growth rates when the current and comparison values are not available.
• Concall commentary when no concall transcript or summary was retrieved.

If a metric is missing, write not available in retrieved Drishti data.

3. Latest Quarter Discipline

If the user does not specify a quarter:

1. Use list_earnings for the company.
2. Identify the latest available reported quarter from returned Drishti records.
3. State the quarter actually analyzed and the filing/report date.

If the user specifies a quarter:

1. Use that quarter in list_earnings or get_earnings_filing.
2. If missing, show the nearest available quarters from list_earnings.
3. Do not silently switch quarters.

4. Separate Facts From Interpretation

Every important claim should be classed as one of:

• Reported fact: Directly from earnings filing, announcement, or structured Drishti table.
• Management commentary: From concall or transcript/search result.
• Market context: From Drishti news, events, or price/volume tools.
• Analyst interpretation: Your synthesis from the above evidence.

5. Source Attribution

Every table and major paragraph must name the source type and date when available. Use dated source notes such as:

Source: Drishti earnings filing for Q1 FY27, filed/reported 2026-07-22.
Source: Drishti concall search, Q1 FY27 transcript/summaries.
Source: Drishti announcements, 30-day window around result date.

When exact dates are not returned, say so rather than fabricating them.

Drishti MCP Tool Usage

Use this exact tool order unless the user request clearly requires a smaller subset.

Step 0: Tool Routing Read

• read_me(mode="data"): call once at the start of a Drishti MCP analysis session when available.
• get_workflow_prompt(workflow="earnings_concall", company_or_symbol="SYMBOL", quarter="qN_YY"): optional helper when the agent needs Drishti's own task-specific workflow guidance before collecting data.

Step 1: Resolve Company

• resolve_symbols(queries=[company_or_phrase]): use for ambiguous company names, short names, brands, or informal names.
• get_symbols(query={"symbols": "SYMBOL"}): use when the user gives a clear symbol.
• get_sectoral_data(symbols="SYMBOL"): fetch sector/industry/company metadata when useful for context.

If multiple plausible symbols remain after resolution, ask the user to choose. Do not proceed with a guessed symbol when ambiguity is material.

Step 2: Find The Earnings Record

• list_earnings(query={"symbols": "SYMBOL", "limit": 8, "detailed": true}): find available quarters and select the requested/latest quarter.
• get_earnings_filing(symbol="SYMBOL", quarter="qN_YY", detailed=true): fetch the detailed filing for the selected quarter.
• If list_earnings returned a specific id, use get_earnings_filing(earnings_id="...") when that is more precise.

Prefer Drishti's earnings_table and extracted fields over manually parsing free text. Use free-text summaries only to explain drivers and context.

Step 3: Add Filing And Event Context

• get_announcements(query={"symbols": "SYMBOL", "lookback_days": 60, "detailed": true}): result announcement, financial results filing, investor presentation, dividend/corporate-action context.
• search_announcements(search_query="financial results investor presentation dividend board meeting", symbols="SYMBOL", days=90): use when the direct announcement list is sparse or when looking for a specific filing type.
• get_events(symbols="SYMBOL", days_count=90): upcoming board meetings, AGM, or event context where relevant.

Step 4: Add Concall Evidence

• search_concalls(search_query="quarter results revenue margin demand guidance outlook risks", symbols="SYMBOL", quarter_selector="LAST", result_mode="chunks", per_page=8): use for latest-quarter commentary when quarter is not specified.
• search_concalls(..., specific_quarters="qN_YY"): use for a specified quarter when the quarter code is known.
• get_concalls(query={"symbols": "SYMBOL", "limit": 4, "detailed": true}): retrieve available concall records and dates.
• get_upcoming_concalls(query={"symbols": "SYMBOL", "detailed": true}): use if results are out but the concall has not happened yet.

Do not claim management said something unless it is in retrieved concall data. Summarize commentary; avoid long quotations.

Step 5: Add News And Market Reaction

• get_news(query={"symbols": "SYMBOL", "lookback_days": 30, "limit": 10}): result coverage and sector context.
• search_news(queries=["SYMBOL results", "Company quarterly results"], symbols="SYMBOL", days=30, per_query_limit=5): use for targeted result news.
• get_price_and_volume(symbols="SYMBOL"): current market data only.
• get_price_and_volume_since(symbols="SYMBOL", days_count=7, series=true): post-result price/volume reaction when requested or useful.

Market reaction is optional. Include it only when the user asks or when it helps explain the result's immediate impact.

Analysis Workflow

The earnings deep dive process follows five phases.

Phase 1: Identify Scope

Determine:

• Company name and symbol.
• Requested quarter, or latest available quarter.
• Whether the user wants brief summary, detailed deep dive, concall takeaways, market reaction, segment analysis, peer context, or source pack.
• Sector-specific metrics that matter.

Common Indian fiscal-quarter notation:

• Q1 FY27 usually maps to q1_27.
• Q2 FY27 usually maps to q2_27.
• Q3 FY27 usually maps to q3_27.
• Q4 FY27 usually maps to q4_27.

Always confirm the actual quarter returned by Drishti.

Phase 2: Collect Evidence

Required minimum evidence:

• Company metadata from get_symbols and/or get_sectoral_data.
• Available earnings records from list_earnings.
• Target quarter details from get_earnings_filing.

Preferred additional evidence:

• Result-related announcements from get_announcements or search_announcements.
• Concall evidence from search_concalls or get_concalls.
• News from get_news or search_news.
• Events/upcoming concalls when concall data is missing.

If Drishti has no earnings filing for the requested quarter, stop and provide:

• What was requested.
• What Drishti returned.
• Nearest available quarters.
• What data would be needed to proceed.

Phase 3: Build Financial Tables

Use actual metric labels from Drishti where possible. Do not force a standard industrial template onto banks, NBFCs, insurers, commodity companies, or platform businesses.

Default financial table:

| Metric | Current Quarter | YoY | QoQ | Interpretation |
|---|---:|---:|---:|---|
| Revenue | Rs X cr | +Y% | +Z% | Short explanation |
| EBITDA | Rs X cr | +Y% | +Z% | Short explanation |
| EBITDA Margin | X% | +Y bps | -Z bps | Short explanation |
| PAT | Rs X cr | +Y% | +Z% | Short explanation |

For financials:

| Metric | Current Quarter | Change | Interpretation |
|---|---:|---:|---|
| Net Interest Income | Rs X cr | +Y% YoY | Short explanation |
| NIM | X% | +Y bps YoY | Margin implication |
| GNPA | X% | -Y bps QoQ | Asset quality implication |
| PAT | Rs X cr | +Y% YoY | Profitability implication |

For sector-specific results:

| Metric/Segment | Current Quarter | Change | What It Means |
|---|---:|---:|---|
| Order Book / AUM / Volume / Occupancy | X | +Y% | Business driver |

If YoY or QoQ is absent, write NA. Calculate only when all values needed for the calculation are retrieved and clearly comparable.

Phase 4: Explain Drivers

Answer these questions:

• What was the headline result?
• What improved versus last year and last quarter?
• What worsened?
• Was growth led by volume, pricing, mix, utilization, recoveries, treasury gains, credit costs, commodity prices, or one-offs?
• Did margins expand or contract, and why?
• Did segment performance diverge?
• Did management commentary confirm, soften, or contradict the reported numbers?
• What are the next-quarter watch items?

Use confidence labels:

• High: directly supported by earnings filing, structured table, or concall.
• Medium: supported by related announcement/news, but not fully quantified.
• Low: reasonable inference from evidence; mark clearly as interpretation.

Phase 5: Produce The Response

Default structure:

**Earnings Deep Dive: Company Name (SYMBOL)**
Quarter analyzed: QN FYYY
Evidence base: Drishti earnings filing, announcements, concall/news if available

**Bottom Line**
- 3-5 bullets with the result verdict, main driver, margin/profit quality, and
  key risk/watch item.

**Key Financials**
| Metric | Current Quarter | YoY | QoQ | Interpretation |
|---|---:|---:|---:|---|

Source: Drishti earnings filing for [quarter], [date if available].

**What Drove The Quarter**
| Driver | Evidence | Interpretation | Confidence |
|---|---|---|---|

**Management Commentary**
| Theme | What Management Indicated | Why It Matters |
|---|---|---|

**Risks And Watch Items**
| Item | Evidence | What To Monitor |
|---|---|---|

**Source Notes And Caveats**
- Dated Drishti sources used.
- Missing data caveats.
- Any external-web source used, clearly labeled.

If the user asks for a brief answer, return only:

• Bottom Line
• Key Financials
• Risks And Watch Items
• Source Notes

If the user asks for concall-only analysis, focus on:

• Management commentary themes
• Demand/pricing/margin outlook
• Capex, balance sheet, guidance, and risks
• Q&A concerns
• Explicit caveat if transcript/search coverage is incomplete

If the user asks for market reaction, add:

**Market Reaction**
| Period | Price Move | Volume Context | Likely Driver |
|---|---:|---|---|

Use get_price_and_volume_since for the table and label "likely driver" as interpretation unless directly supported by news.

Response Quality Bar

A strong answer should:

• Lead with the earnings verdict, not background.
• Quantify the key moves where Drishti provides numbers.
• Use the company's sector-specific metrics.
• Explain both growth and margin/profit quality.
• Include management commentary when available.
• Separate one-offs from recurring drivers.
• Identify next-quarter watch items.
• Cite Drishti source types and dates.
• Clearly state missing data instead of filling gaps.
• Avoid recommendation language.

Common Mistakes To Avoid

• Using training-memory numbers instead of Drishti MCP data.
• Saying "latest quarter" without naming the quarter and date.
• Mixing quarters from filings, concalls, and news without checking dates.
• Calculating YoY/QoQ from incompatible standalone vs consolidated numbers.
• Treating management commentary as fact without sourcing concall data.
• Overloading the response with long company background.
• Providing buy/sell/hold advice or target prices.
• Calling missing data "flat", "weak", or "strong" without evidence.

Verification Checklist

Before final response, verify:

• read_me(mode="data") was called when Drishti tools were available.
• Symbol was resolved or confirmed.
• Quarter analyzed is explicitly stated.
• Earnings filing was retrieved with get_earnings_filing or missing data was disclosed.
• Tables use actual retrieved metrics and labels.
• YoY/QoQ values are sourced or clearly calculated from retrieved values.
• Concall claims are backed by search_concalls or get_concalls.
• Announcement/news/event context is dated where available.
• Missing data is called out plainly.
• No recommendation, rating, target price, or personalized advice is given.

Dependencies

Required:

• Drishti MCP tools for Indian equity data.

Optional:

• Spreadsheet or charting tools if the user asks for exportable tables, model updates, or charts.
• Web search only for external context unavailable in Drishti; label separately.