Inve Blog
Quartr vs Inve: Transcripts vs Guidance Tracking
Quartr vs Inve, Koyfin, Fiscal.ai and Screener.in: compare single-call transcripts with 24,328 Indian and 37,529 US guidance rows, verdicts and MCP tools.
By Inve Research Desk · 11 November 2026
Published by Inve Research Desk
Edited & published by Aravind R · How this was written & verified →
Most investors comparing Quartr, Koyfin, Fiscal.ai, Screener.in, and Inve start by asking which platform has the cleanest transcript reader. That question hides the real choice. A transcript player or financial terminal is built to retrieve what management said today, whereas a multi-quarter guidance ledger is built to verify whether the numeric targets management set three quarters ago were met, revised, or quietly dropped.
That structural difference matters because single-call reading has a blind spot that faster audio playback and one-call AI summaries cannot fix. When an executive sets a margin range or a capacity deadline in Q1, repeats it in Q2, and stops mentioning it in Q4, nothing inside the Q4 transcript looks unusual. You cannot detect a silent withdrawal inside a single document because omission only becomes visible when earlier commitments sit next to later calls.
Inve's Indian production snapshot of 4 September 2026 contained 24,328 management commitments across 1,648 listed Indian companies, drawn from 8,220 parsed concall transcripts across 1,654 companies. Of the 16,842 commitments with a verdict other than new (the remaining 7,486 carried new and had no subsequent call on which to be assessed), 9,734 were kept (57.8% pooled across those 16,842 eligible commitments) and 2,271 were classified ghosted — 13.5% of those 16,842 eligible commitments, or 9.3% of all 24,328 tracked commitments. In the same snapshot, 1,000 of 1,648 companies (60.7% of those 1,648 companies) had at least one ghosted classification.
On the US side (snapshot of 5 September 2026), Inve parsed 6,132 US earnings calls across 3,157 US companies and tracked 37,529 forward commitments across 885 US companies. Even in that tightly scripted market, 35.4% of 1,587 tracked effective tax-rate commitments, 31.0% of 882 interest-expense and other-income commitments, and 23.8% of 951 capital-expenditure commitments were never reconciled aloud on a subsequent call, compared with 10.1% of 8,270 revenue guides, 9.2% of 2,556 EPS guides, and 9.1% of 3,411 margin guides.
The companies and platforms discussed below illustrate an equity research workflow and are not recommendations to buy, sell, or hold any security.
Why transcript players and guidance ledgers solve different problems
Think of a single-call transcript tool as a courtroom stenographer and a multi-quarter guidance ledger as an audit trail of prior affidavits. Both work from the same spoken words, but only the second one asks whether today's testimony matches what the witness signed last year. The companion guide on earnings call summaries versus guidance tracking examines that evidence boundary in detail.
What Quartr, Koyfin, Fiscal.ai, and Screener.in do well
Each major platform in a modern equity research stack is genuinely strong at its core job, and a sound workflow uses each where its evidence is strongest:
- Quartr (Audio Playback, Slide Decks, and Global IR Search): Quartr is purpose-built for live and recorded earnings-call audio, paragraph-synced transcripts, investor-day slide presentations, and mobile listening across global markets. When you want to hear the CFO's cadence during Q&A or inspect slide 14 alongside the spoken paragraph, Quartr handles document retrieval cleanly.
- Koyfin and TIKR (Global Financial Statements and Consensus Multiples): Koyfin's published data documentation covers global financial statements, valuation multiples, and Capital IQ consensus estimates for EPS, revenue, and EBITDA, while TIKR's published plan information includes multi-year financial tables, transcript history, and, on applicable plans, management guidance.
- Fiscal.ai (formerly FinChat) and AlphaSense (KPI Disaggregation and Enterprise Search): Fiscal.ai focuses on company-specific segment KPIs and conversational Q&A over filings, while AlphaSense serves institutional teams searching across broker research, expert calls, and regulatory filings.
- Screener.in, Trendlyne, and Tijori Finance (Indian Screening and Operating Disclosures): For Indian equities, Screener.in is the standard first-stage filter for 10-year ratios and custom query formulas, Trendlyne aggregates screening matrices and exchange links, and Tijori organises operating capacities and market-share disclosures.
Where document retrieval stops and commitment reconciliation begins
Every document-retrieval platform treats an earnings call primarily as a standalone file or media stream. Even when a terminal attaches a single-call AI summary or chat box to a transcript, that tool summarises what management chose to emphasise in that specific quarter.
Inve treats each call as an update to a longitudinal commitment register. When a new transcript is parsed, the system extracts forward-looking statements that name a metric, target, period, and speaker with a verbatim quote, and reconciles the company's earlier commitments against subsequent updates and reported outcomes.
| Research dimension | Quartr | Koyfin / TIKR | Fiscal.ai (FinChat) | Screener.in / Tijori | Inve (inve.money) |
|---|---|---|---|---|---|
| Primary unit of analysis | Single-call audio, slide deck, and transcript search | Multi-year financial statements and consensus estimates | Segment KPI tables and filing chat | 10-year Indian financial ratios and operating metrics | Quarter-by-quarter management guidance and Q&A ledger |
| Earnings call workflow | Synced audio player, slides, keyword search | Raw transcript archive inside a financial terminal | Single-call summaries and prompt chat | Exchange PDF links or single-call notes | Structured call brief, graded Q&A exchanges, and commitment history |
| Multi-quarter guidance tracking | Manual comparison across separate tabs | Consensus estimate revisions (plus management guidance on applicable TIKR plans), not call-to-outcome verdicts | Prompt-dependent retrieval | Manual comparison | Tracked commitment history (achieved, on_track, missed, revised_up, revised_down, ghosted, new) |
Unmentioned target (ghosted) flagging | No | No | No | No | Yes — flags prior call commitments with no parsed follow-up on later calls |
| Q&A response grading | No | No | No | No | Analyst Q&A exchanges graded for directness (plus recurring-dodge records on the India ledger) |
| Native MCP server for Claude & Gemini | Enterprise API (Quartr API) | No | No | No | Native MCP connector (https://api.inve.money/mcp) for Claude, Claude Code, and Gemini |
| Market coverage | Global (strong US and European IR depth) | Global | Global (strong US segment KPIs) | India | India (1,654 companies with parsed calls) and US (3,157 companies with parsed calls) |
What a cross-quarter ledger catches that single-call tools miss
The practical difference between reading one call and reading a multi-quarter commitment history shows up clearly in real Indian and US company sequences — including cases where the automated tracker itself needs a primary-source check.
1. A clean guidance walk — and a false positive that proves why source links matter (Infosys)
On Infosys, the FY26 constant-currency revenue-growth range moved in an orderly, verifiable sequence across five consecutive results calls:
| Actual results call | FY26 constant-currency revenue-growth range |
|---|---|
| Q4 FY25 (April 2025) | 0%–3% (initial guidance) |
| Q1 FY26 (July 2025) | 1%–3% |
| Q2 FY26 (October 2025) | 2%–3% |
| Q3 FY26 (January 2026) | 3%–3.5% |
| Q4 FY26 (April 2026) | Reported full-year constant-currency growth of 3.1% |
Reading those five dates together turns scattered quarterly commentary into a checkable annual cycle. Just as important, Infosys shows why an extracted status label must never be accepted blindly. In Inve's local guidance history, the row for FY26 free-cash-flow conversion (above 100% of net profit) was marked ghosted, with a note claiming the target was not explicitly addressed in Q4 remarks.
Checking the Q4 FY26 Infosys earnings-call transcript on printed page 4 disproves that ghosted label: CFO Jayesh Sanghrajka reported adjusted free cash flow of $3.5 billion for FY26 and adjusted FCF conversion of 106% of net profit (excluding specified Labour Code and income-tax-refund effects), alongside an adjusted operating margin of 21.0%. The FY26 annual-report highlights separately report FCF conversion of 112.3% of net profit on the report's stated basis. A ledger is valuable because it keeps the earlier target (>100% FCF conversion) in view and links you to the primary document, not because its model verdict is infallible.
2. When a consolidated beat coexists with unresolved plant and regional rows (Welspun Corp)
Welspun Corp shows why a strong headline quarter does not close every open question from the prior year. On its Q2 FY26 call, management guided FY26 consolidated EBITDA to ₹2,200 crore, against ₹1,186 crore reported for H1 and ₹1,831 crore for nine months in Q3 FY26. The Q4 FY26 transcript then reported full-year EBITDA around ₹2,370 crore and ROCE around 22.3% — a clear beat against the consolidated target.
In a single-call summary of Q4 FY26, that consolidated beat is the whole story. In the multi-quarter ledger, two narrower commitments from Q1 FY25 remain open beside it: a ductile-iron (DI) pipe capacity expansion target from 500,000 to 600,000 tonnes by the end of Q3 FY25, and a standalone US annual-volume target of 250,000–300,000 tonnes (Inve data, commitments 19452 and 19453). Both rows were classified ghosted in the parsed call history because subsequent combined India-and-US disclosures did not verify those standalone figures. Those classifications do not prove the plant missed its deadline; they tell the analyst which presentation tables and segment notes still need checking.
3. When the reporting perimeter changes between the target and the outcome (Centum Electronics)
Or consider Centum Electronics. On its Q2 FY25 call (14 November 2024), management discussed a consolidated EBITDA-margin objective of 14%–15% over two to three years, compared with an 8.3% consolidated EBITDA margin in its FY25 annual report (where Form AOC-1, printed page 47, records after-tax losses at two Canadian subsidiaries of ₹113 million and ₹10 million, or ₹11.3 crore and ₹1.0 crore). Across the next five calls, that margin objective went unmentioned in Q3 FY25, was recut to 11%–12% in Q4 FY25, restated at 13%–15% in Q1 FY26, and left unmentioned in Q2 and Q3 FY26.
Then the Q4 FY26 transcript placed the Canadian and European operations under discontinued operations and reported a FY26 consolidated EBITDA margin of 14.22% for the continuing business (while the extracted row recorded 14.16% and marked the target achieved). Without a cross-quarter record — and the statutory footnote check described in the four-layer Indian stock research stack — a reader sees a 14.22% margin clearing a 14%–15% target without noticing that loss-making subsidiaries left the denominator.
Across the 4 September 2026 Indian snapshot, ghosted classifications were sharpest in balance-sheet and cash-conversion categories: 71 of 279 eligible working-capital, debtor-day, and inventory-day commitments (25.4% of those 279, out of 372 tracked across 277 companies) and 147 of 750 eligible capital-expenditure commitments (19.6% of those 750, out of 1,041 tracked across 644 companies) had no parsed follow-up on later calls, against the 13.5% corpus baseline (2,271 of 16,842 eligible commitments). The same snapshot recorded 609 persistent-question dodges across 452 companies, defined as topics deflected on two or more consecutive calls (a medium-confidence, model-derived count).
4. What the US record shows: guidance floors and reporting conventions (Centene and S&P Global)
In the US market, the same multi-quarter lens separates genuine guidance resets from ordinary reporting conventions. On Centene, management used the February 2025 (Q4 2024) and April 2025 (Q1 2025) calls to reaffirm full-year 2025 adjusted diluted EPS at "greater than $7.25", then cut the full-year forecast on the July 2025 call to "approximately $1.75" — a drop of roughly 76% — after disclosing a Marketplace risk-adjustment shortfall and higher Medicaid medical costs, before raising the floor to "at least $2.00" in October 2025 and stepping FY26 guidance from "greater than $3.00" to "greater than $3.40" and "greater than $4.80" across February, April, and July 2026. Conversely, S&P Global guided roughly $5.2 billion of full-year adjusted free cash flow on a 2024 call and never reconciled that figure aloud on a subsequent call because US companies routinely close out cash-flow, tax-rate (35.4% of 1,587 tracked), and interest-expense (31.0% of 882 tracked) guides in 10-Q or 10-K tables rather than spoken remarks. The US guide to tracking management guidance across quarters walks through both US cases in full.
Why pasting transcripts into ChatGPT or Claude fails without a structured MCP layer
A common question from investors who already use Quartr or Screener.in is simple: "Why not download four transcript PDFs, paste them into ChatGPT, Claude, or Gemini, and ask the model to build a guidance table?"
The US guide to AI workflows for earnings call research documents what actually broke when we tested unguarded LLM runs at scale:
- Entity and ticker collisions upstream of the model: In an early extraction run without hard identity verification, an analysis of US medical-device maker Stryker (
SYK) ingested transcripts belonging to Straker Limited — a New Zealand translation company with a similar ticker string — and produced an internally consistent, completely invalid summary. The same collision occurred on Altria (MO), where Altri, a Portuguese pulp producer, landed in the file. - Missing quarters and filing close-outs misread as evasion: Roughly a third of the company datasets assembled for the US record had at least one quarter missing, concentrated between April and September 2025, and some companies (such as Mondelez shifting from 46% prepared remarks to 4% in July 2025) publish prepared remarks separately from Q&A. A naive prompt treats a missing file or a 10-Q table close-out as management going silent.
- Unfalsifiable prompts and inferred outcomes: In one early run, a model marked a $100 million target as substantially achieved even though the matched figure appeared nowhere in six transcripts of that company. Without a hard rule that every commitment and every resolution must carry a verbatim quote, a language model will present a plausible inference as a verified fact.
How Inve connects to Claude, Claude Code, and Gemini via MCP
Instead of pasting raw PDFs into a chat window every quarter, Inve exposes its identity-checked database directly to AI assistants through the Model Context Protocol at https://api.inve.money/mcp. Step-by-step connection instructions live in the Inve for Claude setup guide and the Inve for Gemini setup guide.
| MCP tool name | What it reads from Inve | Cost on Inve plan |
|---|---|---|
search_companies | Finds a company in India or the US by exchange-verified ticker or name | Free |
get_company_financials | Structured balance sheet, annual and quarterly income statements, cash flows, ratios (ROE, ROCE, working capital days), valuation panel, and segments | Free |
list_calls | Complete inventory of a company's parsed earnings-call dates and periods in the archive | Free |
get_call_summary | Stored analysis of one earnings call: executive summary, guidance table with speaker and quote, graded Q&A, and risks (guest preview or full read) | Free preview / 1 call credit |
get_guidance | Multi-quarter guidance history from the Guidance Tracker (guest public brief or full signed-in ledger) | Free brief / 1 guidance credit |
get_usage | Monthly credit allowances, remaining balance, reset date, and connected-app usage | Free |
Because the assistant calls search_companies and list_calls first, it never confuses Stryker with Straker or mistakes an unindexed quarter for silence. Each connected account belongs to either the India or the US market based on the phone number used at sign-up, and the tools read only stored research in Inve without browsing the open web or receiving your personal profile details.
Keep the earlier target in view alongside today's call
Read quarterly concall summaries alongside the original guidance target, speaker quote, and subsequent updates before you act on a headline.
Browse concall summariesHow Inve fits alongside your existing research stack
You do not need to cancel Quartr, Koyfin, or Screener.in to use Inve. Each tool handles a distinct stage of the work:
- Screen and verify financial history: Build your initial shortlist in the Indian stock screener (which filters roughly 5,335 Indian companies across 45 FY and TTM metrics from filed statements) or in Screener.in and Koyfin. Then apply the cash-conversion and borrowing checks described in what to check after running a stock screener.
- Compare sector-wide earnings context: Before blaming or crediting management for a margin move, check Indian sector earnings analysis to see whether the top 20 companies in that sector expanded or contracted earnings in the same quarter.
- Read the current call and audit prior guidance: Use Quartr when you want live audio or slide decks on the move. Use quarterly concall summaries for the structured call brief, extracted targets, and graded Q&A exchanges, and open the management guidance ledger to inspect the commitment history and recurring-question record across quarters. Detailed Indian company ledgers require Pro access; on the US side, call briefs and company pages are public, while US Pro (available to accounts registered with a non-
+91phone number) includes one free month before paid billing. - Verify primary filings and test valuation: Always click through from any disputed ledger row to the exchange filing or annual report, and test what growth is priced in with the DCF calculator. Investors comparing domestic and global names can pair that routine with the framework for researching Indian midcaps alongside US stocks.
Inve is describing its own platform alongside adjacent products here, so our incentive to highlight cross-quarter guidance tracking is obvious. The Infosys and Centum examples above show why our own extracted rows still require primary-source verification.
Where a guidance ledger can mislead you
The strongest argument against leaning too heavily on a guidance ledger is that forecasting precision is not the same thing as business quality or capital-allocation skill. A cautious management team in a mature business can set a low bar every April, step over it every quarter, and accumulate a spotless delivery record while under-investing in its competitive moat. Conversely, a candid operator navigating a customer delay or regulatory shock may cut guidance openly and record a missed or revised_down verdict while making sound five-year decisions.
Three data boundaries also govern every statistic in this article:
- Uneven, short time depth: The 4 September 2026 Indian snapshot averaged 4.18 distinct commitment birth quarters per company, and the US archive spans roughly two years of calls. Four or five quarters can reconstruct one annual guidance cycle, not how a management team behaves across a full credit or industrial cycle.
- Call transcripts are not the entire disclosure universe: A
ghostedclassification means the parser did not identify a follow-up on a subsequent earnings call in the corpus. An outcome reported in an annual report, 10-K, 10-Q, investor presentation, or press-release table can remain invisible to a call-only parser, and the Infosys FY26 cash-conversion example shows that the parser can also miss a follow-up spoken on the call itself. - Two model generations without a common holdout: The Indian commitment corpus was extracted across
gemini-2.5-pro(1,469 company runs in the working-paper reconciliation; an earlier note recorded 1,476) andgemini-3.1-pro-preview(179 company runs) without a shared holdout benchmark, and Q&A deflection streaks are model-derived classifications at medium confidence. Category counts (working capital,capex) are matched withILIKEon free-text metric fields and represent fuzzy lower bounds.
A transcript player helps you hear what management wants to talk about today. Before you size a five-year holding, check what happened to the targets management stopped talking about.
Frequently asked questions
How is Inve different from Quartr?
Quartr is built around live and recorded earnings-call audio, slide presentations, and global IR document search. Inve is built around structured post-call analysis and cross-quarter guidance reconciliation: it extracts dated management commitments with speaker quotes, grades analyst Q&A exchanges, and tracks whether prior targets were met, revised, or left unmentioned on later calls.
Can I use Inve alongside Koyfin, TIKR, or Screener.in?
Yes. Financial terminals and screeners such as Screener.in, Koyfin, and TIKR show multi-year financial statements, ratios, and consensus estimates (plus management guidance tables on applicable TIKR plans). Inve adds the multi-quarter commitment-to-outcome ledger and graded Q&A record so you can test whether the operational assumptions behind those financial tables still hold.
Does a `ghosted` label prove management hid a bad result?
No. In Inve's dataset, ghosted means the parser did not find a follow-up on a subsequent parsed earnings call (2,271 of 16,842 eligible Indian commitments, or 13.5%, as of 4 September 2026). The company may have closed the number in a press-release table or annual report, or the parser may have missed the follow-up, as the Infosys FY26 free-cash-flow conversion example demonstrates.
How does Inve's MCP server work with Claude and Gemini?
Inve hosts a Model Context Protocol server at https://api.inve.money/mcp for Claude, Claude Code, and Google Gemini. Once connected through your Inve account, the assistant can call search_companies, get_company_financials, list_calls, get_call_summary, get_guidance, and get_usage directly, avoiding ticker collisions and unverified web-search summaries.
Are Inve's Indian and US guidance ledgers free?
On the US side, company pages and earnings-call briefs are public, while full US Guidance Tracker history is part of US Pro (one free month for eligible non-+91 accounts, then paid). On the India side, public pages and the AI stock screener require no login, while detailed Indian company guidance ledgers require Pro access. Connected MCP reads of get_call_summary and get_guidance return a free preview/brief in guest mode or use one monthly call or guidance credit when signed in.
Inve is a research and analysis platform, not an investment adviser. Nothing here is a recommendation to buy or sell any security. Do your own research or consult a SEBI-registered adviser before investing.