Skip to main content

​
IB Check Deck Skill

​
What is Deck QC?

In investment banking, presentations (called “decks” or “pitchbooks”) are the primary deliverable for client-facing work. A typical pitch deck might contain 30-50 slides with revenue figures, valuation multiples, market data, charts, and narrative text. These decks go through multiple rounds of editing by analysts, associates, VPs, and MDs — each adding their own changes, often under tight deadlines. Deck QC (quality control) is the systematic process of checking a finished presentation for errors before it reaches the client. This is not proofreading in the traditional sense — it is a forensic review that catches the kinds of errors that slip through when multiple people edit a deck over several late nights:
  • Number inconsistencies — Revenue is shown as 485Montheexecutivesummaryslidebut485M on the executive summary slide but 500M on the financial detail slide. Both numbers were correct at different points in the drafting process, but one did not get updated.
  • Data-narrative misalignment — The text says “margins have expanded significantly” but the chart on the same slide shows margins declining. Someone edited the chart with new data but forgot to update the text.
  • Language issues — Casual phrasing like “pretty good performance” in a client-facing IB deck. Or the same concept referred to as “LBO model” on one slide and “leveraged buyout model” on another.
  • Formatting drift — Revenue shown as “500M"ononeslideand"500M" on one slide and "500MM” on another. Or dates formatted as “Jan 2024” on one slide and “January 2024” on another.
Worked example: Consider a real scenario. A 35-slide sell-side pitch deck for a healthcare company goes through five rounds of edits: 
Round 1: Analyst builds the deck. Revenue = $485M (from Q3 10-Q, annualized).
Round 2: VP updates the financial summary with Q4 data. Revenue = $512M (full-year actual).
Round 3: MD edits the executive summary narrative but uses the old $485M number from memory.
Round 4: Associate updates the comps table with new peer data.
Round 5: Analyst re-formats charts but does not re-read the narrative text.

Result: Slide 3 (Exec Summary) says "$485M revenue"
        Slide 8 (Financial Detail) says "$512M revenue"
        Slide 15 (Comps Table footnote) says "$485.0 million"
Three different slides show three different versions of the same number. Without systematic QC, this goes to the client.

​
Detailed Worked Example

Let us walk through a complete QC review of a fictional 20-slide M&A pitch deck for “TechTarget Inc.”
1

Extract All Content

Read every slide and extract all text, numbers, chart data, and footnotes with slide-level attribution. For a 20-slide deck, this produces a structured content map:
Slide 1 (Cover): "Project Alpha - Preliminary Discussion Materials"
Slide 2 (Disclaimer): Standard confidentiality notice
Slide 3 (Exec Summary): "TechTarget revenue of $485M..."
                        "EBITDA of $120M (24.7% margin)..."
                        "Trading at 12.5x LTM EBITDA..."
Slide 5 (Financial Overview): Revenue chart showing $485M, $512M, $540M
                               EBITDA chart showing $120M, $135M, $148M
Slide 8 (Comps Table): TechTarget EV/EBITDA = 13.2x
Slide 12 (Valuation Summary): DCF implies $38-$45 per share
                               Comps imply $35-$42 per share
...
2

Number Consistency Check

Normalize all numbers and look for conflicts:
MetricSlide 3Slide 5Slide 8Slide 12Consistent?
LTM Revenue$485M$512M (chart Year 2)——NO — Slide 3 uses old number
LTM EBITDA$120M$135M (chart Year 2)——NO — Same issue
EBITDA Margin24.7%———Check: 120M/120M / 485M = 24.7%. But if revenue is $512M, margin would be different
EV/EBITDA12.5x (Slide 3)—13.2x—NO — Two different multiples
DCF range———38−38-45Need to verify against DCF model
Comps range———35−35-42Need to verify against comps output
Findings:
  • CRITICAL: Revenue appears as 485M(Slide3)and485M (Slide 3) and 512M (Slide 5). The financial detail slide was updated but the executive summary was not.
  • CRITICAL: EBITDA appears as 120M(Slide3)and120M (Slide 3) and 135M (Slide 5). Same root cause.
  • CRITICAL: EV/EBITDA shows 12.5x on Slide 3 and 13.2x on Slide 8. These cannot both be correct for the same period.
  • IMPORTANT: EBITDA margin of 24.7% on Slide 3 was calculated using 120M/120M/485M. If the correct figures are 135M/135M/512M, the margin is 26.4%.
3

Data-Narrative Alignment

Check every claim against supporting data:
Claim (Slide)Supporting DataAligned?
”Strong revenue growth trajectory” (Slide 3)Revenue chart shows 420M>420M > 485M > $512M (Slide 5)Yes — growing
”Margins have expanded significantly” (Slide 3)EBITDA margin 22.3% > 24.7% > 26.4% (if using updated numbers)Yes — but the margin on the same slide is wrong
”#2 player in the enterprise market” (Slide 6)Market share chart shows CompA 35%, TechTarget 18%, CompB 15%Yes — #2 by share
”Trading at a discount to peers” (Slide 3)Peer median EV/EBITDA = 14.5x, TechTarget = 13.2x (Slide 8)Inconsistent — Slide 3 says 12.5x which would be a larger discount. Need to resolve which multiple is correct.
Finding:
  • IMPORTANT: The “discount to peers” claim is directionally correct but the magnitude depends on which multiple is right (12.5x vs 13.2x). At 12.5x, the discount is 13.8%. At 13.2x, the discount is 9.0%. Both support the claim but the narrative should use the correct number.
4

Language Polish

Scan for IB register violations:
SlideIssueCurrent TextSuggested Fix
3Vague quantifier”significant growth""26% revenue CAGR over 3 years”
6Casual phrasing”pretty strong market position""second-largest market participant with 18% share”
9Contraction”doesn’t compete directly""does not compete directly”
14Exclamation point”Highly attractive opportunity!""Highly attractive opportunity.”
7Inconsistent terminology”LBO model” (Slide 7) vs “leveraged buyout analysis” (Slide 12)Pick one and use it throughout
11Vague quantifier”a lot of synergy potential""$25-35M of identified cost synergies”
5

Visual and Formatting QC

SlideIssueSeverity
5Chart missing source citationImportant
5Chart Y-axis label missing (“$M”)Important
8Comps table uses “MM"butSlide3uses"MM" but Slide 3 uses "M”Minor
10Date format “Jan 2024” but Slide 14 uses “January 2024”Minor
12Font size 10pt in footnote (below firm minimum of 12pt)Minor
15Missing “Source:” attribution on market dataImportant
6

Final Report

QC FINDINGS -- Project Alpha Deck
Date: [Current Date]
Reviewed by: [Analyst Name]

CRITICAL (3 findings -- must fix before client delivery):
1. Revenue inconsistency: $485M (Slide 3) vs $512M (Slide 5)
2. EBITDA inconsistency: $120M (Slide 3) vs $135M (Slide 5)
3. EV/EBITDA inconsistency: 12.5x (Slide 3) vs 13.2x (Slide 8)

IMPORTANT (5 findings -- should fix):
4. EBITDA margin calculated on stale revenue (Slide 3)
5. Six language issues (vague quantifiers, contractions, inconsistent terms)
6. Missing chart source citations (Slides 5, 15)
7. Missing Y-axis label (Slide 5)
8. Unit style inconsistency ($M vs $MM)

MINOR (3 findings -- polish):
9. Date format inconsistency (Jan vs January)
10. Font size below minimum in footnote (Slide 12)
11. Inconsistent terminology (LBO model vs leveraged buyout analysis)

No critical issues found in: Calculations, market share data,
valuation ranges, disclaimer presence.

​
Why It Matters

In investment banking, a single wrong number can have serious consequences:
  • Credibility damage — If a client spots an inconsistency, they question all the analysis. Trust, once lost, is extremely hard to rebuild.
  • Deal impact — Incorrect financial data in a CIM or management presentation can lead to mispriced transactions, deal breakdowns, or post-closing disputes.
  • Regulatory risk — In SEC filings and fairness opinions, data errors can have legal implications.
  • Career impact — Analysts and associates are expected to catch these errors. Missing them reflects poorly and can affect reviews and promotions.
Junior bankers typically spend hours doing manual QC — cross-referencing every number on every slide, checking footnotes, validating calculations. This skill automates that process across four dimensions simultaneously.

​
Key Concepts

TermDefinitionWhy It Matters
Number ConsistencyThe same metric should show the same value on every slide where it appears.The most common QC failure. Revenue cannot be 485Monslide3and485M on slide 3 and 500M on slide 15.
Data-Narrative AlignmentClaims in the text must be supported by the data on the same or nearby slides.Someone edits the chart but forgets the text — or vice versa. This is how decks “lie.”
IB RegisterThe professional, formal tone expected in investment banking documents. No contractions, no casual phrasing, no vague quantifiers.”Very strong performance” is not IB quality. “Revenue grew 23% YoY to $485M” is.
Unit NormalizationConverting different representations of the same number to a common format (500M,500M, 500MM, $500 million = same number).Simple find-replace misses variants. QC must normalize units to catch all instances.
Near-Miss NumbersTwo values that are close but not identical (e.g., 485Mand485M and 486M), which may represent different periods or rounding of the same figure.Harder to catch than obvious mismatches. Require judgment: is this a rounding difference or a genuine error?
Derived NumbersValues calculated from primary numbers (growth rates, margins, market share percentages).When primary numbers change, derived numbers become stale. QC must flag these second-order effects.

​
How It Works

Triggers when: the user asks to review, check, QC, proof, or do a final pass on a deck, pitch, or client materials.
This is read-and-report only — Claude does not edit the deck. Review the findings first, then request specific changes.

​
Step 1: Read the Deck

Pull text from every slide with slide-level attribution. For each finding, Claude identifies which slide(s) are involved. For decks with 30+ slides, write the extracted text to a structured file so the analysis can process it systematically.

​
Step 2: Number Consistency

Claude normalizes unit variants (500Mvs500M vs 500MM vs $500,000,000), categorizes values (revenue, EBITDA, multiples, margins), and flags when the same metric shows conflicting values on different slides. Also checks:
  • Calculations are correct (totals sum, percentages add up, growth rates match endpoints)
  • Unit style is consistent throughout (pick one of MorM or MM)
  • Time periods are aligned (FY vs LTM vs quarterly, explicitly labeled)

​
Step 3: Data-Narrative Alignment

Maps claims to supporting data:
  • Trend statements (“declining margins”) — does the chart actually go that direction?
  • Market position claims (“#1 player”) — do revenue and share data support it?
  • Plausibility — “#1 in a 100Bmarket"with100B market" with 200M revenue implies 0.2% share; that is not #1

​
Step 4: Language Polish

Scans for anything that breaks the IB register:
  • Casual phrasing (“pretty good”, “a lot of”, “very significant”)
  • Contractions (shouldn’t, can’t)
  • Exclamation points
  • Vague quantifiers without numbers (“significant growth”)
  • Inconsistent terminology across slides

​
Step 5: Visual and Formatting QC

  • Missing chart source citations
  • Missing axis labels
  • Typography inconsistencies (font size, weight, color)
  • Number formatting drift (1,000 vs 1K)
  • Date format drift (Jan 2024 vs January 2024)
  • Missing footnotes or disclaimers

​
Output Format

Findings are categorized by severity:
SeverityDescription
CriticalNumber mismatches, factual errors, data contradicting narrative — these block client delivery
ImportantLanguage issues, missing sources, terminology drift — should fix before sending
MinorFont sizes, spacing, date formats — polish
If no critical issues are found, Claude explicitly states “no number inconsistencies found” — that is a finding, not an absence of one.

​
Common Mistakes

The mistake: Reviewing only the 5 slides you changed instead of the entire deck.Why it happens: Time pressure. You edited slides 8-12 and assume slides 1-7 and 13-30 are fine because someone else already reviewed them.The fix: Always QC the entire deck, not just your changes. Your edits may have created inconsistencies with slides you did not touch. The executive summary (slide 2-3) is especially vulnerable because it summarizes data from later slides and is often written first and updated last.
The mistake: Seeing “485Mrevenue"and"485M revenue" and "120M EBITDA (24.7% margin)” and not verifying that 120M/120M / 485M actually equals 24.7%.Why it happens: The margin looks reasonable and the numbers appear internally consistent.The fix: Verify every calculation. 120M/120M / 485M = 24.74%, which rounds to 24.7% — correct. But if someone updated revenue to 512Mandforgottoupdatethemargin,theclaimed24.7512M and forgot to update the margin, the claimed 24.7% would be wrong (120M / $512M = 23.4%). Always re-derive percentages, growth rates, and ratios from their components.
The mistake: Checking that the chart label says “485M"butnotverifyingthatthebarinthechartactuallyrepresents485M" but not verifying that the bar in the chart actually represents 485M.Why it happens: Chart labels are visible text that gets checked. Source data is hidden in the chart’s data table and requires clicking into the chart to inspect.The fix: For every chart, open the data source and verify the numbers match the labels. A chart where the label says “485M"butthebarrepresents485M" but the bar represents 512M (because the data was updated but the label was not) is particularly dangerous — it creates a visual that contradicts its own annotation.
The mistake: Not checking the speaker notes for outdated numbers.Why it happens: Speaker notes are not visible in the slide view and are easy to forget.The fix: Many firms use speaker notes as talking points for the meeting. If the notes reference “485Mrevenue"buttheslideshows"485M revenue" but the slide shows "512M”, the presenter will quote the wrong number live in the meeting. Check speaker notes if your firm uses them.
The mistake: Missing that “485M"and"485M" and "486M” both appear as “revenue” on different slides.Why it happens: The difference is small enough to seem like rounding. But 485Mand485M and 486M cannot both be the exact same revenue figure.The fix: When two similar but not identical numbers appear for the same metric, investigate. Is one rounded differently? Is one from a different period (Q3 annualized vs. full-year actual)? If they represent different things, label them clearly. If they should be the same, fix the discrepancy.
The mistake: The deck is dated “November 2024” but contains December 2024 data, or the deck is undated entirely.Why it happens: The cover slide date is often set when the deck is first created and never updated.The fix: The cover date should reflect either the date of the meeting or the “as of” date for the data. Verify it matches the data throughout the deck. An undated deck creates ambiguity about when the analysis was performed.
The mistake: Revenue shown as “485M"onslide3,"485M" on slide 3, "485 million” on slide 7, and “$0.485B” on slide 12.Why it happens: Different people edit different slides and use their own preferred formatting.The fix: Establish a convention at the start: “M"or"M" or "MM” or ”$ million”. Apply it throughout. During QC, search for all variants and flag inconsistencies. The deck should use exactly one format for dollar amounts.
The mistake: Revenue was updated from 485Mto485M to 512M, but the “+15% YoY growth” on the same slide was not updated. If the prior year was $421M, the correct growth rate is now 21.6%.Why it happens: The QC reviewer checks that the revenue number is correct and moves on, not noticing that the growth rate next to it was derived from the old revenue.The fix: When a primary number changes, trace all derived numbers: growth rates, margins, market share percentages, multiples. Each derived number needs to be recalculated. Flag any derived number that sits near an updated primary number.

​
Daily Workflow

It is 11:30 PM. The deck for tomorrow’s 8 AM client meeting just finished its final round of edits. The VP asks you to do a final QC pass.Workflow:
  1. Start with the executive summary (Slide 2-3) — this is the highest-risk slide because it contains the most numbers and is read most carefully
  2. Extract all financial figures and build a quick reconciliation: revenue, EBITDA, margins, EV, multiples — do they match across slides?
  3. Check every chart: do labels match data? Are source citations present?
  4. Scan for language issues: contractions, casual phrasing, exclamation points
  5. Check formatting consistency: MvsM vs MM, date formats, font sizes
  6. Report findings by severity. Recommend fixing critical issues immediately and flagging important issues for a post-meeting revision
  7. Total time: 30-45 minutes for a 30-slide deck
The deck supports a fairness opinion and will be filed with the SEC. The stakes are higher than a typical pitch because errors can have legal consequences.Workflow:
  1. Verify every financial figure against the source data (10-K, 10-Q, or company-provided data)
  2. Check that all valuation ranges are supported by the underlying analysis (DCF output, comps output, precedent transactions)
  3. Verify that the board presentation date matches all “as of” dates in the data
  4. Check for required legal disclaimers and regulatory language
  5. Verify that the peer set in comps is defensible (no obviously non-comparable companies)
  6. Document every finding with the exact slide, cell, and source reference
  7. Flag any finding that could be material for legal review
Your team is preparing a CIM for a sell-side mandate. The document is 80+ pages and you are responsible for the financial section QC.Workflow:
  1. Extract all financial data from the CIM’s financial section
  2. Reconcile every number against the company’s audited financials
  3. Verify that historical figures match the 10-K exactly (CIM numbers face buyer scrutiny)
  4. Check that projections are internally consistent (revenue growth rates match the dollar amounts)
  5. Verify that the management case and the bank case (if presented) are clearly distinguished
  6. Check all charts and tables for source citations and consistent formatting
  7. Flag any number that differs from the audited financials with an explanation

​
Practice Exercise

Scenario: You receive a 15-slide pitch deck for “Project Beacon” — a sell-side advisory presentation for a consumer products company. The deck has been through 4 rounds of edits by different team members. Slides you can see (summarized):
  • Slide 1: Cover — “Project Beacon: Preliminary Discussion Materials, January 2025”
  • Slide 3: Exec Summary — “FY2024 revenue of 1.2B,growing141.2B, growing 14% YoY. EBITDA of 180M (15.0% margin). Trading at 8.5x LTM EBITDA.”
  • Slide 5: Financial Detail — Revenue chart: FY2022 920M,FY2023920M, FY2023 1.05B, FY2024 1.22B.EBITDA:1.22B. EBITDA: 138M, 152M,152M, 183M.
  • Slide 8: Comps Table — Subject company EV/EBITDA = 9.1x. Peer median = 10.5x.
  • Slide 10: Valuation Summary — “DCF implies 22−22-28 per share. Comps imply 24−24-30 per share. Current price: $19.50.”
  • Slide 12: Text reads “margins have been under pressure” alongside a chart showing margin expansion from 15.0% to 15.4%.
Your tasks:
  1. Identify all number inconsistencies (there are at least 4)
  2. Flag any data-narrative misalignment
  3. Identify any calculation errors (check the margins and growth rates)
  4. List any language or formatting issues you notice
  5. Categorize all findings as Critical, Important, or Minor
Start with the revenue: Slide 3 says “1.2B"butSlide5shows"1.2B" but Slide 5 shows "1.22B” — is this rounding or an error? Then check the growth rate: Slide 3 says “14% YoY” but 1.22B/1.22B / 1.05B - 1 = 16.2%. The EBITDA margin: Slide 3 says “15.0%” but 180M/180M / 1.2B = 15.0% while 183M/183M / 1.22B = 15.0% — both round to 15.0% but the dollar amounts differ. And the EV/EBITDA: 8.5x on Slide 3 vs 9.1x on Slide 8.

​
How to Add to Your Local Context

1

Install the Plugin

claude plugin install financial-analysis@financial-services-plugins
2

Add Firm Standards

Edit skills/ib-check-deck/SKILL.md:
## Firm-Specific QC Standards
- All decks must include standard disclaimer on the last slide
- Revenue figures always in USD millions, never billions
- Unit convention: "$M" (not "$MM" or "$ million")
- Date format: "January 2024" (not "Jan 2024" or "1/2024")
- Source citation format: "Source: [Provider], as of [Month DD, YYYY]"
3

Add Required Disclaimers

## Required Disclaimers
- Fairness opinion decks: Include regulatory disclaimer on every slide
- M&A pitches: Include confidentiality notice on cover slide
- All client-facing decks: Include "Preliminary and Confidential" header

​
Best Practices

These are critical habits for effective deck QC:
  • Check the executive summary first — This is the slide most likely to have outdated numbers because it gets written early and edited last.
  • Verify chart source data — A chart label might say “$485M” but the actual bar in the chart might represent a different value because the source data was not updated.
  • Watch for near-miss numbers — 485Mand485M and 486M might both be “revenue” but from different periods. This is harder to catch than 485Mvs485M vs 500M.
  • Check speaker notes — Some firms use speaker notes as talking points. If the notes reference outdated numbers, the presenter will quote wrong figures in the meeting.
  • Always QC the full deck — Even if you only edited 3 slides, your changes may create inconsistencies with the other 27 slides.
  • Build a reconciliation table — For decks with many financial figures, extract all numbers into a simple table organized by metric. This makes inconsistencies immediately visible.
  • Check the comps table carefully — Comps slides are the densest data in any deck. Verify that multiples reference the correct financial metrics and that all companies use the same time period.
  • Verify valuation ranges — If the DCF implies 38−38-45 and comps imply 35−35-42, make sure these ranges appear correctly on the valuation summary slide. Transposed endpoints are a common error.
  • Run the check twice — A second pass catches errors you missed the first time. The first pass finds the obvious issues; the second pass catches the subtle ones that were masked by your focus on the first set of findings.