Banks measure the cost of everything except the artifact they produce most. Ask what a pitchbook costs and you get a shrug, because nobody's timesheet has a line for "rebuilt slide 14 again". So we modeled it: we built the cost calculation once, then ran it 200,000 times, each run drawing every input from a documented range instead of a single guessed number. Every range is either tied to a public source or openly labeled as an assumption, and all of it is published.
One thing before the numbers: this is a model, not a measurement. No public timesheet data for deck work exists. What a transparent model buys you is honesty about ranges and drivers instead of a fake-precise point estimate.
What is solid, and what is assumed
The model's inputs are not equally trustworthy, so here is the honest inventory before any numbers:
Well documented: what bankers and consultants earn, and how many hours they actually work. These come from large published surveys: two compensation reports covering 656 PE professionals and roughly 200 IB data points, and a 531-person working-conditions survey (74-hour weeks on average). Every number we took from them was verified against the original documents.
Reasonably grounded: how long a deck is. We cross-checked the range against , where the typical deck has run about 26 pages for a decade, plus a hand-collected sample of 21 filed board decks (median 25 pages).
Openly assumed: how many hours one page takes to build, how much senior review time a deck consumes, and what share of a working week goes into decks at all. No reliable public data exists for any of these, so the model uses wide, clearly labeled ranges. Every result below should be read with that in mind, and the sensitivity section shows exactly how much each assumption matters.
The shape of the number
Three typical deck types, each run through the simulation. Because the inputs are ranges, the honest output is a range too: a low, a typical and a high estimate rather than one number.
| Deck type | Low | Typical | High | Team-hours (low–high) |
|---|---|---|---|---|
| Sell-side pitchbook (IB) | $6.6k | $14k | $26k | 48–180 |
| IC deck (PE) | $4.1k | $7.8k | $14k | n/a |
| Client deck (consulting) | $7.7k | $14k | $25k | 81–240 |
Two readings of the same typical pitchbook are worth separating. Priced at junior rates only, it costs about $6.8k ($3.3k to $13k). The full number adds senior review, which the model prices at VP-level rates and which accounts for roughly half of the typical total. Both readings are in the published results; quote whichever matches your question, but quote it with its range.
Pages, not pay, drive the cost
We stress-tested every input: hold everything else steady, move one input across its full range, and watch how far the typical cost moves. The result is unambiguous: the two structural inputs, how many pages you build and how long a page takes, dominate everything, while the well-documented pay inputs barely register. This finding is the most robust thing in the model, because it survives almost any choice of ranges.
Deck length and effort-per-page dominate; compensation barely matters
Monte Carlo sensitivity, seed 20260706, n=200,000, github.com/Gedonus-AI/deck-economics
Roughly half the money buys rework, not thinking
The model separates a deck's hours into first draft, revision churn and senior review. Churn, the v4-to-v40 cycle of rebuilt charts and reflowed pages, takes 27% to 47% of the hours across deck types. Convert hours to cost and attribute review time to the churn that triggers it, and 36% to 62% of deck cost (typically around half) is churn. On the strictest reading, pricing only junior rework hours, it is still 19% to 41%.
The interesting part is what churn is not: it is rarely new analysis. The underlying model moved, and the deck, which has no live link to it, had to be rebuilt by hand. That failure mode is measurable elsewhere too: in , the median deck's last save landed just hours before filing.
A seat, a team, a year
Scale the per-deck numbers up to a year and they arrive at: one IB analyst seat represents $57k to $110k per year of presentation-production cost (typically $83k); a six-junior coverage group, $0.88M to $1.9M per year. For consulting, deck production runs around $47k per team-month at cost, roughly 7% to 13% of a typical engagement-month price.
What this means for deal teams
Measure the churn, not the deck count. The model says the money is in revisions. If your team tracks anything, track how many times a "final" page gets rebuilt; that number converts to dollars faster than any headcount discussion.
Attack geometry before rates. Shorter books and less manual effort per page are the only levers with five-figure swings. Standardized components, live links from model to slide, and diff-based review shrink exactly those two parameters.
Distrust point estimates, including ours. Anyone quoting "a pitchbook costs $X" without a range is selling something. The honest statement is a distribution, and we published ours with every assumption inspectable.
Methodology
The cost calculation runs 200,000 times (a "Monte Carlo simulation"); each run draws every input from a range with a most-likely value in the middle, and the spread of outcomes gives the low, typical and high estimates. The ranges come from graded public sources: working hours from the Wall Street Oasis 2024 survey (531 bankers) and eFinancialCareers 2025 (2,500+ finance professionals); pay from Heidrick & Struggles' 2025 PE survey (656 professionals), Mergers & Inquisitions 2026 and official US employment-cost data; deck length cross-checked against our and a hand-collected sample of 21 filed board decks. Where no reliable source exists (hours per page, review share, time share), the range is a clearly labeled assumption, made deliberately wide. Hourly costs divide real total pay by the real surveyed hours: the 74-hour weeks make the rates lower, not higher, which keeps the model conservative. The simulation is seeded, so it reproduces exactly.
The core parameters and where they come fromall 30+ rows in METHODOLOGY.md
| Parameter | Distribution (min / mode / max) | Basis |
|---|---|---|
| Pitchbook pages | tri(15, 35, 90) | assumption; guides + n=21 filed-deck sample |
| First-draft hours per page (IB) | tri(0.25, 0.75, 1.5) | assumption; C-grade corroboration only |
| Revision-churn factor (IB) | tri(0.4, 0.9, 2.0) | sourced |
| Senior-review share of junior hours | tri(0.15, 0.30, 0.50) | assumption |
| IB analyst cash comp (USD/yr) | tri(165k, 195k, 225k) | sourced (M&I 2026) |
| IB analyst hours/week | tri(60, 74, 95) | sourced (WSO 2024, n=531) |
| Load multiplier on comp | tri(1.15, 1.30, 1.45) | sourced (BLS ECEC) |
| Time share on presentations (IB) | tri(0.15, 0.30, 0.55) | assumption; the model's weakest input |
Every input is either tied to a graded source or labeled an assumption; nothing hides in code. The repo's METHODOLOGY.md carries the full table with source IDs and the reasoning per range.
Limitations
This is a model, not a measurement; no public timesheet data exists to validate it. The two largest cost drivers rest on labeled assumptions corroborated only by weak-grade material plus our small primary sample. Compensation sources are US- and NY-centric and self-selected. The PE IC-deck archetype is the least reliable of the three. The extremes are likely understated; no plausible-range model contains the v44 all-nighter. All results are reported as low/typical/high ranges, deliberately rounded, for exactly these reasons.
Data & code
Model, parameter-to-source table, graded source list and every chart's data: github.com/Gedonus-AI/deck-economics. You do not need to read code to check the work: the methodology document lists every input, its range and its source in one table. If you disagree with an assumption, change one line and the model gives you your own range. If you use these numbers, cite this page and the repo, and keep the ranges attached.
gedonus attacks the two parameters that matter: pages that rebuild themselves from the model, and review that reads a diff instead of re-reading a deck. See it inside PowerPoint.
