Verification Report 2 — Chapters 13, 14 and 15

Audit date: September 15, 2026 Scope: 13-endgame-and-ownership-strategy.md, 14-exit-deal-mechanics.md, 15-post-mortem-library.md. Method: Independent recomputation of every worked numerical example in Python; direct fetch of primary/near-primary sources (srsacquiom.com, site.warrington.ufl.edu (Ritter), federalregister.gov, arnoldporter.com, klgates.com, en.wikipedia.org, plus trade-press searches); mechanical cross-referencing of chapters 13/14/15 against 07, 09, 10 and 11. Standards applied: same as the first report — recompute rather than re-read; verify against the source the chapter itself cites; where a source cannot be reached, mark UNVERIFIABLE rather than guess.

Headline: the two Chapter 14 proceeds waterfalls — the highest-risk arithmetic in the three chapters — are exact on every line. The defects are concentrated in Chapter 13's decision-framework examples and Chapter 15's synthesis tabulations.

1. Defects found

D1 — ERROR — Chapter 13 Example 2 overstates the required exit by roughly 85%

• File/location: 13-endgame-and-ownership-strategy.md, §"Example 2 — Seed-funded AI startup, $500K ARR", Step 3 (line 952).
• Claim as stated: "At 48% combined ownership, clearing $5M each requires a sale above roughly $30M after the preference — call it a $40M+ exit at current ownership."
• What I found: the stated cap table is founders 48%, pool 15%, investors ~37%, preference $8.5M, 1x non-participating. Working that model:
  • Preferred converts (and the preference stops binding) above a sale of $22.97M (0.37 × S > $8.5M).
  • Below that, founders take 48/63 of the residual. Solving for $10M combined gives a required sale of $21.6M, not $40M.
  • Even reading the founders' "$5M each" as after tax at 28.8% gives $26.9M.
  • At the chapter's own "$40M+", the preferred converts and founders receive 48% × $40M = $19.2M, or $9.6M each — nearly double the stated Number. Neither reading of "above roughly $30M after the preference" produces $40M either: $30M of residual would require a ~$38.5M sale and deliver ~$22.9M to founders.
• Severity: ERROR. This is a load-bearing number: it drives the example's recommendation to "take a serious meeting with two or three plausible acquirers before raising," and the overstatement makes a small sale look far less viable than it is.
• Recommended correction: "clearing $5M each requires a sale of roughly $22M — and above about $23M the preferred converts and stops mattering, so the founders' 48% applies to the whole price. Call it a $22–27M exit at current ownership, pre- and post-tax, and considerably more after a Series A's dilution."

D2 — ERROR — Chapter 15 Synthesis 3's capital bands sum to 49, not 53

• File/location: 15-post-mortem-library.md, §"Synthesis 3 — Time-to-death patterns" (line ~1185), the "Median lifespan by capital raised" table.
• Claim as stated: "computed from the 53 entries," with band counts n = 6, 12, 11, 15, 5.
• What I found: 6 + 12 + 11 + 15 + 5 = 49. Four entries are unaccounted for. This is the only table in the chapter that fails to reconcile — Synthesis 1 (11+10+9+7+6+4+3+2+1 = 53) and Synthesis 2 (19+22 = 41 attributable, +12 non-attributable = 53) both close exactly, and I confirmed the chapter contains exactly 53 numbered entries (## 1. through ## 53.).
• Severity: ERROR.
• Recommended correction: re-derive the five band counts so they total 53, or state explicitly that four entries had no reliable published raise figure and were excluded (Relay's own entry concedes "full totals not published," so an explicit exclusion note may be the honest fix).

D3 — ERROR + INCONSISTENT — Chapter 13's three SRS earn-out figures cannot all be true, and the headline differs from Chapter 14's

• File/location: 13-endgame-and-ownership-strategy.md lines 369–379 (§A5); compare 14-exit-deal-mechanics.md lines 912–916 (§H2).
• Claim as stated (Ch 13, flagged [Verified]): "Just over 50% of deals with earnouts see any payment at all... Among deals that do pay, the average payout is about 50 cents on the dollar... Across all deals with earnouts, including the non-payers, the blended outcome is about 21 cents on the dollar."
• What I found: two problems.
  • (a) Internal arithmetic. A payment rate "just over 50%" multiplied by a payer average of "about 50 cents" yields a blended 25–26 cents, not 21. To reach 21 cents at a ~50% payment rate, the average among payers must be ~42 cents. The three figures are presented as a consistent set and are not.
  • (b) Cross-chapter and source. The SRS Acquiom page the chapter cites states only "closer to one out of five dollars gets paid across all deals with an earnout" — i.e. 20 cents. Chapter 14 §H2 reports this correctly as "roughly 20 cents on the dollar" and, prudently, does not repeat the 50%/50-cent decomposition. Chapter 13 states 21 cents four times (lines 15, 373, 379, 975) and builds a worked expected-value calculation on it ($25M + 0.21 × $15M ≈ $28.2M — arithmetically correct at 28.15, but resting on the unsupported 21). I could not find the 50%/50-cent breakdown published anywhere on the cited page; it is UNVERIFIABLE as stated.
• Severity: ERROR (internal arithmetic) + INCONSISTENT (13 vs 14 vs source). The [Verified] tag overstates what the cited page supports.
• Recommended correction: harmonise on the sourced figure. Replace the three-figure decomposition with: "Across all deals with earnouts, including the non-payers, SRS Acquiom's own framing is that 'closer to one out of five dollars gets paid' — roughly 20 cents on the dollar. Roughly half of earn-out deals pay nothing at all. [Verified for the aggregate; the split between payment rate and average payer recovery is not published.]" Then change "21 cents" to "20 cents" at lines 15, 373, 379 and 975 and rerun the EV example ($25M + 0.20 × $15M = $28.0M).

D4 — ERROR — Chapter 15, entry 41: Jasper's Series A was $125M, not $131M

• File/location: 15-post-mortem-library.md, ## 41. Jasper, "Capital raised" line (~line 849).
• Claim as stated: "$131M Series A (October 2022, Insight Partners, Coatue, Bessemer) at a $1.5B valuation."
• What I found: Jasper's October 18, 2022 Series A was $125M at a $1.5B valuation — per Jasper's own announcement and PR Newswire release ("Jasper Announces $125M Series A Funding Round, Bringing Total Valuation to $1.5B"), TechCrunch and SiliconANGLE. $131M is Jasper's cumulative total raised, including the earlier ~$6M. The chapter has attached the lifetime total to the round label. The $1.5B valuation and the investor list are correct.
• Severity: ERROR (minor magnitude, but it is a cited headline figure in an entry the chapter singles out as its only valuation-collapse case).
• Recommended correction: "$125M Series A (October 2022, Insight Partners, Coatue, Bessemer) at a $1.5B valuation; ~$131M raised in total."

D5 — ERROR — Chapter 15, entry 39: Neeva's search shutdown date is wrong

• File/location: 15-post-mortem-library.md, ## 39. Neeva, "Founded / Died" line (~line 808).
• Claim as stated: "consumer product shut down 20 June 2023; company acquired by Snowflake May 2023."
• What I found: Neeva announced the shutdown on May 20, 2023, and the search engine actually stopped on June 2, 2023. The stated "20 June 2023" matches neither date and appears to be a conflation of the May 20 announcement with the June shutdown month. The Snowflake acquisition (announced May 24, 2023) and the "~$77.5M raised" figure are both correct.
• Severity: ERROR (date).
• Recommended correction: "shutdown announced 20 May 2023; consumer search switched off 2 June 2023; company acquired by Snowflake, announced 24 May 2023."

D6 — INCONSISTENT — Chapter 15's Synthesis 2 capital buckets do not reconcile with Synthesis 3 or with its own list of non-attributable entries

• File/location: 15-post-mortem-library.md, Synthesis 2, structural observation 1 (line 1174), read against Synthesis 2's opening paragraph and Synthesis 3's band table.
• Claim as stated: "Among entries under $10M raised, the stated cause matched the evidence in 12 of 15 attributable cases (80%). Among entries over $100M raised, it matched in 3 of 14 (21%)."
• What I found: the two percentages compute correctly (12/15 = 80.0%, 3/14 = 21.4%), but the denominators conflict with the rest of the chapter:
  • Synthesis 3 puts 18 entries under $10M (6 bootstrapped/<$1M + 12 in $1M–$10M), yet Synthesis 2 finds only 15 attributable there. Synthesis 2 names all 12 non-attributable entries — Cerebral, LendUp, Frank, Nate, Builder.ai, HeadSpin, Hipmunk, One Kings Lane, Jawbone, Beepi, Modsy, Katerra — and every one of them raised well over $10M. So all 18 under-$10M entries should be attributable, not 15.
  • Symmetrically, Synthesis 3 puts 20 entries over $100M ($100–500M: 15, >$500M: 5). Eight of the 12 non-attributable entries raised over $100M, implying 12 attributable there, not 14. Because Synthesis 3's bands are themselves broken (D2), at least one of the two tabulations must be re-derived; they cannot both stand.
• Severity: INCONSISTENT. The 80%-vs-21% contrast is quoted as one of the chapter's three structural findings, so the denominators should be defensible.
• Recommended correction: rebuild both tables from a single per-entry table of (capital raised, attributable y/n, match y/n) and publish the counts that fall out. If the 80/21 split survives, keep it; if not, restate it.

D7 — INCONSISTENT — Chapter 15 hardens "alleged" fraud into "evidenced" fraud in the Synthesis 1 table

• File/location: 15-post-mortem-library.md, Synthesis 1 table, row "Fraud or material misrepresentation | 4 | 8% | Frank, Nate, Builder.ai, HeadSpin" (line 1133), read against entries 47 and 48.
• Claim as stated: Synthesis 1 states that "each entry was assigned one primary evidenced cause," and assigns fraud to all four.
• What I found: the individual entries are carefully and correctly hedged — entry 47 (Nate) says "Alleged criminal fraud... As of September 2026 the case is pending; these remain allegations"; entry 48 (Builder.ai) says "The company denied fraud; investigations were ongoing as of 2026." I verified both: Albert Saniger was charged by DOJ and SEC in April 2025 (pending), and Builder.ai entered insolvency in May 2025 with Bloomberg reporting a ~300% overstatement of sales to creditors, no adjudication. Frank (Javice) and HeadSpin (Lachwani) did result in criminal convictions. The synthesis table flattens all four into an evidenced category without the qualifier the entries themselves insist on.
• Severity: INCONSISTENT (the chapter contradicts its own carefully-drawn standard). Not an ERROR — the entries are right; the table is looser than they are.
• Recommended correction: relabel the row "Fraud or material misrepresentation (adjudicated or alleged)" and footnote it: "Frank and HeadSpin resulted in convictions; the Nate and Builder.ai matters were unresolved as of September 2026 — see entries 47 and 48."

D8 — MINOR — Chapter 13 misreads the 2025 row of Ritter's SPAC table

• File/location: 13-endgame-and-ownership-strategy.md, §A12 (line 746).
• Claim as stated: "The 2025 vintage is structured more conservatively than 2021... but the returns column above does not yet show any improvement."
• What I found: the table's data is exactly right (see §2), but the two columns are indexed differently. "SPAC IPOs" counts vehicles by IPO year; "avg. 1-year post-merger return" covers de-SPACs measured one year after their merger. SPACs that IPO'd in 2025 cannot yet have completed a year post-merger, so the −57.1% in the 2025 row cannot be evidence about the 2025 IPO vintage. The sentence draws a conclusion the column cannot support.
• Severity: MINOR (interpretation, not data).
• Recommended correction: "...the 2025 vintage is structured more conservatively than 2021. Its post-merger record is not yet observable — the returns column reflects de-SPACs completing their first year, not the cohort that IPO'd that year — so treat the structural improvement as untested."

D9 — MINOR — Chapter 13's $100M cap table sums to 75%, under a claim that "arithmetic is exact"

• File/location: 13-endgame-and-ownership-strategy.md, §"The $100M sale where the founder got nothing", lines 299–323.
• Claim as stated: header "[Analysis — illustrative worked example; arithmetic is exact]"; cap table "founders 18%, employees/option pool 14%, Seed 4%, A 8%, B 13%, C 18%, remainder to later angels and secondary buyers rolled into the preferred classes."
• What I found: the listed percentages total 75%, leaving 25% unallocated. The hedge is self-defeating: if the remainder is genuinely rolled into the preferred classes, those classes' percentages would already include it and the remainder would be zero. As written a reader cannot tell whether that 25% carries additional preference, which would change the waterfall. The waterfall arithmetic itself is exact and the conversion logic is sound (see §2) — only the cap table fails to close.
• Severity: MINOR (the stated numbers all follow; the setup is underspecified while claiming exactness).
• Recommended correction: raise the six percentages so they total 100% (e.g. Seed 5%, A 11%, B 17%, C 25%, founders 18%, pool 14% — recheck each conversion test afterwards), or drop "arithmetic is exact" in favour of "the waterfall steps below are exact on the stated inputs."

D10 — MINOR — Chapter 13 Example 3: "$1.8M" is equal to, not larger than, two years of distributions

• File/location: 13-endgame-and-ownership-strategy.md, Example 3 (line 972).
• Claim as stated: moving the multiple from ~3x to ~5x on $900K of earnings "is $1.8M of created value, larger than two years of total distributions."
• What I found: the $1.8M is right ((5−3) × $900K). But total distributions are $540K + $360K = $900K/yr, so two years is exactly $1.8M. Equal, not larger. Every other figure in Example 3 checks out (see §2).
• Severity: MINOR.
• Recommended correction: "roughly equal to two full years of the owners' combined distributions — and it is permanent, where the distributions are not."

D11 — MINOR — Chapter 15's bootstrapped-entry count does not reconcile, and renames two entries

• File/location: 15-post-mortem-library.md, Synthesis 1 sample-bias warnings (line 1143).
• Claim as stated: "Roughly 46 of 53 raised institutional capital; only five entries (Apollo, Tapbots, The Iconfactory, HouseFresh, and most of the wrapper cohort) were bootstrapped."
• What I found: 46 + 5 = 51, leaving two entries in neither category; "roughly" only partly covers this. Separately, the two companies are listed in the Synthesis 1 table under their product names — Tweetbot (Tapbots) and Twitterrific (The Iconfactory) — so a reader cross-referencing the two lists will not match them.
• Severity: MINOR.
• Recommended correction: make the counts complement to 53 and use one naming convention throughout, e.g. "47 of 53 raised institutional capital; six (Apollo, Tweetbot/Tapbots, Twitterrific/The Iconfactory, HouseFresh, Darklang, and most of the wrapper cohort) were bootstrapped."

D12 — MINOR — Chapter 14's §1202 section omits the 28% rate trap that Chapter 11 flags

• File/location: 14-exit-deal-mechanics.md, §J2, "Where founders lose QSBS by accident" (line ~1220); compare 11-legal-operations-and-founder-sustainability.md line 634.
• What I found: the two chapters' QSBS tables are fully consistent — $10M→$15M per-issuer cap, $50M→$75M gross-assets ceiling, both inflation-indexed from 2027, tiered 50%/75%/100% at 3/4/5 years, grandfathering for stock acquired on or before July 4, 2025 (see §2). But Chapter 14 describes the tiering purely as "a 3-year sale gets 50%, not 100%," while Chapter 11 §10 adds the material point that "the non-excluded portion of gain for stock held three or four years is taxed at the 28% §1202 rate, not the usual 15%/20% long-term capital gains rates. The partial exclusions are worth less than the headline suggests." A reader of Chapter 14 alone will overvalue a 3-year exit.
• Severity: MINOR (omission, not contradiction).
• Recommended correction: add one sentence to Chapter 14 §J2 carrying Chapter 11's 28%-rate point, with a cross-reference.

D13 — MINOR — Chapter 14's S-corp tax line ignores the §1411(c)(4) NIIT limitation

• File/location: 14-exit-deal-mechanics.md, §I2, Founder A's outcome table (line 1062).
• Claim as stated: "Federal LTCG 20% + NIIT 3.8% + state 5%, on near-zero basis" — a flat 28.8% on $6,417,000.
• What I found: the arithmetic is exact (28.8% × $6,417,000 = $1,848,096, stated as $1,848,000). But the example's Founder A is an owner-operator of an S corporation, and §1411(c)(4) limits net investment income on the disposition of an interest in an S corporation in which the shareholder materially participates to the gain attributable to the entity's passive assets. A full-time founder would plausibly owe little or no NIIT on this sale, making 28.8% an over-estimate and the net proceeds an under-estimate by up to ~$244K.
• Severity: MINOR (conservative in the seller's favour, and the chapter's teaching point survives either way; but it is stated as settled).
• Recommended correction: footnote it: "NIIT is applied here for simplicity; under §1411(c)(4) a materially participating S-corporation shareholder may owe little or none of it on a stock sale — another reason to model this with a tax adviser rather than a rate card."

D14 — MINOR — Chapter 14's second waterfall asserts, but does not demonstrate, that Seed/A/B decline to convert

• File/location: 14-exit-deal-mechanics.md, §I3 (line 1116).
• Claim as stated: "Seed, A and B (45%) take their preferences and do not convert, because conversion would pay them far less."
• What I found: the aggregate is right — 45% of $55,030,000 is $24.8M against $30M of combined preference — and under the natural dollar-proportional reading of the 45% (Seed 3%, A 12%, B 30%) each round individually also declines to convert ($1.65M vs $2M; $6.60M vs $8M; $16.51M vs $20M). So the assertion holds. But the conversion test is per-class, the chapter never gives the per-round split, and the conclusion is sensitive to it: if Seed held 5% rather than 3% it would convert, the residual would rise to ~$2.1M, and Founder A's $28,364 would become roughly $426,000 — a 15x swing. Everything else in this waterfall is exact (see §2).
• Severity: MINOR (underspecified input behind a deliberately dramatic headline number).
• Recommended correction: add the per-round as-converted split to the preference-stack table (e.g. Seed 3%, A 12%, B 30%, C 25%, common+options 30%) and one line: "each of Seed, A and B is individually better off taking its preference at this price — check this per class, not in aggregate, because a single early round converting can change the common residual by an order of magnitude."

D15 — MINOR / UNVERIFIABLE — Chapter 15's Forward Health capital figure is contested in the trade press

• File/location: 15-post-mortem-library.md, ## 38. Forward Health (~line 791).
• Claim as stated: "~$650M total, including a $100M Series E in November 2023."
• What I found: the $650M figure is widely repeated, but Fierce Healthcare's shutdown report — the closest thing to a trade-press record of the event — says Forward shuttered "after raising $400M." I could not resolve the discrepancy from a primary source (Forward filed no public financials and the company is gone). The November 2024 shutdown date and the 2016 founding are correct.
• Severity: MINOR / UNVERIFIABLE.
• Recommended correction: "~$400–650M total (reported totals differ materially: Fierce Healthcare says ~$400M, other coverage ~$650M), including a $100M Series E in November 2023."

2. Checked and found correct

2.1 Arithmetic — all worked examples recomputed independently in Python

The arithmetic is, with the single exception of D1, correct. Specifically:

Chapter 14, Worked waterfall 1 (bootstrapped $12M sale) — every line exact, 22/22 checks. EV $12.0M + cash $0.4M − debt $0.9M = equity value $11,500,000; − $150K NWC = $11,350,000. The advisor fee is a correctly computed double Lehman (10/8/6/4% of the first four $1M tranches + 2% of the remaining $8M = $440,000). Net proceeds $10,655,000; escrow at 10% of transaction value $1,200,000; cash at closing $9,455,000. The option treatment is right: aggregate strike 1,000,000 × $0.22 = $220,000 added to the pot gives ($9,455,000 + $220,000) / 10,000,000 = $0.9675/share exactly; Founder A $5,805,000, Founder B $2,902,500, options $967,500 − $220,000 = $747,500 net of strike, and the three sum precisely to $9,455,000. Escrow release $1,200,000 − $180,000 = $1,020,000, split 612/306/102 (correctly pro rata on shares, with the strike already paid). Founder A gross $6,417,000 = 89.1% of the $7.2M naive expectation (stated 89%). Tax at 28.8% = $1,848,096 (stated $1,848,000); net $4,568,904 (stated $4,569,000). The two teaching claims both verify: the gap to naive is $783,000 = 10.9% ("about 11%"), and the net is 38.07% of the $12M headline ("about 38%"). The "60% owner nets ~38% of headline" conclusion is sound.

Chapter 14, Worked waterfall 2 (the "$75M" venture-backed sale) — every line exact, 18/18 checks. Preference stack: Series C 1.5 × $15M = $22.5M plus an 8% cumulative dividend on the original issue price for two years ($15M × 0.08 × 2 = $2.4M) = $24,900,000; total stack $54,900,000 on $45.0M raised. Proceeds: $65M + $3.5M cash − $6M venture debt = $62,500,000; banker fee 2% of EV = $1.3M; RWI $195K (3% of a $6.5M limit) + a $40K underwriting fee = $235K; after expenses $60,030,000; less the $5M retention pool carved out of the price = $55,030,000; escrow at 2.8% of transaction value = $1,820,000, leaving $53,210,000 at closing. Residual = $55,030,000 − $54,900,000 = $130,000. The participation treatment is correct. Series C is participating, so it takes its full $24.9M preference and shares the residual; Seed/A/B are non-participating and take preference without converting, so they are correctly excluded from the residual pool — leaving a 55% base (C's 25% + common's 30%). C takes 25/55 × $130,000 = $59,091; common takes 30/55 = $70,909; the two sum to $130,000. Founders' 18/30 = $42,545; Founder A's 12/30 = $28,364. The final number follows. The carve-out variant is also exact: 8% of $55,030,000 = $4,402,400, leaving $50,627,600 against a $54,900,000 stack, a $4,272,400 shortfall borne by the preferred. Founder A's 30% of the pool = $1,320,720; at 37% + 5% + 2.35% = 44.35%, the net is $734,981 (stated ~$735,000). The 2.35% is correctly 1.45% Medicare + the 0.9% Additional Medicare surtax for a high earner. One framing point worth preserving: the $28,364 and the $735,000 are alternative scenarios, not additive — under the carve-out the remaining $50.6M falls short of the $54.9M stack, so common's residual is genuinely zero, and the chapter says exactly that ("Common receives nothing from the waterfall proper"). This is internally consistent and correctly reasoned. The four "why this happens" bullets also check: $65M − $5M retention = $60M of distributable value against a $75M headline; $6M of venture debt off the top; $15M of Series C investment converted into a $24.9M claim.

Chapter 13, the $100M sale (~$1.4M each) — the waterfall steps are exact. $100M − $50M (senior Series C) = $50M − $30M (B) = $20M − $12M (A) = $8M − $3M (Seed) = $5M residual. Founders' 18/32 × $5M = $2,812,500, i.e. $1,406,250 each ("$1.4M each"); employees 14/32 × $5M = $2,187,500 ("$2.2M"). Each conversion test is correct (C: 18% × $100M = $18M < $50M; B: $13M < $30M; A: $8M < $12M). The Seed step is the subtle one and the chapter gets it right: Seed's 4% of $100M = $4M does exceed its $3M preference, but as the chapter says, a converting Seed would share only the $8M residual alongside common's 32% — 4/36 × $8M = $888,889, well below $3M — so Seed correctly takes the preference. Only the cap table's failure to total 100% (D9) mars this example.

Chapter 13, Examples 1, 3 and 4 — correct. Example 1: $1.0M owner earnings at Acquire.com's median 3.9x = $3.9M, consistent with the stated "~$3.5–4.5M" range and with §A5's own citation of that median. Example 3: $5M × 18% = $900,000 owner earnings; the 60/40 split is $540,000 / $360,000; 3–5x gives $2.7M–$4.5M; the multiple expansion is $1.8M — all exact (only the "larger than two years of distributions" comparison is off, D10). Example 4: ownership sums to 100% (16 + 16 + 68); at $60M the preferred correctly takes preference rather than converting (68% × $60M = $40.8M < $52M), common splits $8M, founders' 16/32 = $4M total, $2M each — exact; and "at $45M, common gets nothing" is right ($45M < $52M). 3–5x on $15M ARR = $45M–$75M, exact.

Chapter 15, Syntheses 1 and 2 — the tabulation is internally consistent. Synthesis 1's nine category counts sum to 53, matching the 53 numbered entries; all nine percentages round correctly (11/53 = 21%, 10/53 = 19%, 9/53 = 17%, 7/53 = 13%, 6/53 = 11%, 4/53 = 8%, 3/53 = 6%, 2/53 = 4%, 1/53 = 2%); and the named entries in each row match each row's count exactly (11, 10, 9, 7, 6, 4, 3, 2, 1 companies listed). The derived claim "57% of these failures (categories 1, 2 and 3 combined)" = 30/53 = 56.6%, correct. Synthesis 2 also closes: 41 attributable + 12 non-attributable = 53; 19 matching + 22 diverging = 41; 22/41 = 53.7% → "54%" and 19/41 = 46.3% → "46%", both correct; the direction-of-bias breakdown 18 + 4 + 0 = 22; and the 12 non-attributable entries are individually named and number exactly 12. The user-supplied framing of "54% of 41 entries" is right, and the chapter is careful to state that the 41 are a subset of 53. The "80% under $10M / 21% over $100M" percentages compute correctly from their stated denominators (12/15, 3/14) — it is only those denominators that fail to reconcile (D6).

2.2 Source verification

2.3 Overlap check (Chapter 15 vs Chapter 09) — clean

I extracted all 53 Chapter 15 entry names and word-boundary-matched each against 09-case-studies.md. There is no substantive duplication. Only two names appear at all, and neither is a Chapter 09 case study:

• Odeo — 3 incidental mentions in Chapter 09, in the pivot discussion (Case 11, Slack), not a case of its own.
• Fast — 1 incidental match, a false positive on the ordinary word.

Chapter 09's failure-side cases (Quibi, Convoy, IRL, Bench, Olive AI, Theranos, Nikola, WeWork, 23andMe, Ginkgo) appear nowhere in Chapter 15, and Chapter 15's 53 entries appear nowhere in Chapter 09's 36 cases. The instruction to avoid duplicating Chapter 09's companies was followed.

2.4 Other cross-chapter consistency (Part C) — no conflicts found

• Valuation multiples. Chapter 13's figures are internally consistent and used consistently: Acquire.com's median 3.9x profit appears at lines 505, 567, 879 and 938 with the same value and the same "below $10M enterprise value" qualifier; the SEG public SaaS index at 4.8x EV/revenue (Q4 2025, down from 6.3x) is used identically at lines 287 and 981; PE involved in 58% of the record 2,698 SaaS M&A transactions in 2025 appears identically at lines 287, 497 and 988. Chapter 14 does not restate any of these, so there is nothing to conflict.
• Earn-out prevalence. Chapter 14 §H1 presents SRS's 24% and the ABA's 18% side by side and attributes each correctly rather than blending them — the right handling of two studies with different samples ($25M–$900M for the ABA). Chapter 13 cites only the SRS 24%.
• Earn-out statistics elsewhere in the library. Chapters 07, 09, 10 and 11 contain no competing earn-out payout statistic, so D3's inconsistency is confined to 13 vs 14.
• Non-competes. Chapter 14 §G9 correctly defers the employment-side analysis to Chapter 11 §7 and confines itself to sale-of-business non-competes, where its statement of California Bus. & Prof. Code §16601 (goodwill/ownership-interest exception) is accurate black-letter law and does not conflict with Chapter 11's account of §16600.5.
• Chapter 14's own source note (line 1460) correctly identifies the two deal-terms studies as the most reliable sources in the chapter and flags the survivorship bias in the advisory-firm and marketplace benchmarks — consistent with the standard Chapter 13 §A sets out and with the first report's methodological findings.

3. Unverifiable

These are findings in their own right, not omissions:

1. Chapter 13, the 50%-pay / 50-cents-among-payers decomposition (line 371–372). Not published on the SRS Acquiom page cited. Flagged [Verified]; it is not. (See D3.)
2. Chapter 13's SPAC table, the 2019 and 2020 rows (−2.0% and −3.0%). The Ritter PDF fetch returned 2021–2025 only. The 2021–25 rows are exact, so these two are probably fine, but they are unconfirmed.
3. Chapter 13's "low dozens" of US direct listings since 2018. The chapter already self-flags this ("I was unable to retrieve the current count at research time — treat as an estimate"), which is the correct handling and worth preserving.
4. Chapter 14's SRS Acquiom sub-statistics — 91% no-shop, 58% appraisal-rights closing condition, 72% separate PPA escrow, 39% worksheet approach, 89% excluding tax items, 54% with both stand-alone and back-door MAC, median earn-out 34% of closing payment, median duration 21 months, metrics revenue 69% / EBITDA 65%. The study's existence, sample and headline framing are confirmed, but these individual data points sit inside a gated PDF I could not open. They are attributed to a real study with a matching sample description and are plausible for that dataset; none was independently confirmed.
5. Chapter 15, Relay's capital raised (~$8–10M). The entry itself says "full totals not published," which is honest; the figure cannot be confirmed.
6. Chapter 15, Forward Health's total raised. Sources conflict between ~$400M and ~$650M (D15).

4. Summary

The two Chapter 14 proceeds waterfalls — the most consequential arithmetic in the three chapters, and the material a reader is most likely to rely on — are exact on every line, including the preference stack, the participation treatment, the option-strike mechanics, the escrow release and the after-tax conversions. Chapter 13's $100M waterfall and its Examples 1, 3 and 4 are also arithmetically sound. The single arithmetic error of consequence is D1, in Chapter 13's Example 2.

Chapter 15's two main synthesis tables (frequency and divergence) are internally consistent and correctly computed; the defects there are in the third table and in the capital-bucket denominators that depend on it. The chapter's handling of the two unadjudicated fraud matters is careful at the entry level and should simply be carried up into the summary table.

Priority of repair: D1 and D3 first (both change numbers a reader would act on), then D2 and D6 together (they must be re-derived from one underlying per-entry table), then D4, D5 and D7.