A 70% Error Rate Across the Top-20 Parsed Deal Values Invalidates July 2026 Aggregate Transaction Totals as Deal Type Spans 2.79 Points From Private Placements at 7.60 to Licensing at 4.81

InnoDexis has published its latest Corporate Intelligence Report — What Was Paid — analyzing transaction type and deal value fields across 5,529 valid Corporate-stream records during July 2026, covering 1,248 records carrying a transaction type across 11 categories and 413 parsed deal values. The report reveals that a manual audit of the 20 largest parsed values found 5 correct, 2 partially correct, and 12 misattributed — a 70% error rate — that no aggregate deal total is reported as a result, and that private placements lead all transaction types at a mean InnoDexis score of 7.60 against a stream mean of 5.52.

Key Findings

The central finding of the report is a field reliability failure concentrated precisely where it causes the most analytical damage. The top 5% of parsed deals account for 87.5% of the summed value — and a manual audit of the 20 largest values found 12 wrong, 2 partially correct, and only 5 verified as correct. The failure modes are systematic rather than random: three captured government programme or budget totals, three captured unrelated organisational scale including a bank balance sheet cited in a sponsorship announcement, two captured annual sector investment aggregates, one captured a company valuation rather than the round size — a USD 6.8 billion valuation reported instead of the USD 500 million raised — and one was a currency conversion failure in which a Hong Kong dollar listing value was treated as US dollars and overstated by approximately 7.8 times. Because the error and the value mass are co-located in the same tail, no aggregate total, sector capital allocation, or month-on-month value comparison is reported anywhere in this document.

The five verified correct values establish the upper anchor of what the report is willing to publish: HII's submarine construction contracts for Block VI Virginia-class attack and Build II Columbia-class ballistic missile submarines at USD 76.6 billion, AT&T's spectrum acquisition from EchoStar at USD 23 billion, the Hologic take-private by Blackstone and TPG at up to USD 18.3 billion enterprise value, a Lockheed Martin logistics and sustainment contract for US Special Operations Command at USD 10.5 billion, and Grant Thornton Advisors' acquisition of CBIZ at USD 5 billion supported by New Mountain Capital. All five are individually verified against their source announcements.

Value disclosure rate varies from 96.7% for private placements and 79.1% for debt financings down to 20.9% for acquisitions and 12.2% for licensing agreements — a spread that sorts almost perfectly along a regulatory line. Transactions touching securities markets disclose because they are required to. Transactions governed by private commercial agreement disclose only when a party benefits from it. The acquisition disclosure rate of 20.9% is the most consequential: four in five acquisitions in this dataset do not state a price, meaning any M&A value analysis built on this stream observes a non-random fifth of the market — specifically the fifth where at least one party is publicly listed or the buyer wants the number known.

Transaction type predicts assessed quality across a clean 2.79-point range. Private placements lead at 7.60 and strategic investments at 7.51 — both more than two points above the stream mean of 5.52. Joint ventures reach 7.16. Licensing agreements score 4.81 at the bottom. The ordering matches the partnership type quality gradient documented in the companion partnership report and for the same structural reason: transactions where a party takes an equity position or commits to building something jointly attach to more substantive announcements than transactions transferring an existing right. Deals disclosing a value score 6.47 against 5.99 for those that do not — a gap the report identifies as a correlate of deal type rather than an independent signal, since the high-disclosure types are also the high-scoring ones.

The untyped Other category holds 330 records — 26.4% of typed transactions — and carries the second-highest value disclosure rate at 45.5%. The report identifies this as a taxonomy gap rather than a genuine residual: recurring structures visible on review include sponsorship and naming-rights arrangements, government funding allocations that are neither grant nor contract, real-estate and project financing, and settlements. A quarter of the deal data is landing in a category that effectively means unclassifiable, and the fix is vocabulary extension rather than reclassification of existing categories.

The stream contains 183 market-research press releases — 3.3% of valid records — published by firms whose product is the report itself rather than any corporate event. They score 4.85 against 5.54 for the rest of the stream and carry large dollar figures in fields intended for transaction values. One firm accounts for 140 of the 183. All 183 are excluded from every figure in this report. The report identifies a source-level ingestion filter as the appropriate remedy, since a small exclusion list on publisher names would remove most of the contamination before it reaches the analysis layer.

Strategic Insight and Trend Analysis

The most consequential structural finding of the What Was Paid report is methodological rather than substantive: the transaction value field is not producing deal consideration figures — it is producing any monetary figure associated with an announcement — and the distinction between those two things cannot be recovered after the fact without manual review. This is identified as a schema design problem rather than an extraction failure. The field is performing its specified function faithfully; the specification is wrong. Splitting transaction value into three separate fields — consideration amount, currency code, and figure type distinguishing consideration from valuation, programme total, and contextual figure — is the architectural resolution, and it is a prerequisite for any deal-value analysis rather than an incremental improvement to the current one.

The disclosure asymmetry between regulated and private transactions is structural and will persist regardless of any schema remediation. Securities transactions disclose because law requires it. Private commercial transactions disclose when a party benefits. The 20.9% acquisition disclosure rate and 12.2% licensing disclosure rate are not data quality problems — they are accurate measurements of how commercial negotiation works. Any product feature or intelligence output promising comprehensive deal-value coverage will be systematically incomplete in private markets, and the honest framing is coverage of disclosed transactions rather than of transactions. This structural reality also explains why strategic investments at 53.7% disclosure and 7.51 mean score represent the best available combination of quality and transparency in the deal data: they sit at the boundary between regulated and private markets, and they are the deal type most likely to indicate a corporate acquirer establishing a strategic position before a full acquisition.

The standing top-20 manual audit introduced in this report carries an insight that extends beyond transaction data. The error mode it caught — a plausible figure correctly parsed from the wrong context — is invisible to range checks, completeness checks, and automated validation of any kind. It requires a human reading the source announcement and verifying that the figure belongs to the transaction being reported. The report recommends this audit become a standing monthly control on any newly extracted numeric field, not only transaction value.

Global and Industry Implications

For corporates and R&D teams, the five verified large transactions each carry downstream intelligence signals that headline value figures do not surface. The HII submarine contract at USD 76.6 billion is identified as a decade of predictable demand for a specific supplier network — meaning supplier-level announcements referencing Virginia and Columbia programmes in subsequent months are the observable trace of how that capital distributes. The Lockheed Martin sustainment contract at USD 10.5 billion is identified as one of the larger enterprise data undertakings announced in July 2026 in practical terms, since sustainment increasingly means predictive maintenance, parts availability modelling, and fleet analytics — software problems attached to hardware contracts. The Grant Thornton and CBIZ combination at USD 5 billion is identified as a buyer-side consolidation event for the Technology-to-Financial-Services flow documented in the companion market report, since professional services firms are among the largest buyers of enterprise software and AI tooling and consolidation concentrates that purchasing without reducing its volume.

For investors and capital allocators, the strategic investment category at 7.51 mean score across 108 records and 53.7% value disclosure is identified as the highest-quality and most transparent deal type in the data — the one most likely to indicate a corporate acquirer establishing a position before a full acquisition. The AT&T spectrum acquisition at USD 23 billion is identified as a leading indicator for network capital expenditure: an operator that has committed USD 23 billion to frequency rights has committed to the equipment and deployment spend that makes them useful, typically visible in vendor contract announcements within two to four quarters. The Hologic take-private structure — contingent value right indicating buyer-seller valuation disagreement about a specific material outcome — is identified as a signal that divestment activity rather than product investment is the likely innovation-relevant development over the next 18 months.

For policymakers and national innovation bodies, the 7.5% of value-stating records using currencies outside the conversion table — including Hong Kong dollar, Singapore dollar, Korean won, Brazilian real, Norwegian and Danish krone, Polish zloty, Mexican peso, and South African rand — each treated as US dollars until the schema is corrected, represents a systematic geographic coverage gap that would skew any regional capital allocation analysis built on current extraction. The market-research press release contamination — 183 records at 3.3% of valid Corporate records, concentrated in one publisher accounting for 140 entries — illustrates the class of source-level data quality intervention that national innovation data infrastructure programmes should treat as a governance requirement rather than a post-hoc analytical correction, with source-level filtering at ingestion as the appropriate policy instrument.

InnoDexis Statement

"The most useful outcome this month's deal data produced was a precise account of what it cannot be used for — and that precision, including the explicit rejection of a USD 1.36 trillion aggregate that an earlier pass produced, is the standard of analytical honesty that transaction intelligence requires," noted InnoDexis in its latest intelligence report.

Conclusion

The What Was Paid report establishes that the transaction value field is unreliable for aggregation due to a 70% misattribution rate at the top of the value distribution where 87.5% of summed value sits, that deal type is a clean and high-discriminating field spanning 2.79 points of InnoDexis score requiring no remediation, and that the disclosure asymmetry between regulated and private transactions is a structural property that will persist regardless of extraction improvement. Across 1,248 typed transactions and 413 parsed values, the evidence confirms private placements and strategic investments as the highest-scoring and most transparent deal types, a 26.4% Other residual requiring vocabulary extension, and five individually verified transactions — spanning submarine construction, spectrum acquisition, diagnostics take-private, military sustainment, and professional services consolidation — as the only deal values this report is willing to publish. As the transaction value field is split into consideration, currency, and figure type at the schema level, the top-20 audit becomes a standing monthly control, and the currency conversion table is extended to cover all represented currencies, the What Was Paid framework will produce the most structurally reliable transaction intelligence the InnoDexis platform has yet attempted. The complete What Was Paid Transaction and Deal Intelligence July 2026 Report is available to InnoDexis subscribers and enterprise clients.

About InnoDexis

InnoDexis is a global Innovation Intelligence platform that tracks, analyzes, and interprets breakthrough innovations, prototypes, and emerging technologies across industries and countries. Its intelligence helps corporates, investors, and policymakers understand the true structure and direction of global innovation. Learn more at innodexis.ai.