Original research · BAU AI · 2026

US bid package risk patterns 2026: study design and illustrative preview

This study measures how often three costly risk patterns appear in US commercial bid packages: liquidated damages without a stated cap, scope gaps that only surface when specs are read against drawings, and broad-form indemnity. The analysis is not final, so every figure on this page is an illustrative placeholder, not a finding.

No results yet. This page shows the study design. Every figure marked DATA is a placeholder (XX) that shows how results will be reported. Do not cite them. Final figures will replace them here, and the page will say so.

LD clauses without cap XX%

Share of solicitations with LDs that state no cap.

Placeholder
Broad-form indemnity rate XX%

Share of contracts with indemnity covering the owner's own negligence.

Placeholder
Spec-to-drawing scope gaps XX%

Share of packages with a gap found only by reading specs against drawings.

Placeholder

Why run this study?

Preconstruction risk sits in language buried deep in the bid set: uncapped liquidated damages, broad-form indemnity, and scope gaps that only show when specs and drawings are read together. We found no public US benchmark for how often these patterns appear. This study is designed to build one.

How is the study designed?

Anonymized US commercial bid packages, aggregated only after identifying information is removed. Each finding is tagged against the standard text it departs from: AIA A201, ConsensusDocs, EJCDC or FAR 52.

01 · Source
Anonymized bid packages

US commercial bid packages, fully anonymized before aggregation. Sample size and period will be published with the final figures.

02 · Mix
Private, public and federal work

All CSI MasterFormat divisions, with extra attention to Divisions 03, 09, 22, 23 and 26.

03 · Standards
AIA A201, ConsensusDocs, EJCDC, FAR 52

Each finding cites the section and page in the source document and the standard clause it departs from.

04 · Review
Human review of sampled findings

A sample of findings is reviewed by people against the source documents before any figure is published. Reviewers will be named with the final results.

Illustrative preview

What will the results look like?

The table shows the patterns the study counts and how each will be reported. XX marks a placeholder, not a measured result.

Pattern Freq. Risk level Median exposure Cites against
Liquidated damages, no stated cap XX% HIGH $XXX,XXX AIA A101-2017 § 4.5
Broad-form indemnity covering owner negligence XX% MEDIUM Variable, uncapped AIA A201-2017 § 3.18
Scope gap between specs and drawings XX% HIGH $XXX,XXX CSI MasterFormat
Pay-when-paid, no time stipulation XX% MEDIUM XX-day cash flow gap State prompt-pay statutes
Performance and payment bond at 100% XX% STANDARD Expected, budget for it FAR 52.228-15
Retainage (retention) above 10%, no step-down XX% MEDIUM XX% of contract sum State retention caps

What should preconstruction teams check now?

Liquidated damages without a cap

Find the LD clause, the daily rate and the cap before you price. On AIA projects the rate sits in A101-2017 § 4.5, not in A201.

Dora flags LD clauses by default, checks whether a cap is stated, and cites the clause and page. See how Dora works.

Broad-form indemnity

Negotiate it down to the limited form A201-2017 § 3.18 uses. In Texas, Insurance Code chapter 151 voids most indemnity for the other party's own negligence; cite it in your redline.

Scope gaps between specs and drawings

Read specs against drawings division by division, then level sub bids to that scope. Nova flags the gaps no single sub bid shows.

Pay-when-paid with no time limit. Without a backstop, payment can stall while the owner disputes. In Texas, cite the Prompt Pay Act (Property Code chapter 28) in the subcontract redline.

Frequently asked questions

How is the dataset assembled?

The design uses anonymized US commercial bid packages from private, public and federal work. Identifying information is removed before analysis. Sources, sample size and period will be published with the final figures.

Why does the study count uncapped liquidated damages?

An uncapped LD clause can turn a thin margin into a loss when the schedule slips. It has to be caught before signature. Dora flags LD clauses by default.

Can we see the underlying data?

The full dataset is private and anonymized to protect contributor identities. Aggregate statistics, methodology and per-finding summaries will be published openly once the study is final.

Is this study peer-reviewed?

No. Findings are reviewed internally against the source documents before any figure is published. We welcome independent review and will publish corrections inline if methodology or findings need refinement.

How do you handle the data?

Customer documents are processed in tenant-isolated environments on US infrastructure, encrypted in transit and at rest. We do not train models on customer data. SOC 2 Type II audit underway. Research datasets are anonymized before any aggregation.

Which agents check for these patterns?

Each pattern maps to the agent that checks for it on your own documents.

See how the agents work together.