Production Errors Are Sanctions Waiting to Happen
Courts have shown an increasing willingness to impose severe sanctions for production errors in eDiscovery. Adverse inference instructions, fee-shifting, and even default judgments are all on the table when a party fails to produce documents in compliance with its discovery obligations. The consequences are not hypothetical — they happen regularly, and they happen to sophisticated parties represented by experienced counsel.
The most common production errors fall into a handful of categories that are entirely preventable with proper quality control. Bates numbering gaps and duplicates create confusion about the completeness of a production and can suggest spoliation. Visual-only redactions — where the redaction appears on screen but the underlying text remains selectable and searchable — expose privileged and confidential information. Privilege leaks, where privileged documents are inadvertently produced, trigger clawback obligations and potential waiver arguments. Metadata mismatches between load file fields and the actual document properties undermine the reliability of the entire production.
Consider the Federal Production Remediation case that DecoverAI handled: all six of these failure categories were present in a single production. The producing party had to remediate over 360,000 documents under federal court scrutiny, at enormous expense and with significant reputational damage. Every one of those errors could have been caught with a systematic pre-production QC process.
The checklist that follows is designed to be run before every production, regardless of size. Small productions deserve the same rigor as large ones — a single privilege leak in a 500-document production can be just as damaging as one in a 500,000-document production. Print this out, assign responsibility for each section, and do not authorize a production until every item has been verified.
Checklist: Bates Numbering Verification
Bates numbering is the backbone of any production. It provides the unique identifier that allows every party, the court, and witnesses to refer to specific documents without ambiguity. When Bates numbering fails, the entire production becomes unreliable, and courts take notice. Gaps in Bates ranges suggest missing documents, while duplicates create confusion about which version is authoritative.
Start by verifying that the Bates numbers are sequential with no gaps and no duplicates. Export the full Bates range from your review platform and sort numerically. Any gaps should be investigated — they may indicate documents that were removed from the production set without the numbering being regenerated, or they may indicate a processing error. Duplicates are equally problematic: two documents with the same Bates number make it impossible to cite either one with confidence.
Verify that the digit count is consistent across the entire production. If your numbering convention is ABC-000001, every document should follow that pattern. Mixed formats (ABC-1, ABC-00001, ABC-000001) create sorting problems and suggest a lack of process control. Check that the cover letter or production notice matches the actual Bates range stamped on the documents. It is surprisingly common for the cover letter to state one range while the documents contain a different range, particularly when productions are regenerated after errors are discovered.
Ensure that family groups are sequential — a parent email and its attachments should have consecutive Bates numbers. When family members are scattered across the Bates range, it becomes difficult to understand document relationships and can lead to incomplete privilege designations. As a practical matter, sample at least 5% of the production or 100 documents, whichever is greater, to visually confirm that the Bates stamp appears correctly on each page, is legible, and is positioned consistently.
Checklist: Redaction Verification
Redaction failures are among the most damaging production errors because they directly expose the information you were trying to protect. The distinction between a data-layer redaction and a visual-only redaction is critical. A visual-only redaction places a black box over text on the rendered page, but the underlying text remains in the PDF content stream. Anyone who selects the text, searches the document, or opens it in a text editor can read the "redacted" content.
For every redacted document in your production, attempt to select the text behind each redaction. Open the PDF, click on the redacted area, and try to drag-select. If you can highlight text beneath the black box, your redaction has failed. This is not a theoretical concern — it is the single most common redaction error, and it has led to sanctions, malpractice claims, and waiver of privilege in numerous reported cases.
Verify that every redacted PDF has been properly flattened. Flattening removes the annotation layer and bakes the redaction into the document content, making it irreversible. An unflattened PDF with redaction annotations can have those annotations removed by anyone with a PDF editor, fully exposing the redacted content. Most modern redaction tools flatten automatically, but this should be verified rather than assumed.
Finally, confirm that no unredacted native versions have been produced alongside the redacted image versions. This is the error that occurred in the Federal Production Remediation matter: redacted TIFFs were produced, but the corresponding native files — containing the full unredacted text — were included in the same production package. Check your production directory structure and load file to ensure that only the redacted versions are referenced and included.
Checklist: Privilege Log Cross-Reference
A privilege log that does not match the actual documents withheld from production is an invitation for a motion to compel. Courts expect a one-to-one correspondence between withheld documents and privilege log entries, and discrepancies in either direction create problems. Every document withheld on privilege grounds must appear on the privilege log, and every entry on the privilege log must correspond to an actually withheld document.
Check for orphaned entries — privilege log entries that reference documents which were actually produced. This happens when documents are initially designated as privileged during review but are later redesignated as non-privileged and produced, without the privilege log being updated. Orphaned entries suggest sloppy process and can lead to confusion about what was actually withheld.
Equally important, check for missing entries — documents that were withheld from production but do not appear on the privilege log. This is the more dangerous error, because it means you have withheld documents without providing any basis for the withholding. Courts may order production of these documents or draw adverse inferences about their content. Run a comparison between your production Bates range and your privilege log Bates range to identify any documents that fall in neither set.
Review the privilege log descriptions to ensure they are specific enough to support the claimed privilege. Generic descriptions like "Email re: legal matter" are routinely rejected by courts. Each entry should identify the author, recipients, date, and a description sufficient to establish the elements of the claimed privilege without revealing the privileged content. If your descriptions are too vague, you will face a log challenge, and you may lose the privilege designation for hundreds of documents at once.
Checklist: Metadata and Load File Validation
The load file is the roadmap to your production. It tells the receiving party how to ingest your documents into their review platform, and it contains the metadata fields that make the documents searchable and sortable. A load file with errors can render an entire production unusable or, worse, misleading. Metadata validation should be a mandatory step before any production is finalized.
Verify that all required metadata fields are present and populated. At a minimum, this typically includes document date, author/sender, recipients, subject/title, custodian, file type, and Bates range. Check the ESI protocol or court order governing the production to confirm which fields are required. Missing fields should be investigated — they may indicate processing errors or gaps in the source data that should be documented.
Confirm that date formats are consistent and correct across the entire load file. Mixed date formats (MM/DD/YYYY in some rows, YYYY-MM-DD in others) create sorting and filtering problems for the receiving party. Check for obviously incorrect dates — dates in the future, dates before the relevant time period, or dates of 01/01/1970 (which typically indicate a metadata extraction failure). Custodian fields should be accurate and consistent: the same custodian should be identified the same way throughout the load file, without variations in spelling or formatting.
Check for encoding issues that can corrupt text fields, particularly in productions that include documents in multiple languages or documents with special characters. Open the load file in a text editor that displays encoding and verify that the file encoding matches what your load file specification requires (typically UTF-8 or ANSI). Finally, verify that all file paths in the load file are valid and point to files that actually exist in the production package. A load file that references files in directories that do not exist will cause ingestion failures for the receiving party.
Checklist: Native File and Image Quality
The final QC checkpoint focuses on the files themselves. Processing and conversion errors can produce files that are corrupted, truncated, or unreadable, and these problems are often not apparent from the load file or Bates numbering alone. A systematic check of file integrity is essential before any production is released.
Verify that all native files open correctly in their associated applications. Spot-check a representative sample by actually opening documents in Word, Excel, PowerPoint, and other native applications. Files that fail to open may have been corrupted during processing or export. Check that files are not truncated — compare file sizes in the production to file sizes in the source collection to identify any significant discrepancies that might indicate incomplete exports.
For image productions, confirm that TIFF resolution is sufficient for readability. The industry standard is 300 DPI for black-and-white TIFFs, but some platforms default to lower resolutions that produce blurry or illegible images. Open several TIFFs at 100% zoom and verify that the text is clear and legible. Check that OCR text has been properly extracted and included in the production — without OCR, image-only productions are unsearchable, which may not comply with your production obligations.
Finally, verify consistency between native and image versions of the same document. If you are producing both natives and images, the content should match. Page counts should be the same, and the text content of the image (as captured by OCR) should correspond to the text content of the native. Discrepancies may indicate that the wrong version of a document was converted, or that the native was modified after the image was generated.
How DecoverAI Prevents Production Failures
DecoverAI builds automated quality control into every stage of the production workflow, so that the errors described in this checklist are caught before they reach opposing counsel or the court. Rather than relying on manual spot-checks after a production is assembled, DecoverAI runs continuous validation throughout processing, review, and production, flagging issues in real time so they can be corrected before they compound.
The platform's Bates numbering engine enforces sequential numbering with no gaps or duplicates by design. Redactions are applied at the data layer and automatically flattened, eliminating the possibility of visual-only redactions reaching a production. Privilege log entries are generated and cross-referenced against production and withholding decisions in real time, so orphaned and missing entries are flagged immediately rather than discovered during a log challenge.
In the Federal Production Remediation matter, DecoverAI remediated over 360,000 documents that had been produced with errors by a previous vendor. The remediation included correcting Bates numbering, replacing visual-only redactions with data-layer redactions, reconciling the privilege log, validating metadata, and repackaging the entire production — all under federal court deadlines and scrutiny. The fact that those errors existed in the first place demonstrates why automated QC is essential.
Every DecoverAI engagement includes an embedded LegalOps specialist who performs a final manual QC review before any production is released. This human review layer supplements the automated checks, catching edge cases and context-dependent issues that automated tools may not flag. The combination of automated validation and expert human review ensures that productions meet the highest standards of accuracy and defensibility.