Processing. Cull hard, before a lawyer reads the first document.
Review is where budgets are spent. Processing is where budgets are made. The right work at this phase can compress a million-document dataset into a review population an order of magnitude smaller, defensibly.
- Phase
- 04 of 06
- Discipline
- Processing
- Standard
- ACPO · ISO 27001
Plate · ProcessingCompression, without compromise.
Every reduction step is defensible. Deduplication is family-aware. Threading preserves the underlying messages, they are suppressed for review, not deleted. Search-term work is negotiated with counsel and, where relevant, with the other side. Nothing is discarded. Suppressions are reversible, and the audit trail shows exactly why any document is not in the review population.
Processing takes the collected material and prepares it for human review. De-duplication removes identical copies across custodians. Email threading collapses long chains into their inclusive messages. Near-duplicate detection groups minor variants. OCR turns image-only PDFs into searchable text. Language identification routes non-English material to the correct reviewer.
Done well, processing is the difference between a review that finishes on schedule and one that overruns by a factor of three.
How processing runs on our matters.
- 01
Ingest & indexing
Native and container files decompressed, embedded objects extracted, metadata normalised, full-text indexed. Processing exceptions logged and addressed rather than silently dropped.
- 02
De-duplication
Global or custodian-level deduplication using MD5 and content-based hashing, applied family-aware so that a duplicate parent does not orphan its attachments.
- 03
OCR & language ID
Image-only content OCR'd with quality thresholds recorded. Language identification per document routes non-English material to the appropriate review team.
- 04
Threading & near-dupe
Email threading identifies inclusive messages. Near-duplicate detection groups variants so that a reviewer sees one canonical version and applies coding to the group.
- 05
Search terms & filters
Date filters, custodian filters, domain filters and negotiated search terms applied with hit reports at each step. Terms iterated against a sample rather than run once and left.
- 06
Analytics & predictive coding
Concept clustering, communication analysis and, where the volume justifies it, technology-assisted review to prioritise likely-relevant material and de-emphasise noise.
What processing produces.
What arrives at the end of the phase, in a form the next phase, or a court, can rely on.
- Processed, deduplicated and threaded review population
- Hit reports per search term with counts before and after culling
- Processing exception report, with each item resolved or accepted
- OCR quality report for image-heavy custodians
- Analytics workspace ready for review to begin
- Documented processing protocol, defensible to the other side
Where processing goes wrong.
The recurring failure is running processing as a black box, tools with default settings, hit reports never inspected, exceptions written to a log no one reads. The document population reaches review inflated by duplicates, missing OCR text for whole custodian batches, and search terms that were never tested against a sample. The remedy is not more tooling. It is treating processing as a phase that produces its own outputs, reviewed by a human before review begins.
Continue through the workflow.
Bring us in early.
Defensibility is built, not retrofitted.
Whether you are responding to a regulator, preparing for disclosure, or scoping an internal investigation, start the chain of custody with a short, confidential conversation.