Discovery Document Management When the Volume Gets Real

There is a moment in most contested matters where the document population stops being something a person can hold in their head. Up to a few hundred pages, a competent associate can tell you what is in the file, roughly where it came from, and which items nobody outside the firm should ever see. Past a few thousand, that stops being true, and the firms that get hurt are the ones who do not notice the moment it stops being true. They keep working the same way, with folders named after whoever collected them, privilege calls made in someone's memory, and productions assembled the night before they are due.

The damage almost never shows up as a dramatic failure. It shows up as a partner unable to answer a simple question during a meet and confer about whether a particular batch of emails went out in the second production or the third. It shows up as the same document appearing under two different Bates ranges because two people numbered two overlapping collections. It shows up as a privileged internal memo sitting in a production set because the reviewer who flagged it wrote the flag in a spreadsheet that was superseded a week later. None of those are exotic disasters. They are bookkeeping failures, and bookkeeping failures at volume are what discovery sanctions and malpractice claims are actually built from.

What follows is not a review platform pitch. Most firms outside the largest litigation shops are handling discovery volume inside their practice management system, their document storage, and whatever review tooling the matter can justify. The question is how you structure the work so the record is defensible regardless of which tools you assembled. That structure has four load-bearing parts: how documents enter, how they get numbered, how privilege is decided and recorded, and how production sets are built and remembered.

The Volume Problem Is Not About Storage

Storage is the cheapest part of discovery and it is where firms spend most of their attention. Someone worries about whether the cloud drive has capacity, whether the scans are good enough, whether the PDFs open on the paralegal's laptop. Meanwhile the actual failure mode is that nobody can reconstruct, six months later, which files came from which custodian, on which date, through which method, and what happened to each one between arrival and production. Capacity is not the constraint. Provenance is.

The practical test is uncomfortable and worth running on your current matter. Pick a document from the middle of your largest active file and try to answer four questions without asking a colleague: who gave it to us, when, was it reviewed for privilege and by whom, and has it been produced. If you cannot answer all four from the system itself in under a minute, your discovery process is running on institutional memory. Institutional memory is fine at fifty documents and negligent at five thousand, and the transition between those two states happens quietly, usually in the middle of a matter that started small.

Intake Is Where Discovery Is Won or Lost

Every document that enters the matter should be captured with its origin attached at the moment of entry, not reconstructed later. That means the custodian or source, the date and method of collection, the person at the firm who received it, and the original filename and folder path as it existed on the client's side. The last one gets skipped constantly and it matters enormously, because the folder structure a client used is often the single best evidence of what the client understood a document to be. Strip it during intake and you cannot get it back.

The discipline that makes this stick is refusing to accept a batch that is not documented. If a client's operations manager forwards forty attachments with no context, the answer is not to file them and move on. It is to send back a short note asking which system they came from and who pulled them, and to hold the batch until that answer exists. Firms hate this because it feels like friction with a paying client. It is far less friction than a deposition where your own client is asked where a document came from and the answer is that the law firm does not know either. In Casely, every document carries a comment field recording what changed and why, which is exactly the field intake notes belong in, and the AES-256 encryption sits on the document with a per-firm key so the provenance record and the file itself stay together under the firm's own control.

Bates Numbering Is a Commitment, Not a Label

A Bates number is the only stable identity a document has for the rest of the matter's life. Everything downstream refers to it: deposition exhibits, motion citations, expert reports, the privilege log, the trial binder. That makes the numbering scheme a decision you make once, deliberately, before the first production, and never revisit. The failure pattern is firms that treat numbering as a formatting step applied at production time, which guarantees that internal work product, correspondence and pleadings all reference documents by names and paths that will not survive.

Number at intake, not at production. Assign the range as documents enter the matter, with a prefix that identifies the producing party and the matter, and keep the sequence unbroken and firm-wide rather than per-custodian. A firm-wide sequence means a number is never ambiguous, which is the entire point. If you number per custodian and later discover two custodians shared a mailbox, you now have a reconciliation problem that costs real hours. Keep the numbering register in the matter itself so anyone can see the last number issued, and treat that register as the authoritative source rather than whichever spreadsheet is open on someone's screen.

Privilege Tagging as You Go, Never at the End

The single most expensive habit in small-firm discovery is deferring privilege review to a dedicated pass before production. It feels efficient because it concentrates the work, and it is catastrophic because it concentrates the risk. A reviewer working through eight thousand documents against a deadline makes different decisions in hour nine than in hour one, and the marginal calls, the ones involving a partially privileged attachment or a communication copied to a business contact, are precisely the ones that get resolved by fatigue rather than judgment.

Tag at first touch instead. Whoever opens a document for any purpose records a privilege status before closing it, using a fixed vocabulary the whole team agreed on before the matter started. Something like privileged, work product, partially privileged with redactions needed, not privileged, and unresolved. That last category matters more than the others because it is honest. A junior reviewer who is unsure should be able to say so in the system rather than guessing, and unresolved items should surface as a queue a supervising attorney clears deliberately. The specific scope of privilege and work product protection differs meaningfully between the United States, England and Wales, Canada and Australia, particularly around in-house counsel communications and litigation-anticipated material, so build the vocabulary against your own jurisdiction's standard and confirm it locally rather than importing a checklist from somewhere else.

The Privilege Log Is Built From the Tags, Not From Memory

If tagging happens at first touch and the tag carries the reviewer's name, the date, and a one-line basis, the privilege log is a report rather than a project. That is the whole return on the discipline. The log needs enough detail for the other side to assess the claim without disclosing the protected content itself, and jurisdictions differ on how granular that has to be. Some courts expect document-by-document entries with author, recipients, date and a description; others accept categorical logging for defined groups of material. Check what your court actually requires before assuming, because the difference between those two standards is weeks of work.

The one-line basis is the part firms skip and the part that saves them. Writing "advice from outside counsel re: indemnity clause" at the moment of review takes fifteen seconds and produces a defensible log entry. Reconstructing that same basis four months later, from a tag that says only "privileged", means reopening the document, rereading it, and hoping the reviewer's reasoning is recoverable from the text. Multiply that across a few hundred logged items and you have built a second review at full cost, which is the outcome the tagging was supposed to prevent.

Production Sets Are Records, Not Exports

Most firms treat a production as an action. Files get gathered, numbered, stamped, converted and sent, and then the matter moves on. The correct mental model is that a production is a permanent record of a state: this specific set of documents, in this specific form, went to this specific party on this specific date under this specific agreement. That record needs to persist inside the matter with the same weight as a filed pleading, because it will be cited, challenged, and supplemented, and every one of those events depends on knowing exactly what the earlier set contained.

Give each production a name and a number, record the recipient, the date, the Bates range or ranges included, the format produced, and whether it was made under a protective order or confidentiality designation. Then freeze it. A production set that can be quietly edited after the fact is worse than no record at all, because it looks authoritative while being unreliable. This is the same reasoning behind Casely voiding corrections rather than deleting them in trust accounting, where a reversed entry stays visible with its reversal attached. The principle transfers directly: in a system of record, the correction is part of the record, never a replacement for it.

Knowing What Was Produced to Whom

Multi-party matters are where production tracking stops being administrative and starts being substantive. In a case with a plaintiff, three defendants, an insurer and an expert under a protective order, the same underlying document may go out four times in four different forms, redacted differently for each recipient according to different designations. If your record is a folder called "Production 3 FINAL", you cannot answer which version the expert received, and you will eventually be asked.

The workable structure is to treat recipient as a first-class field on the production record rather than something implied by the folder name. Every set names its recipient explicitly, and where the same document went out in different redacted forms, each form is its own numbered item tied to its own set. Related matters compound this further, and Casely's connected matters feature exists for exactly this shape of problem, linking related files with the reason for the link stated so a coverage action and the underlying claim can be reasoned about together without collapsing their document populations into one undifferentiated pile.

Chain of Custody From Collection to Production

Defensibility is not a document you produce when challenged. It is a property of the record you kept while the work happened. A chain that holds up shows an unbroken line from the moment material left the client's control to the moment it left yours, with each transition attributable to a named person on a known date. Nothing about that requires expensive tooling. It requires that every movement of a document be a logged event rather than a drag between folders.

The specific pressure point is conversion and processing. Documents get flattened to PDF, images get OCR'd, spreadsheets get printed to static form, and each of those steps changes the artifact. If the original is not preserved alongside the processed version, and the processing step is not recorded, you have a gap that opposing counsel can widen. Keep the native file. Record what was done to produce the version that went out. The comment field on every Casely document is the natural home for that note, and because the encryption key is per-firm, the processing history lives under the firm's control rather than in a vendor's undifferentiated pool.

Ethical Walls Inside a Document Population

Discovery is where ethical walls get tested, because the volume creates pressure to pull in whoever is free. A contract reviewer brought on for two weeks, a paralegal borrowed from another team, an associate who previously worked on a matter adverse to one of the parties: each is a wall question, and each is a question the firm usually answers with an instruction rather than a control. Telling someone not to open a matter is not a wall. It is a hope.

A wall that means anything is enforced where the data is served, not in the interface. If a walled reviewer runs a global search across the firm's documents, the restricted material must not appear in the results, must not appear on a shared calendar entry, and must not open from a link a colleague pastes into a message. Casely enforces walls at the server and data-access layer for exactly this reason, so a walled user genuinely cannot reach a restricted matter by any path. Under a deadline, with a large document set and a temporary team, the difference between a UI-level wall and a server-level one is the difference between a policy and a fact.

Deadlines, Stages and the Rolling Production Problem

Discovery deadlines rarely arrive as a single date. There is a substantial completion date, a set of rolling production milestones that may exist only in a stipulation between counsel, a privilege log deadline that often trails the production it describes, and supplementation obligations that persist as new material surfaces. Firms that track only the court-ordered date miss the informal commitments, and the informal commitments are what erode credibility with the other side long before they become motion practice.

Attach every one of those dates to the matter itself rather than to an individual's calendar, and let the next date auto-track so that completing one milestone surfaces the following one. Casely's deadline diary is built around exactly that attachment, and the matter stage tracker gives the discovery phase a visible position so anyone picking up the file can see whether the matter is in collection, review, production or supplementation without asking. On a matter running eighteen months across three attorneys, that visibility is not a convenience. It is how the file survives a staffing change.

Getting There From Where You Are

If you are mid-matter with a population that is already disordered, do not attempt a full retroactive cleanup. Draw a line at today. Everything entering from this point gets intake metadata, a Bates range from a single register, and a privilege tag at first touch. Then work backwards selectively, starting with the documents most likely to be cited: anything already produced, anything referenced in a pleading, anything on a witness list. Most of the population will never need reconstruction, and the discipline of not trying to fix all of it is what makes fixing the important part achievable.

The structural point underneath all of this is that discovery bookkeeping is not a specialist function that appears when volume arrives. It is the same set of habits that make a matter legible at any size, applied when the stakes make them mandatory. Provenance at intake, one numbering register, tags with authors, frozen production sets, and a wall you can actually rely on. Firms that build those habits at fifty documents do not have a crisis at five thousand, because there is nothing to change except the pace.

None of it requires a discovery platform to begin. It requires that your legal document management software treat a document as a record with a history rather than a file in a folder, and that the matter system holding the deadlines, the walls and the production log be the same system holding the documents themselves. Casely is cloud-native with a free plan, so the honest way to test this is to structure one live matter the way described here and see whether the questions that used to require a phone call now answer themselves.