The version of this that keeps you up is a deadline you find out about late. A production is due Thursday. On Tuesday, someone realizes the custodian's mailbox was never exported, the shared-drive folder has three files in it instead of thirty, and the spreadsheet tracking all of it was last touched by an associate who is out on leave. The legal calls are the lawyers' to make. Right now, though, the status of the work lives in four places, and none of them agrees with the others.
If you run litigation support or manage matters at a firm handling more disputes than it used to, discovery tracking is where small gaps turn into real exposure. Requests come in from the other side. Custodians get named. Documents get collected, reviewed, produced, withheld, or explained. Objection windows close. Production dates move. The work is not hard to describe. It is hard to see all at once, across every open request, at the moment a deadline is bearing down.
This guide walks through giving every request one home: what it asks for, what the firm agreed to produce, whose files answer it, what has been collected, who owns the next step, and how close each clock is. The legal judgment stays with the lawyers. The tracking stops being a guess.
What a discovery tracker has to answer
A tracker earns its keep when it lets you answer, without opening five systems and calling three people, a short set of questions about any request in front of you:
- Which matter and issue does this request belong to, and what date range and document types does it actually cover?
- What did the firm agree to produce, and what did it object to?
- Whose files answer it, where those files sit, and how much has been collected so far?
- Who owns collection, who owns review, and who owns the privilege call?
- Which requests are blocked, by what, and how close is the next deadline?
The tracker makes the status honest and current. The lawyers still make the legal and strategic decisions. That split is the whole point: nobody is asking software to decide what to withhold or what to fight, only to stop the status of forty requests from living in someone's memory.
A quick way to test whether you have this today. Open one request three days from its production date and ask whether you can see its owner, its source, what is still missing, and whether anyone has made the privilege call, in under five minutes. If you have to reconcile a spreadsheet against a review platform against someone's inbox, the work is hiding, and it will surface at the worst possible time.
Follow one request from service to production
Before you build anything, trace one real request the way it actually moves through the firm today. The gaps are between the steps, not inside them.
- A request arrives from opposing counsel and gets logged, usually in a spreadsheet, sometimes just in an email folder.
- A lawyer works out what the request really covers: the date range, the custodians, the document types, and where it is broader than the firm will agree to.
- Responses and objections get served by the response deadline, and a scope gets settled, sometimes after a meet-and-confer.
- Someone identifies the custodians and where their files sit.
- Mailboxes, drives, chat exports, and file shares get collected.
- Documents go to a review platform for a relevance and privilege pass.
- A lawyer makes the privilege or confidentiality call on the flagged material.
- The set is produced with Bates numbers, withheld on a log, or explained, often in rolling batches.
Each handoff is a place where the trail can break. A custodian gets missed. A folder looks collected but is half empty. A privilege call sits in a comment nobody reopens. An agreed narrowing of scope never reaches the person doing the collection. The tracking exists to catch exactly these gaps before a deadline does.
Every request runs on two clocks
Most trackers record one deadline per request. Discovery usually has two, and they belong to different people at different moments.
The first is the response deadline: the date by which the firm has to serve its written responses and objections. Miss it, or serve boilerplate that does not preserve the right objection, and you can waive an objection you needed. That clock is a lawyer's to answer, but it is easy to lose because it runs early, while the collection work has barely started and attention is elsewhere.
The second is the production deadline, and in most matters it is not a single date. Productions roll. You agree to produce one custodian's documents by one date and another set two weeks later, and each of those is its own clock. A request can look healthy on the production side while its response window has quietly closed, or the reverse: the objections are clean but the documents are late. If the tracker shows one date per request, it is hiding one of the two clocks that can actually hurt you.
Scope is what you check everything against
After objections and any meet-and-confer, each request has an agreed scope: a date range, a custodian list, document types, sometimes negotiated search terms. That agreed scope is the thing you measure collection and production against. Without it written down, "we produced everything" means nothing, because nobody can say what "everything" was supposed to be.
This matters more in discovery than in most tracking work because the scope is contested and it moves. The other side asks for six years; you agree to three. They name ten custodians; you agree to four. A month later, a deficiency letter argues the four should have been six. If the agreed scope lives only in a lawyer's memory of a phone call, the person collecting cannot tell whether a source is in or out, and the person defending the production cannot show what was promised. Record the scope as a checkable definition, updated whenever it is renegotiated, and half the arguments about completeness resolve themselves.
Map each request to its custodians and sources
A request does not get satisfied by a folder. It gets satisfied by specific people's files in specific systems. So each request needs its custodians named and each custodian tied to where their material actually lives: a mailbox, a shared drive, a chat or messaging export, a ticketing or CRM system, physical files, or a third party who holds the records.
The failure here is quiet. A custodian gets identified in a kickoff call and then nobody is assigned to pull their mailbox. A shared-drive folder is marked collected because someone opened it, but it holds three files where the real material is thirty. A system with responsive records has no clean export path, so it keeps getting deferred until it is the last thing standing between you and the deadline. Naming the source, and marking whether it has actually been pulled, is what turns a list of custodians into something you can rely on.
Build one request tracker before you automate anything
The first thing to build is a tracker tied to the sources and the owners, not another status email that goes stale the moment it is sent. The difference between a status spreadsheet and a tracker is that a spreadsheet tells you what someone typed last, while a tracker shows what has happened, what still needs a lawyer's review, and which request is one missing export away from slipping.
At a minimum, each request should carry the matter and issue it belongs to, its agreed scope, the custodians and sources with working links, the state each source is in, the nearest of its two clocks, the owner of the next step, and whatever is blocking it. Here is what that looks like for a few requests inside one matter. The rows are illustrative, not a real client.
| Request | Source and custodian | Nearest deadline | State | Next owner | What is blocking it |
|---|---|---|---|---|---|
| Emails, one custodian, agreed date range | Custodian mailbox, not yet exported | Production, tight | Collection not started | Litigation support | No export scheduled |
| Signed contract versions | Document management folder | Production, comfortable | In review | Associate | Privilege pass not done |
| Meeting notes and attachments | Shared drive and notes app | Objection deadline, this week | Scope not settled | Matter lead | No source linked yet |
Notice what the table forces into the open. One request has a production date and no collection started. One is collected but stuck on a privilege call. One has an objection deadline this week and no source even identified. That last one is the dangerous row, and on a plain status list it would look identical to the others.
A worked example: forty requests across three matters
Say a litigation team is handling forty outstanding requests spread across three matters, with a rolling production on Friday and a hard production on the following Thursday. This scenario is invented to show the shape of the problem, not drawn from a real client.
On Monday, the picture in the spreadsheet is a wall of rows, all of them technically "in progress." That tells the team nothing about where to spend the week. The useful move is to stop looking at the forty as a list and group them by what is actually in the way, then read the nearest clock for each group.
| What is blocking them | Requests in this state | Nearest clock | First move this week |
|---|---|---|---|
| Waiting on collection | 11 | Production in nine days | Schedule the two mailbox exports that hold most of it |
| Waiting on review | 14 | Rolling production Friday | Assign reviewers to the oldest batch first |
| Waiting on a privilege call | 6 | Production in four days | Get the flagged set in front of the partner |
| Waiting on the other side | 5 | Meet-and-confer Thursday | Confirm the narrowed date range in writing |
| Ready to produce | 4 | Due Thursday | Bates-stamp and send |
Grouped this way, the week plans itself. The six requests waiting on a privilege call are the real fire, because their clock is four days and only a lawyer can clear them, so they go in front of the partner today rather than Wednesday. The eleven waiting on collection have nine days, so they can wait a beat, but the two big mailbox exports get scheduled now because exports take time you do not control. The five stuck on the other side are not the firm's to move alone and should not eat the team's attention until Thursday's meet-and-confer settles the scope. Same forty requests, but now you know which one to touch first and which ones are safe to leave.
Track the objections and agreements, not just the documents
A lot of what determines whether a production is defensible never appears on a document list. It lives in the negotiation: what the firm objected to, what the other side pushed back on, what got agreed, and what is still in dispute and possibly headed for a motion to compel. That trail usually sits in emails and meet-and-confer letters, which means it is invisible to the person doing the collection.
When the negotiation is not tracked next to the request, the scope drifts silently. A custodian gets added in a phone call and never collected. A date range gets narrowed and the reviewers keep reading documents that are now out of scope. The team produces on the old understanding and gets a deficiency letter for its trouble. Keeping the meet-and-confer outcome attached to the request, with the current agreed scope and any open dispute flagged, is what keeps the collection and the negotiation from telling two different stories.
Keep collected, reviewed, and produced apart
Three words get used as if they mean the same thing, and conflating them hides most of the remaining work. "Collected" means the material has been pulled from its source. "Reviewed" means a human has made a relevance and privilege pass over it. "Produced" means it has been Bates-stamped and sent to the other side. A request can be fully collected and nowhere near produced.
The trap is a status field that says "in progress" or, worse, a green checkmark that means "we have the documents" being read as "we are done." The gap between collected and reviewed is where the real hours are, because review is slow and it is where privilege gets caught. If the tracker cannot show which of the three states each source is in, the team keeps discovering late that "collected" was never "reviewed," and the review it still owes will not fit in the days that are left.
Sort the open requests by what is blocking them
A long list of open requests tells you nothing about where to spend the afternoon. What helps is grouping them by what stands in the way, the same move that made the worked example legible. The buckets are stable across most matters:
- Waiting on collection: the mailbox, drive, or custodian has not been pulled yet.
- Waiting on review: the documents are in, a reviewer has not read them.
- Waiting on a privilege or confidentiality call: flagged, and it needs an accountable lawyer.
- Waiting on the client or a third party: you have asked, the response is pending.
- Waiting on the other side: the scope or a dispute is unresolved.
- Ready to produce: reviewed, cleared, and it can go out.
Grouped this way, a Monday-morning look tells you which blocker to clear first and which deadlines are safe to leave alone until later in the week. It also tells you which blockers are yours to move and which depend on a lawyer, the client, or opposing counsel, so you stop spending energy on the ones you cannot unblock alone.
Keep a production record that ties Bates ranges to requests
The question that ends a lot of discovery disputes is simple: did you actually produce what you said you would? Answering it means keeping a production record that maps Bates ranges back to the requests they satisfy, with the date each set went out and what remains outstanding under a rolling schedule. Without it, a deficiency letter turns into a scramble through old production logs to reconstruct what was sent when.
| Request | What was agreed to produce | Bates range produced | Production date | Still outstanding |
|---|---|---|---|---|
| Custodian A emails, agreed date range | All non-privileged material in range | ABC000001 to ABC004120 | First rolling set | None |
| Vendor contracts | Executed versions and amendments | ABC004121 to ABC004890 | First rolling set | Two late-signed amendments |
| Board materials | Minutes and decks in agreed range | Not yet produced | Scheduled, next set | Privilege review in progress |
This record does double duty. It is your answer when the other side claims a gap, and it is your own early warning that a rolling production has an open item with no scheduled date. When the withheld material goes on a privilege log, the same record shows what was produced against what was held back, so the two views cannot drift apart.
Where AI helps inside discovery tracking
Once the tracker, the scope, and the states are clear, AI can take the reconciling and summarizing work off your team. It is genuinely useful for reading a request and pulling out its date range and document types, classifying the request type, pulling the response and production deadlines out of the letters, grouping near-duplicate requests that ask for the same thing in different words, flagging a custodian with no source linked, and drafting an internal status note so the Monday update writes itself.
It can also draft first-pass entries, a privilege log line or a status summary, from firm-approved inputs, with a link back to the source every time, so a reviewer is checking a draft against a document rather than starting from a blank page. The rule that keeps this safe is that AI only ever works from approved sources and always shows its link back, so nothing it produces is trusted until a person has seen where it came from.
Where the legal judgment stays
Everything in the last section is extraction, tracking, and drafting. None of it is a legal decision, and the line between the two has to stay bright. AI does not decide what to object to, what is privileged, what is confidential, what to withhold, or what is worth fighting in a motion. Those calls belong to an accountable lawyer, every time, and this guide is not legal advice about how to make them.
The value of getting the tracking right is that it protects the lawyer's attention for exactly those judgments. When the status is honest and current, the privilege reviewer is not spending an hour figuring out which set is even ready to review before they can start. The system keeps the status straight so the lawyer's time goes to the calls only a lawyer can make.
How the tracking breaks
The failures are consistent enough to name, and naming them is usually enough to design around them.
One spreadsheet, one owner
When the tracker is a spreadsheet only one person updates, the truth walks out the door when they take leave. Everyone else is guessing until they are back.
Collected blurred with reviewed
A single "done" state hides the gap between having the documents and having read them. The review you still owe is invisible until the days to do it are gone.
A privilege call parked in a comment
A flagged item with the privilege decision sitting in a comment nobody reopens before production is the classic way privileged material either slips out or holds up a whole set at the deadline.
The response deadline tracked nowhere
When only the production date is visible, the response and objection window closes unwatched, and an objection gets waived that the firm needed to preserve.
The production record living only in the vendor tool
If the only place that ties Bates ranges to requests is the review platform, a deficiency letter turns into archaeology, and rolling productions lose track of their own open items.
AI added with nowhere to check it
An extraction or a draft that no one is assigned to verify is not a time saver. It just produces confident-sounding status faster than anyone can confirm it is true.
What to watch while there is still time
A few signals tell you the tracking is slipping while you can still do something about it. Watch for requests with no source linked yet, requests whose owner has not changed in a week, the days of slack left before each response deadline and each production deadline separately, requests reopened because a document or a note went missing, and open items in a rolling production that have no scheduled date. None of these is a status report anyone reads for pleasure. Each one is a specific gap you can close before it becomes a filing.
Rolling out the first matter
Do not try to track every matter at once. Start with one, ideally one where the tracking already scares you a little, so the first build is solving a real problem rather than a hypothetical one.
- Pick one matter, or one request set inside a matter.
- List the requests, the agreed scope for each, the custodians, where their files sit, and both clocks: the response deadline and the production dates.
- Build the tracker fields and the three states, so collected, reviewed, and produced are never the same box.
- Connect only the systems the tracker needs to stay current: the document store, the review platform, the shared drives, and the calendar.
- Add one AI-assisted step, such as pulling deadlines out of the request letters, with a person checking every result.
By the end of a few weeks on that one matter, you should be able to open the tracker and see every request, its owner, its source, and how close each deadline is, without asking three people first. Once it holds up under one real matter, the pattern moves to the next one without being rebuilt.
How Ubisar would build this with you
In week one, we would pick one matter or request set with you and map it end to end: the requests, the agreed scope, the custodians, where the documents sit, both sets of deadlines, who owns collection and review, and what is blocked. The first thing you would have is a tracker showing each request's scope, source, owner, nearest deadline, state, and what is missing.
In the following weeks, we would connect only what the tracker needs to stay current: your matter system, the review platform, the shared drives, and the calendar. AI would help summarize the requests, classify issue types, pull deadlines out of the letters, and flag stale owners and unlinked sources, while the objection, privilege, and production calls stay with your lawyers. By the end of the first month, your matter team should be able to run a live request review from one tracker and know, without a scramble, which deadline is closest and which blocker to clear next. It feeds straight into client status reporting when the client wants a regular update. Keep going if owners, sources, and deadlines are visible earlier than they used to be; narrow it if request definitions or source access are still being settled.
When you want a hand building it, that is the whole of our implementation service, starting from $4,000/month, cancel anytime. Send us the one matter where the tracking scares you most through our contact page, and we will map out the first month with you.