eDiscovery, short for electronic discovery, is the process of identifying, preserving, reviewing, and exchanging digital evidence during a lawsuit. Nearly every business record now lives in electronic form, so almost every modern case involves sorting through emails, chat logs, databases, and other digital files to decide what each side is entitled to see. The process is governed primarily by the Federal Rules of Civil Procedure, and getting it wrong can result in sanctions severe enough to lose a case outright.
What Counts as Electronic Evidence
Electronically stored information, or ESI, covers any data created or stored on a computer or electronic device. Federal Rule of Civil Procedure 34(a) defines the scope broadly to include documents, images, sound recordings, and “other data or data compilations” stored in any medium.1Legal Information Institute. Federal Rules of Civil Procedure Rule 34 – Producing Documents, Electronically Stored Information, and Tangible Things In practice that means emails, text messages, word processing files, spreadsheets, databases, social media posts, voicemails, and cloud-stored files all qualify.
A key difference between paper and ESI is metadata, the embedded information about a file that you typically can’t see by opening it. Metadata records details like when a file was created, who last edited it, and the exact time an email was sent. A printed document shows only the final product; the electronic version reveals the full history. That history can be decisive in disputes over whether someone altered a contract or backdated a message.
The definition has expanded well beyond office files. Data from Internet of Things devices, including smart home speakers, GPS trackers in fleet vehicles, and wearables like Fitbit and Apple Watch, is increasingly showing up in discovery requests. A pacemaker’s log can contradict testimony about physical activity, and a vehicle’s GPS data can place someone at a specific location at a precise time. Courts treat sensor readings, location data, voice commands, and usage logs from these devices as discoverable ESI.
The Stages of the eDiscovery Process
The widely adopted roadmap is the Electronic Discovery Reference Model, or EDRM, which breaks the work into nine stages: information governance, identification, preservation, collection, processing, review, analysis, production, and presentation.2EDRM. EDRM Model A small contract dispute might barely touch analysis; a large antitrust case can spend months in review alone. The model is a framework, not a checklist.
Information governance happens before any lawsuit exists. Organizations that maintain clear data retention policies, know where their information lives, and can quickly identify relevant custodians spend far less time and money when litigation arrives.3EDRM. Information Governance Reference Model
Once a case is underway, identification locates potential sources of relevant data, preservation protects them from alteration or deletion, and collection extracts the data without changing it. For high-stakes matters, collection often means forensic imaging: a bit-for-bit copy of a storage device created with hardware that prevents anything from being written back to the original drive.
Processing then converts files into formats review software can handle and reduces volume through de-duplication, date filtering, and keyword culling. The reduction is often dramatic. A collection of 500 gigabytes might shrink to 50 after duplicates, system files, and clearly irrelevant material come out. Review and analysis are where attorneys actually read the surviving documents, and production delivers the responsive ones to the other side. Presentation covers the use of the evidence at depositions, hearings, and trial.
When the Duty to Preserve Starts
The duty to preserve relevant evidence attaches the moment a party reasonably anticipates litigation, not when a complaint is filed. The test is objective: would a reasonable organization in the same position have expected a lawsuit? For a plaintiff, the answer is almost always yes well before filing. For a defendant, it might be the day a demand letter arrives or a regulatory investigation begins.4Legal Information Institute. Federal Rules of Civil Procedure Rule 37 – Failure to Make Disclosures or to Cooperate in Discovery
Once that duty attaches, the organization must issue a litigation hold: a written notice to employees who might possess relevant data, directing them to stop routine deletion, hardware recycling, or backup overwriting. The hold should name specific data types, systems, and custodians, and give a contact for questions. A vague one-size-fits-all notice won’t cut it.
Ephemeral Messaging and Auto-Delete
Holds that mention only “email and electronic documents” are increasingly treated as inadequate. Apps like Signal, WhatsApp, and Slack with disappearing-message features are common business tools, and courts expect holds to address them explicitly. If a hold is in effect, custodians must disable auto-delete on these platforms. Courts have imposed sanctions even where companies sent reminders about auto-delete settings, holding that organizations have an affirmative duty to monitor compliance rather than simply issue a notice and hope.
In 2025, the Seventh Circuit affirmed the dismissal of a case where the plaintiff intentionally deleted Signal messages discussing the litigation. Federal agencies including the DOJ and FTC have updated their preservation letters to warn that destroying ephemeral messages can trigger spoliation sanctions or obstruction of justice charges. Signal’s design makes ignorance hard to claim, since changing the disappearing-message setting is a deliberate act that notifies everyone in the conversation.
Sanctions for Spoliation
Rule 37(e) spells out what happens when ESI that should have been preserved is lost because a party didn’t take reasonable steps to protect it and the data can’t be recovered through other means. The consequences depend on the party’s mental state.4Legal Information Institute. Federal Rules of Civil Procedure Rule 37 – Failure to Make Disclosures or to Cooperate in Discovery
If the loss was negligent and caused prejudice, the court may order measures to cure the harm, but nothing more than necessary. If the destruction was intentional, the court may presume the lost information was unfavorable to the spoliating party, instruct the jury to draw that same presumption, or dismiss the case entirely and enter a default judgment. The distinction matters more than most people realize. An adverse inference instruction is available only on a finding of intent to deprive the other side of the information. Sloppy record-keeping alone doesn’t get that remedy, though lesser measures remain available. State of mind here directly drives the sanction, so documenting preservation efforts in real time is critical.
The Rule 26(f) Discovery Conference
Before formal discovery begins, the parties must meet and confer under Rule 26(f) to develop a discovery plan. In cases involving ESI, the rule directs them to discuss production formats, preservation issues, and procedures for handling privilege claims after documents have been exchanged.5Legal Information Institute. Federal Rules of Civil Procedure Rule 26 – Duty to Disclose; General Provisions Governing Discovery
This is where the most consequential eDiscovery decisions get made. Agreeing up front on search terms, date ranges, custodians, and production formats avoids expensive disputes later. It’s also the natural point to negotiate a clawback agreement for inadvertently produced privileged documents and submit it to the court as a protective order. Parties who treat the 26(f) conference as a formality tend to pay for it in motion practice down the road.
Review, Privilege, and Clawback Protections
Document review is where attorneys examine the surviving files to decide what gets handed over. Each document is evaluated on two questions: is it relevant to the claims or defenses, and is it protected by attorney-client privilege or work product doctrine? Relevant, non-privileged documents move forward to production. This stage is by far the most expensive part of eDiscovery. A RAND Corporation study found review accounts for roughly 73 percent of total production costs, with processing making up about 19 percent.6RAND Corporation. The Cost of Producing Electronic Documents in Civil Lawsuits
Privilege Logs
Withholding a document on privilege grounds requires more than silence. Rule 26(b)(5)(A) requires a party to expressly claim the privilege and describe the withheld document in enough detail for the other side to evaluate the claim, without revealing the privileged content itself.5Legal Information Institute. Federal Rules of Civil Procedure Rule 26 – Duty to Disclose; General Provisions Governing Discovery In practice that means a privilege log listing each withheld document with its date, author, recipients, subject matter, and the specific privilege asserted. In large cases, these logs can run to thousands of entries and become their own source of disputes.
Clawback Orders Under Rule 502(d)
No review is perfect, and accidentally producing a privileged document is a common nightmare. Federal Rule of Evidence 502(d) offers a safety net. A court can enter an order declaring that producing privileged material during the litigation does not waive the privilege, either in the current case or any other proceeding.7Legal Information Institute. Federal Rules of Evidence Rule 502 – Attorney-Client Privilege and Work Product
Without a 502(d) order, accidental production is governed by Rule 502(b), which forces the producing party to prove it took reasonable precautions and acted promptly to correct the mistake. With a 502(d) order, the document can simply be clawed back; the only remaining question is whether it was actually privileged. Entering this order early, ideally right after the 26(f) conference, is one of the smartest protective moves available, and there’s rarely a reason for either side to oppose it.
Technology-Assisted Review
Technology-assisted review, or TAR, uses machine learning to identify relevant documents instead of relying entirely on human reviewers reading every file. The two main approaches are predictive coding and continuous active learning. In predictive coding, reviewers code an initial set of documents, and the software uses those decisions to predict relevance across the rest. In continuous active learning, the system watches reviewer decisions in real time and continuously reranks the unreviewed population, pushing likely relevant documents to the top of the queue.
Federal courts have accepted TAR since at least 2012, when a Southern District of New York judge approved it as “an acceptable way to search for relevant ESI in appropriate cases.” Later decisions, including the widely cited Rio Tinto v. Vale opinion in 2015, reinforced that TAR protocols are permissible and often preferable to keyword searching or manual review for large document sets.8Justia Law. Rio Tinto PLC v Vale SA Generative AI tools have begun entering the market as well, with per-document pricing for AI-assisted review typically falling between $0.11 and $0.50 per document.
TAR doesn’t eliminate human judgment. Attorneys still validate the system’s decisions through sampling and quality checks, and privilege calls still require a human eye. For cases with hundreds of thousands or millions of documents, though, TAR dramatically cuts the time and cost of review and often achieves higher accuracy and consistency than teams of contract reviewers working file by file.
How Documents Are Produced
After review, responsive documents must be delivered to the opposing party in an agreed format. When the parties haven’t specified one, Rule 34(b) provides a default: ESI must be produced either in the form in which it is ordinarily maintained or in a reasonably usable form, and a party doesn’t have to produce the same information in more than one format.1Legal Information Institute. Federal Rules of Civil Procedure Rule 34 – Producing Documents, Electronically Stored Information, and Tangible Things
The most common formats are TIFF images and searchable PDFs, which preserve the appearance of the original while allowing text searching. Some files, especially spreadsheets and databases, are better produced in their native application format because converting them to images strips out formulas, hidden columns, and sorting. Productions typically include a load file, a data map that tells the receiving party’s review software how to import and organize the documents with their metadata and coding intact.
Proportionality and What eDiscovery Costs
Discovery is not unlimited. Rule 26(b)(1) requires discovery to be proportional to the needs of the case, and courts weigh six factors: the importance of the issues, the amount in controversy, the parties’ relative access to information, the parties’ resources, the importance of the discovery to the issues, and whether the burden of production outweighs its likely benefit.5Legal Information Institute. Federal Rules of Civil Procedure Rule 26 – Duty to Disclose; General Provisions Governing Discovery Spending $500,000 on discovery in a $100,000 case is hard to justify; constitutional questions get more latitude than routine contract disputes.
Rule 26(b)(2)(B) adds a separate protection for ESI stored in formats that are genuinely difficult or expensive to retrieve, such as disaster recovery backup tapes or decommissioned legacy systems. A party can identify these sources as “not reasonably accessible because of undue burden or cost” and decline to search them. The requesting party can challenge the designation, but the responding party has to prove the inaccessibility, and the court can require good cause before ordering production. When production is ordered anyway, the court can shift some or all of the cost to the requesting party.
The eDiscovery industry generates roughly $20 billion in annual spending. Vendor fees for processing raw data typically run from a few dollars to $25 per gigabyte depending on complexity. Monthly hosting fees generally fall below $10 per gigabyte without analytics features and up to $15 with them. Review software license fees commonly range from $50 to $100 per month per reviewer. The real expense, though, is human review. Contract review attorneys typically charge between $30 and $100 per hour, and a large case can require dozens of reviewers working for months. That’s why TAR has gained traction: when review represents nearly three-quarters of total production costs, any technology that reduces review hours delivers outsized savings. Negotiating scope limitations, phased discovery, and reasonable search terms at the Rule 26(f) conference remains the single most effective way to keep costs from spiraling.