# What Is E Discovery? Stages, Costs and Tools Explained URL: https://legalysis.co/journal/what-is-e-discovery Type: blog Locale: en Published: 2026-11-02 Updated: 2026-11-02 --- > E-discovery explained without jargon: what ESI is, the stages of the process, why review costs most, and what to prepare before a dispute reaches you. E-discovery is the process of finding, preserving, reviewing and handing over electronically stored information (emails, chat messages, files, database records) when a lawsuit, investigation or regulatory request makes that data evidence. If you are asking what is e discovery, think of it as document disclosure for the digital age: each side must hand over relevant data, and the way it does so is itself regulated. Here is what the term actually covers, in plain English, with the stages, the costs and the traps. ## What counts as electronically stored information? The legal term is ESI, electronically stored information. It is deliberately broad. Under the US Federal Rules of Civil Procedure, which shaped the practice everywhere else, ESI covers any information kept in electronic form that could bear on a dispute. In practice, that means more than email. A typical matter pulls in: - Email and attachments, including archived mailboxes - Chat and collaboration tools such as Slack, Teams or WhatsApp threads - Shared drives, cloud documents and their version history - Databases, CRM records and system logs - Text messages and files on phones and laptops - Metadata: who created a file, when it was changed, who opened it Metadata matters more than most people expect. A contract amendment that "was always there" looks different once the file properties show it was created three days after the dispute began. ![Desk with laptop, USB drive, smartphone and paper folders representing sources of electronic evidence](https://fdzlnqpwsaniezitwiuw.supabase.co/storage/v1/object/public/cms-media/legalysis/2026-11/0d4f7d-i1.webp) ## What are the stages of e-discovery? Practitioners use the EDRM model (Electronic Discovery Reference Model) as the shared map. It lists nine stages: identification, preservation, collection, processing, review, production, presentation, disposition, with analysis running across all of them. In real projects the middle stages carry the weight. - **Identification.** Work out who holds relevant data and where it lives. People, called custodians, and systems both go on the list. - **Preservation.** Stop relevant data from being changed or deleted. This is where a legal hold comes in. - **Collection.** Copy the data out in a way that keeps its metadata intact. - **Processing.** Convert files to a reviewable form, remove duplicates, extract text. - **Review.** Lawyers decide what is relevant, what is privileged, what is confidential. - **Production.** Hand the agreed set to the other side in an agreed format. - **Presentation.** Use the evidence at a deposition, hearing or trial. Disposition, the last stage, is the one people forget: once the matter ends, the held data can be released or deleted under a defensible policy. ## Why does a legal hold come first? A legal hold is a written instruction to stop deleting anything that might be relevant. The duty to preserve usually starts when litigation is reasonably foreseeable, not when the complaint arrives. That timing catches companies out. In practice, this means that the automatic 30-day chat deletion setting your IT team configured years ago has to be switched off for the people involved, quickly. Courts can impose sanctions when relevant ESI is lost because a party failed to take reasonable steps; the text of [Federal Rule of Civil Procedure 37(e)](https://www.law.cornell.edu/rules/frcp/rule_37) sets out how that works for lost electronic data. This is not a theoretical risk. It is what litigators see regularly: the dispute itself is manageable, the missing messages are what hurt. ## How much does e-discovery cost, and why is review the expensive part? Review is the costly stage because it is the only one that scales with the number of documents and needs lawyers' judgement. Collection and processing are largely technical and priced per gigabyte. Review is priced per hour, and a mid-sized dispute can mean hundreds of thousands of documents. That is why the industry moved from linear review (a person reads every document) to technology-assisted review, often shortened to TAR. A reviewer codes a sample as relevant or not, the software learns the pattern, and it ranks the rest. Courts in the US and UK have accepted this approach when it is documented and validated. Newer tools go further. Language models can summarise a document, suggest a relevance call and explain it in a sentence. They shorten the first pass; they do not replace the lawyer who signs off on privilege. Skip any vendor who promises that review can run unattended. ![Lawyer marking up printed pages next to a laptop during document review](https://fdzlnqpwsaniezitwiuw.supabase.co/storage/v1/object/public/cms-media/legalysis/2026-11/42c8e3-i2.webp) ## What does proportionality mean for scope? Proportionality is the principle that discovery should be sized to the case. Under [Rule 26(b)(1)](https://www.law.cornell.edu/rules/frcp/rule_26), US courts weigh the importance of the issues, the amount at stake, each side's access to the information, the burden of producing it and whether the benefit justifies the expense. Two consequences follow. First, you rarely have to search everything. Agreeing on custodians, date ranges and search terms early cuts volume sharply. Second, "it would be expensive" is not an automatic defence: you have to show the burden with specifics. The meet-and-confer discussion under Rule 26(f) is where these choices get made. Parties who arrive with a data map and a realistic proposal usually keep control of the process. Parties who arrive without one tend to have the scope set for them. ## Is e-discovery only a US thing? No, but the vocabulary differs. English courts use "disclosure" and have an e-disclosure regime under Practice Direction 57AD in the Business and Property Courts, with a structured disclosure review document. Switzerland and most civil-law EU countries have no US-style discovery at all: parties generally submit the evidence they rely on, and requests for production are narrower. This matters for cross-border work. A Swiss or Austrian company sued in a US court may face broad discovery duties while data protection law (GDPR, the Swiss FADP) limits what can be moved abroad. Before transferring personnel data to a US vendor, check the rules with your counsel. The answer depends on the jurisdiction, and a generic checklist will get it wrong. ## What does a realistic timeline look like? Take a mid-sized commercial dispute between a software supplier and a customer. A hold notice goes to eight people on day one. IT confirms within a week which mailboxes, chat workspaces and shared folders are in scope. Collection takes a few days, processing another few, and the review set lands in a platform before the end of the first month. Then the real work starts. A first-pass team codes documents for relevance, a second pass checks privilege, and the senior lawyer samples the results to confirm quality. If the parties agreed on search terms early, the set is a fraction of what a blunt "collect everything" approach would have produced. If they did not, expect weeks of argument about scope before a single document is produced. The lesson is plain: the decisions made in the first ten days, who is a custodian, which systems count, what the date range is, set the cost of the next six months. ## Which mistakes show up again and again? Four errors appear in most post-mortems, and none of them needs sophisticated technology to avoid. - **Late holds.** The notice goes out after the dispute is public, and weeks of routine deletion have already run. - **Forgotten sources.** Teams preserve email but miss the ephemeral messaging app the project team actually used. - **Over-collection.** Pulling every mailbox in the company inflates review costs without improving the evidence. - **No documentation.** Nobody records why a search term was chosen or a source excluded, so the choice cannot be defended later. Each one is a process failure, not a technology failure. A one-page protocol, owned by someone named, prevents most of them. ## Where do contracts fit into e-discovery? Contracts are often the first documents requested, and the clause text decides what happens next. Three kinds of terms deserve attention before any dispute starts: - **Record retention and audit clauses**: how long you must keep records, and what a counterparty may inspect. - **Confidentiality and data-location clauses**: they affect whether you may hand a counterparty's data to a review vendor. - **Cooperation and notice clauses**: some agreements require you to tell the other side when a claim is threatened. A contract analysis tool can flag these clauses in a portfolio in minutes, so that when a hold notice goes out you already know which agreements add constraints. The tool analyses; the decision stays human, and anything that could affect your preservation duties is worth checking with your counsel. ![Two colleagues planning discovery scope at a meeting table](https://fdzlnqpwsaniezitwiuw.supabase.co/storage/v1/object/public/cms-media/legalysis/2026-11/fc3ba5-i3.webp) ## Which e-discovery tools are worth knowing? The market splits into three groups. Dedicated e-discovery platforms handle collection through production at scale. General legal AI assistants such as Harvey and CoCounsel help with summarising, research and first-pass review. Contract-focused tools such as Kira Systems extract clauses from agreements. Choose on four criteria: where the data is hosted, whether the vendor trains on your documents, whether results are explained so a reviewer can verify them, and how exports are handled for production. Skip any tool that cannot answer the second question in writing. ## What should you do this week? You do not need a litigation to prepare. Three actions pay off quickly: - **Write a data map.** List systems, owners and retention periods for email, chat, shared drives and phones. - **Draft a legal hold template.** Have counsel approve it now, so you are not writing one under pressure. - **Review the retention settings** on chat and email, and know how to suspend deletion for specific people within a day. That is the practical answer to what is e discovery: a discipline of knowing where your data is, keeping it when it matters, and reviewing only what the case needs. Have your counsel check the most sensitive point before you act. ## FAQ ### What is e-discovery in simple terms? E-discovery is the process of identifying, preserving, collecting, reviewing and producing electronic data such as email, chat and files as evidence in a lawsuit, investigation or regulatory request. ### What does ESI stand for? ESI means electronically stored information. It covers email, messages, documents, databases, logs and metadata, wherever they are stored, including phones and cloud services. ### What are the stages of e-discovery? The EDRM model lists identification, preservation, collection, processing, review, production, presentation and disposition, with analysis running throughout. Review is usually the most expensive stage. ### What is a legal hold? A legal hold is a written instruction to preserve potentially relevant data and suspend routine deletion. The duty generally starts when litigation is reasonably foreseeable. ### Is e-discovery the same as e-disclosure? They overlap. E-disclosure is the English term, governed in the Business and Property Courts by Practice Direction 57AD. US practice uses e-discovery under the Federal Rules of Civil Procedure. ### Can AI do e-discovery review? AI and technology-assisted review can rank and summarise documents, which shortens the first pass. A lawyer still has to validate results and decide on privilege. ### Does e-discovery exist in Switzerland or the EU? Most civil-law systems have no US-style discovery. Parties submit the evidence they rely on, but cross-border cases can still trigger US obligations alongside data protection limits.