Counsel Ops HQLitigation Matter & Spend Tech
eDiscovery

eDisclosure in England and Wales: PD 57AD, Software Choices, Costs and Timelines

PD 57AD decides what your eDisclosure platform has to do and roughly what it will cost. Here's how the Extended Disclosure Models, the Disclosure Review Document and TAR expectations turn into software requirements, budgets and timelines.

By the Counsel Ops HQ editorial team · November 2026 · 9 min read

What PD 57AD Actually Demands From Your eDisclosure Software

If your dispute sits in the Business and Property Courts of England and Wales, Practice Direction 57AD sets the disclosure rules. It has applied since October 2022, when it replaced the PD 51U pilot, and it ends the old default of standard disclosure. Parties now choose an Extended Disclosure Model for each issue in dispute and record those choices, along with their data sources and search methods, in the Disclosure Review Document (DRD). The right edisclosure software therefore depends less on feature checklists and more on whether it can deliver, and defend, the specific models and methods you sign up to.

This is a country guide for partners, in-house counsel and legal operations leads funding or managing a mid-size commercial dispute in London, including foreign parties litigating in England. Many of you will have custodians abroad, so the cross-border points later on apply directly. Three duties matter most for platform choice:

  • Preservation. Once proceedings are contemplated, documents must be preserved, and legal representatives must notify clients of that duty in writing. You need defensible legal hold and collection tooling before any review platform is chosen.
  • Initial Disclosure. The key documents a party relies on, plus those needed for the other side to understand the case, are served with the statements of case. There are exceptions, such as where the parties agree to dispense with it or the volume would be excessive. Early case assessment tools help here.
  • Known adverse documents. These must be disclosed whatever model applies. Your platform should let you tag and track them as their own category, separate from responsiveness.

Matching Models A to E With Platform Capabilities

Disclosure is set issue by issue, and the models carry very different technical loads. The table shows what each generally requires from your tooling. One dispute often uses two or three models across different issues, so pick the platform for the heaviest model you expect.

Advertisement
ModelWhat it requiresSoftware capability that matters
ANo disclosure beyond known adverse documentsIssue tagging, privilege logging, production of a small set
BLimited disclosure: Initial Disclosure-type documents, no search-based exerciseFast ingestion, simple review, clean production numbering
CRequest-led search for defined categories of documentsTargeted search, saved searches per request, audit trail tying each request to its results
DNarrow search-based disclosure, with or without narrative documentsSearch-term hit reports, deduplication, email threading, TAR or continuous active learning, validation sampling
EWide search-based disclosure, including train-of-inquiry documents (exceptional)Everything in D, plus scale, concept clustering and strong analytics

Our position: for a mid-size dispute, budget and specify for Model D. Model C sounds lighter, but it often becomes a run of contested requests, each needing its own documented search. A platform limited to keyword search and linear review will struggle the moment one issue moves to D. Model E is rare, and courts generally need persuading that it's justified.

Under Model D, narrative documents (background material relevant only to the story, not the issues) can be excluded. That cuts review volume substantially, but only if reviewers can code the distinction consistently in the platform.

Filling In the Disclosure Review Document Without Overcommitting

The DRD is where your software choices go on the record. Section 1A lists the issues for disclosure and each party's proposed model. Section 1B records Model C requests. Section 2 is a questionnaire covering data sources, custodians, date ranges, search terms, deduplication and use of technology. The claimant drafts the List of Issues for Disclosure shortly after the statements of case close, the parties work towards an agreed DRD, and it goes to the court ahead of the case management conference (CMC). Check the current practice direction for the exact deadlines.

Before you answer Section 2, your vendor should give you:

  • A custodian and data map: mailboxes, Teams or Slack, shared drives, mobile devices, and cloud systems such as Microsoft 365 or Google Workspace
  • Volume estimates before and after processing, deduplication and date filtering
  • Search-term hit reports with unique hits per term, which let you push back on over-broad terms from the other side
  • A clear statement of which analytics (threading, near-duplicate detection, TAR) you'll use and how you'll validate the results

Red flag: committing in the DRD to a methodology your vendor hasn't priced. Once the court has approved the DRD, changing course usually means agreeing it with the other side or going back to court. The parties and their lawyers also owe each other a duty to cooperate, and a Disclosure Guidance Hearing is available if you can't agree on scope.

When the Court Expects Technology Assisted Review

English courts have accepted technology assisted review since Pyrrho Investments v MWB Property (2016), where the High Court approved predictive coding for a review population in the millions. PD 57AD goes further and encourages it. The DRD questionnaire asks parties to consider TAR, and where the review set is likely to be large (a figure of around 50,000 documents is commonly cited), expect to explain why you aren't using it.

Courts don't require any particular algorithm. They look for a process that is proportionate and defensible. In practice that usually means:

  • Continuous active learning (CAL, sometimes called TAR 2.0) rather than older seed-set predictive coding. It adapts as reviewers code documents and suits rolling data collections
  • A documented stopping point backed by statistical validation, typically an elusion test on the documents not reviewed, with recall estimates you can report
  • Disclosure of the method to the other side, ideally agreed in the DRD before review starts

Most major platforms include CAL in their analytics tier, among them RelativityOne, Reveal, Everlaw, DISCO and Nuix Discover. The real differences lie in how clear the validation reporting is and how much of the work the vendor's own team does. Generative review features that summarize or first-pass-code documents are spreading. If you plan to use them, describe them as plainly in the DRD as you would TAR, and keep human-verified validation on the record. Whether a given approach is acceptable on your facts is a question for your solicitors and counsel.

Typical eDisclosure Costs and Timeline for a Mid-Size Dispute

Here's a hypothetical. Say a contractual dispute in the Commercial Court has 12 custodians on each side. Collection yields 400 GB. Processing, deduplication and date filtering reduce that to about 90 GB, or roughly 350,000 documents. Agreed search terms leave a review set of 140,000, and most issues are on Model D. We assume reviewers work at 40 to 60 documents an hour. The ranges below are typical 2026 market ranges. They vary considerably by contract, data type, region and vendor, so confirm current figures before you budget.

Cost driverTypical range (USD)Example total
Forensic collection (including mobile devices)$300–$1,500 per custodian or device$7,000–$35,000
Processing (often bundled; may be priced on collected volume)$0–$75 per GB$0–$30,000
Hosting with analytics$5–$20 per GB per month$5,000–$32,000 over 12–18 months
First-level review with CAL (perhaps 30–45% of the set reviewed)$40–$90 per hour for managed review$30,000–$140,000
Linear review of the full set, for comparisonSame hourly rate$90,000–$315,000
Project management, QC and production$125–$250 per hour$15,000–$50,000

Supervising-solicitor time and second-level privilege review sit outside these figures and are often the largest single cost. Where costs management applies, your Precedent H disclosure phase has to reflect the methodology in the DRD, so price the TAR approach before you file it.

A realistic sequence: preservation and collection take several weeks. Initial Disclosure is served with the statements of case. The List of Issues follows, and the DRD is negotiated over a month or two ahead of the CMC. The court then sets the Extended Disclosure timetable, often a matter of months, and the Disclosure Certificate is signed on completion. Lose a month to slow processing and your review window gets squeezed.

Choosing a Platform and Vendor: Criteria and Red Flags

For most mid-size disputes, a hosted SaaS platform with built-in CAL, run by a vendor with a London project team, is the better fit. The deciding factor is speed. You need hit reports and volume estimates quickly to negotiate the DRD, and on-premise or self-managed tools rarely deliver them fast enough.

Decision criteria

  • Data residency. Hosting in a UK region simplifies compliance with UK GDPR and the Data Protection Act 2018. Custodians elsewhere bring their own rules: EU transfer rules, India's Digital Personal Data Protection Act, data-localization requirements in parts of the Middle East and Southeast Asia, and bank-secrecy regimes. Plan in-country collection if local counsel advises it.
  • Validation reporting. Ask for an anonymized sample elusion and recall report from a past matter, of the kind you could hand to the court.
  • Pricing structure. Prefer all-in per-GB pricing that includes analytics, with reduced hosting rates once the matter goes dormant.
  • Production formats. The platform should handle load files, native production and redaction to the specification agreed with the other side.

Red flags

  • Analytics or TAR priced as an add-on after you've signed
  • No hosting region option, or a vague answer on sub-processors
  • A vendor that can't produce unique-hit search reports within a few days of processing

Next step: before the List of Issues is due, get two vendors to quote against the same custodian map and a Model D assumption, and require each to include a written TAR validation plan. If a vendor can't produce one, don't put its methodology in your DRD.

Questions buyers ask

Does PD 57AD apply to every civil claim in England and Wales?

No. It applies to proceedings in the Business and Property Courts, such as the Commercial Court, the Chancery Division and the Technology and Construction Court, subject to some exceptions. Other civil claims generally still follow CPR Part 31, where Practice Direction 31B and the Electronic Documents Questionnaire deal with electronic documents. Your solicitors can confirm which regime applies to your claim.

Is technology assisted review mandatory under PD 57AD?

No, but it's strongly encouraged. For large review sets, parties should expect to explain in the Disclosure Review Document why they aren't using it. Courts look for a proportionate method that can be defended and validated, not a particular product.

Who signs the Disclosure Certificate, and why does that matter for software choice?

A party with knowledge of the disclosure process signs it, and the legal representative signs a separate confirmation. Both attest that disclosure followed what was ordered. Your platform needs an audit trail showing the searches, models and validation actually match what the DRD recorded.

Can we host disclosure data outside the UK?

Often yes, but transfers must comply with UK GDPR transfer rules, and the laws of the custodians' own countries can add restrictions. Hosting in a UK region usually makes a London dispute simpler. For cross-border custodians, take advice from data-protection counsel in each relevant jurisdiction.

How much can TAR realistically save on eDisclosure costs?

In a mid-size Model D review, continuous active learning can mean reviewing a fraction of the documents a linear review would cover, often somewhere between a third and a half. On a review set like the hypothetical above, that can reduce first-level review spend by tens of thousands of dollars or more. Actual savings depend on how rich the data is, the quality of the search terms and the validation thresholds you agree, so get vendor estimates based on your own volumes.

Written by the Counsel Ops HQ editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.