eDiscovery platform comparison: what the pricing models reveal

Comparing eDiscovery platforms: the dimensions that actually separate them

This comparison offers no verdict, ranking, or recommendation. The platforms overlap heavily on headline capability, so the differences that matter are the pricing model, the operating model, and the fit with a particular matter profile—not a feature checklist that changes with every release. This is not a review, and nothing was tested, trialed, benchmarked, or hands-on reviewed.

The search for “best eDiscovery software” often produces a list because a single answer would be misleading. The more useful question is which platform’s pricing and operating assumptions match the data, people, and workflow in a specific matter. Public pricing pages support only part of that analysis, but they reveal enough to establish one hard distinction: the units used to frame price.

This page provides procurement and process guidance, not legal advice. Requirements that depend on a court order, jurisdiction, protective order, or the parties’ discovery agreement need matter-specific analysis.

The verified pricing distinction#

Each vendor’s pricing page was checked in November 2026. None of Everlaw, Relativity, Logikcull, DISCO, or Nextpoint publishes a price figure on its pricing page. Every one routes the reader to a sales conversation, so no dollar comparison can be made from the published pages.

What can be compared is the pricing unit each page names, because that unit indicates which cost driver is central to the vendor’s published framing.

Platform	Published pricing unit	What the public page states
Everlaw	No unit named	No price figures. The page presents use cases including eDiscovery, early case assessment, legal holds, depositions, data privacy, internal investigations, and FOIA and public records. The call to action is to schedule a meeting.
Relativity	No unit named	No price figures. The call to action is to talk to sales. The page is titled for Relativity aiR, and the site states that what was RelativityOne is now presented under that name, so current product documentation should be checked.
Logikcull	Per-GB and per-month	No price figures. The page references per-GB and per-month units, offers a free trial, and directs the reader to contact the company.
DISCO	Per-document	No price figures. The page references a per-document unit and directs the reader to contact the company.
Nextpoint	Per-user, per-GB, and per-document	No price figures. The page is titled around simple per-user software pricing and also references per-GB and per-document units. It directs the reader to contact the company.

What each pricing unit rewards and punishes#

A pricing unit is more than a labeling convention. It reveals what the vendor expects to grow with the matter and, consequently, where cost pressure is likely to appear. The published units show the model’s incentives; they do not establish the final quote, allowances, minimums, overages, or contract terms. Those details require a written proposal from each vendor.

Per-GB pricing#

A per-GB model rewards early scoping because the amount of data processed affects the pricing calculation. It punishes matters in which a large collection contains little material to review, especially when the collection includes media, duplicates, or files outside the likely population of relevant material.

That model tends to fit early case assessment, broad custodial collections, and matters where collection size is the principal uncertainty. It creates pressure for defensible data minimization before ingestion. It can be less natural for a record-centered matter with a stable, narrow population, where page or document counts may reflect review scope more directly than source volume.

Per-month pricing#

A per-month model rewards a stable, continuously active workflow because the pricing period does not directly track each file, collection, or user action. It punishes short matters that consume intensive effort in a brief window or organizations whose activity rises and falls sharply.

That framing tends to fit recurring legal-hold, investigation, or case-management work with a predictable monthly cadence. It can suit teams that want budget predictability more than consumption-based alignment. The proposal still needs to define what a month includes, particularly for unusually large collections, additional processing, or temporary staffing changes.

Per-document pricing#

A per-document model rewards precise population control because the number of documents placed into the workflow can affect the pricing calculation. It punishes collections dominated by media, native files, or other large items that create substantial processing effort without producing a comparable number of reviewable documents.

That model tends to fit matters with a defined document population and a clear path from collection through review. It can be less suitable for early case assessment in which the first pass deliberately includes a wide collection before the relevant population is known.

Per-user pricing#

A per-user model rewards a stable user population and makes staffing assumptions visible. It punishes broad access arrangements, temporary reviewers, specialist reviewers, or infrequent users if access is priced regardless of actual use. Its incentives favor matters with a predictable number of people working in the platform at a given time.

That framing can fit a matter with a stable review team, including regular internal access by legal and litigation-support personnel. It can be less aligned with a collection-heavy phase followed by a short production sprint, particularly if the model charges for named users rather than active users.

No published unit#

Everlaw and Relativity publish neither a price figure nor a pricing unit. That absence makes direct public cost comparison impossible and puts more weight on proposal structure: which inputs count, which activities are bundled, and which events can change the price after deployment.

The absence of a named unit does not mean that a platform lacks a pricing model. It means the model cannot be identified from the pricing page. A buyer should ask for the unit, treatment of each cost driver, and the events that trigger additional charges before treating either platform as economically comparable with a per-GB, per-month, per-document, or per-user offer.

Everlaw and Relativity: what can—and cannot—be compared#

For readers asking “Everlaw vs. Relativity,” the public pages do not support a better-versus-better conclusion. Their pricing pages provide the central verified finding: neither publishes a price figure or names a pricing unit, and both direct the buyer into a sales conversation.

Everlaw’s page frames the platform around a broad set of use cases: eDiscovery, early case assessment, legal holds, depositions, data privacy, internal investigations, and FOIA and public records. It names AI features including coding suggestions, a writing assistant, a review assistant, predictive coding, and clustering. It also states that the platform connects to external AI tools so that AI outputs are grounded in evidence held in the platform.

Relativity’s public page carries a naming caveat that matters for any comparison. It is titled for Relativity aiR, while the site states that what was RelativityOne is now presented under that name. Product naming is therefore in flux, and current documentation should be checked rather than assuming that a label on an older comparison or contract still identifies the current offering.

Relativity also references a separate server policy alongside its cloud platform and a FedRAMP-certified option for government. Those are checkable points of orientation, not a substitute for comparing deployment architecture, contractual commitments, or workflow with the matter at hand.

The practical comparison therefore begins with the proposal and the matter, not a feature verdict. Without disclosed pricing units and like-for-like test results, claims about which product is “better” would be assumptions presented as conclusions.

Dimensions that require a like-for-like evaluation#

The most consequential differences cannot be established from these pricing pages. They must be tested against a common data profile and a common sequence of tasks. Each dimension below can produce a different result depending on the collection, team, and legal requirements.

Ingestion breadth on the actual file population#

Ask whether the platform can ingest the organization’s real file types, including the native formats, archives, messaging exports, databases, audio, video, images, and encrypted or corrupted files that may appear in the matter. A broad product statement does not answer whether a specific combination of formats, metadata, encoding, or file sizes processes consistently.

The comparison corpus should include files with relevant content but awkward technical characteristics. Record the result, warnings, exceptions, and any need for manual steps. Ingestion failures discovered after review begins can change both cost and staffing assumptions.

Processing behavior and exception rates#

Processing should be compared as a sequence, not as a single speed indicator. Include deduplication, extraction, indexing, OCR or text conversion where applicable, metadata handling, and error handling. For each dataset, preserve the exception count, reason for failure, retry behavior, and amount of staff intervention.

“Processed successfully” is not a sufficient answer if the same result requires repeated manual repair. A low headline rate can conceal exceptions concentrated in a small, difficult file population that matters disproportionately to the review.

Search expressiveness#

Test the search syntax, field restrictions, proximity controls, Boolean behavior, stemming, saved searches, and result counts with terms that legal professionals actually expect to use. The same query should be entered in each platform, with filters and date or custodian restrictions applied consistently.

Compare both the visible results and the reasons documents may be excluded. Search behavior that appears broad can still be narrow if a field, metadata condition, or document type silently constrains the population.

TAR workflow auditability#

For technology-assisted review, the evaluation should follow a workflow from training through validation, production review, and any reconsideration. Record who approved each stage, what decision was made, what information was displayed, and whether the platform retained an inspectable history.

The question is not whether AI is present. Everlaw’s page names specific AI functions and describes grounding outputs in evidence held in the platform, but the actual workflow still needs evaluation. Auditability should also be considered under the organization’s own governance requirements, including access controls, retention, and documentation.

Redaction and privilege handling at volume#

Compare the treatment of burn notes, redactions, privileged content, confidentiality designations, and exceptions. Include native redactions where the output format requires them, along with how the platform records decisions, supports QC, and distinguishes a completed redaction from a proposed action.

Volume matters because exceptions that are manageable on a sample can become a staffing problem in a large review. The test should also examine whether reviewers can identify why content was flagged and whether corrections are recorded in a way the team can review.

Production and load-file flexibility#

Compare production formats, field mapping, metadata controls, redaction behavior, Bates numbering, image or file packaging, and the handling of missing or duplicate values. Generate the same production request against the same reviewed population and inspect the resulting structure, not just the successful completion message.

Ask whether load-file definitions can accommodate counsel’s agreed specifications and whether changes can be reproduced without rebuilding the entire production. Requirements vary by matter and governing agreement.

Administration model#

Administering the platform involves user provisioning, matter creation, permission groups, data retention, legal-hold workflows, audit exports, and support access. Compare the number of steps and the extent to which routine work requires vendor personnel, particularly for changes that happen after litigation begins.

The operating model can outweigh a visible pricing-unit difference when an in-house team lacks capacity to manage manual administration or when the required administrator workflow differs materially between platforms.

Export on exit#

Exit planning should begin before selection. Determine which native documents, extracted text, review decisions, tags, redactions, production files, audit logs, and reports can be exported, and in what format. Confirm whether exports preserve the original data, decision history, and chain of associations needed for later review or an incoming vendor’s ingestion.

Also examine contractual restrictions, export fees, format dependencies, and whether the receiving platform can accept the exported material without material loss. A low-friction exit path can reduce disruption if the matter changes direction or the platform is replaced.

How to run a meaningful platform comparison#

A controlled comparison requires one data profile, one pricing basis, the same test corpus, and one record traced through the complete workflow. The purpose is not to manufacture a universal winner; it is to expose differences under conditions the organization actually expects to face.

Use one data profile#

Define the matter’s collection sources, custodians, date range, approximate collection size, file-type mix, estimated review population, expected user count, and production stage. State which figures are known and which are estimates. Keep the same assumptions in every proposal request.

A common profile prevents an apparent pricing difference from reflecting different inputs. If two platforms cannot accept the same assumptions, record that divergence rather than adjusting one scenario silently.

Use one comparable pricing basis#

Ask each vendor to explain its unit, include assumptions, allowances, overages, implementation charges, training, support, and the events that can change the cost. Where units differ, request a common representation or have each vendor translate its proposal using the same documented assumptions.

Translation should preserve uncertainty. If a vendor cannot state how a collection or workflow maps to its pricing, that uncertainty belongs in the evaluation. The published pages do not provide enough detail to make the conversion independently.

Use the same test corpus#

Build a representative corpus with routine files, technically difficult files, potentially privileged content, redactions, duplicates, and records connected to expected searches. Include enough material to exercise review rather than relying on a small demonstration set.

Run the same collection, search, review, production, and quality-control tasks. Capture exceptions and staff intervention in a common worksheet. The resulting data shows how each platform handles the organization’s actual population.

Trace one record end to end#

Select one record with enough complexity to identify meaningful transitions. Follow it through ingestion, text extraction or conversion, indexing, search retrieval, review, any coding or categorization, redaction, privilege treatment, production, and QC.

At each stage, note where the record was transformed, where a field was added or lost, and which action created a reviewable audit entry. Repeat the trace with a second record if the first proves unusually clean. One end-to-end path often reveals integration and export issues that isolated feature demonstrations conceal.

What “best eDiscovery software” really means#

There is no defensible general ranking among these platforms from the published pages. “Best” resolves into “best for what matter profile,” and several profiles lead to different evaluation priorities.

• Collection-heavy early case assessment: Ingestion behavior, exception handling, and the consequences of processing large volumes carry more weight.
• Stable recurring matter management: Administrative effort, user structure, and predictable monthly operating assumptions become more important.
• Narrow record-centered discovery: Document population controls, review efficiency, search precision, and production mechanics may lead the evaluation.
• Broad team or multi-user workflow: Permissions, user administration, auditability, and the fit between staffing and pricing units merit greater attention.

Those profiles do not map automatically to a vendor winner. The published pricing units indicate which cost drivers each vendor emphasizes, but only a proposal and matter-specific evaluation can establish the actual economics and operating fit.

Everlaw and Relativity publish no pricing unit; Logikcull references per-GB and per-month units; DISCO references per-document; and Nextpoint references per-user, per-GB, and per-document. Those are the only hard pricing distinctions established by the pages checked in November 2026.

Verify current pricing and capability with each vendor because these pages change. This comparison ranks nothing, recommends nothing, and tested nothing.

Common questions#

What is the best software for law firms?

The comparison identifies no best platform. It states that a single answer would be misleading and offers no verdict, ranking, or recommendation.

Which pricing units do the platforms publish?

Everlaw and Relativity name no unit; Logikcull names per-GB and per-month; DISCO names per-document; Nextpoint names per-user, per-GB, and per-document.

What dimensions require like-for-like evaluation?

Compare ingestion breadth, processing exceptions, search, TAR auditability, redaction and privilege handling, production, administration, and export on a common data profile and task sequence.

What does a controlled platform comparison require?

Use one data profile, one pricing basis, the same test corpus, and one record traced through the complete workflow.