Relativity and Everlaw alternatives, and the number none of them will print: what does a gigabyte cost?

Start with the thing you can check yourself in ten minutes, because it sets the tone for everything that follows. On August 22, 2026 we requested the pricing page of nine e-discovery platforms with a declared research user agent, followed every redirect, recorded the status code, and counted the dollar figures in the returned document. Here is what came back.

Read the table twice, because the interesting column is the last one. This is not a category that is quietly quote-only and embarrassed about it. It is a category where the marketing copy on the pricing page is about pricing transparency, at some length, without transparency. Everlaw's page carries the headline promise of clear pricing without surprises and a subhead about 100 percent predictable pricing, and every specific it offers is structural rather than numeric: what is included in the per gigabyte rate, which AI actions cost credits, that unused committed credits expire at the end of the term. Logikcull tells you to predict your spend to the penny and then describes two plan shapes, storage-based and matter-based, with no rate for either.

Nobody here is obliged to publish a rate, and plenty of enterprise software does not. The point is narrower and more useful: if you are running a shortlist, six of these nine pricing pages advertise transparency, predictability or simplicity as something the vendor competes on, which means it is fair game on the call, and none of them has given you a figure to anchor against. Walk in with Digital WarRoom's published sheet in hand and you have the only public anchor in the category.

What Digital WarRoom publishes, for the record, since it is the anchor: on-premises Pro software at 2,500 dollars per year per license with unlimited matters, a hosted single matter plan at 500 dollars per month including 50 gigabytes with 1 dollar per gigabyte after that, a hosted private cloud at 2,000 dollars per month including 500 gigabytes on the same overage rate, near-line archive at 50 cents per gigabyte with a 100 dollar minimum, and professional services at 315 to 450 dollars per hour. Its own buyers' FAQ answers the question "Are there any further fees beyond what is listed on this page?" with "No further fees!" and names what that excludes: no processing fees, no matter setup fees, no matter migration fees, no support or training fees in hosted subscriptions. Whether that product suits your matters is a separate question, addressed further down. As a pricing document it is the only one of the nine you can actually read.

Here is the structural fact underneath the whole category, and it is stated openly by the vendors themselves. In the products a firm normally buys, the meter runs on headcount. Practice management, research subscriptions, drafting add-ins, document management: all of them charge per user per month, so the bill is a function of how many people you employ, which is a number the managing partner sets.

E-discovery inverts that completely. Seats are free and data is metered. Everlaw's pricing page lists "Unlimited User Licenses" under core platform features and says pricing is based on the amount of data you manage and your usage, without limits on users or uploads. Logikcull's pay-as-you-go plan reads "Unlimited users, Unlimited projects, Unlimited downloads" and prices per gigabyte of data stored each month. DISCO says it bills on processed data, describes its platform features as included in a per gigabyte rate, and lists unlimited users and unlimited custodians under its Hold product. Nuix describes consumption-based licensing that flexes with your data volumes. Digital WarRoom's single matter plan reads "1 Matter: Unlimited Users" against a gigabyte allowance.

Sit with what that means for a firm of eighty attorneys. Your cost per matter has almost nothing to do with your firm. It is a function of how many custodians the opposing party has, how long they have been at the company, whether they lived in email or in Slack, whether anyone ever ran a retention policy, and whether the case is the kind where somebody kept every attachment. You can be efficient, disciplined and lucky and still receive an invoice four times the size of the one you received last quarter, because a different company's IT department made different choices six years ago.

Two consequences follow, and they are the ones that matter to whoever signs the contract.

The first is that per-seat intuition is worthless here. A firm administrator who has bought software for a decade knows roughly what a seat costs and can multiply. That instinct produces a number with no relationship to the invoice. The forecast has to be built on data volumes per matter type, and if your firm has never measured those, a per gigabyte rate is a number you cannot convert into an annual figure no matter how long you look at it.

The second is that the AI layer is now a second, separate meter. Everlaw includes single-document AI actions and its Writing Assistant in the subscription and charges batch actions on usage, against purchased credits, with the note that unused committed credits expire at the end of the term. That is the important sentence: the batch actions are the ones you would use on a large review, so the AI budget scales with volume too, and the credits are use-it-or-lose-it. DISCO takes the opposite line and says there is no additional charge for unlimited use of Cecilia AI on its platform, while charging its Auto Review product per document. Relativity's pricing page states that RelativityOne includes aiR for Review, aiR for Privilege and aiR for Case Strategy at no additional cost. Three vendors, three different answers to "is the AI extra", and all three answers live on the pricing page rather than in the contract you will actually sign. Ask which one governs.

Every other comparison on this site is a comparison between products. This one is not, quite, because a large part of what you are buying is a position under three rules. We quote them from the rule text rather than from anyone's summary, because the wording is the part that does the work.

Rule 26(b)(1) makes your own resources a factor. The scope of discovery is any nonprivileged matter relevant to a claim or defense "and proportional to the needs of the case, considering the importance of the issues at stake in the action, the amount in controversy, the parties' relative access to relevant information, the parties' resources, the importance of the discovery in resolving the issues, and whether the burden or expense of the proposed discovery outweighs its likely benefit." Read the list slowly. The parties' relative access to relevant information is in it. The parties' resources are in it. Whether the burden or expense outweighs the likely benefit is in it. Proportionality is not a slogan a vendor put on a landing page; it is a six-factor test in which what the review will cost you is an express factor. A firm that can state its per gigabyte cost, its review rate and its projected volume is arguing proportionality with numbers. A firm that cannot is arguing it with adjectives.

Rule 37(e) attaches before anybody has priced anything. Where electronically stored information "that should have been preserved in the anticipation or conduct of litigation is lost because a party failed to take reasonable steps to preserve it, and it cannot be restored or replaced through additional discovery", the court may order measures no greater than necessary to cure prejudice; and only on a finding that the party acted with intent to deprive another party of the information's use may it presume the lost information was unfavourable, instruct a jury to presume it, or dismiss the action or enter default judgment. Note the sequencing. The duty runs from anticipation of litigation, which is well before the platform decision, often before the engagement letter. This is why every serious vendor on the list ships a separate legal hold product, and why the hold question and the review question get bought at different moments by different people inside the firm.

Rule 502(d) is the cheapest risk control in the category and it is not software. A federal court may order that privilege or protection is not waived by disclosure connected with the litigation pending before it, and where it does, the disclosure is also not a waiver in any other federal or state proceeding. A 502(d) order is a paragraph. The privilege review it protects you against is among the most expensive per-document work in the matter. Any vendor conversation that gets deep into privilege detection accuracy without anyone at the table asking whether there is a 502(d) order in place is a conversation about the wrong control.

None of this is legal advice and none of it is novel to a litigator who does this weekly. It is here because it is the part that gets left out of the procurement conversation, where the questions tend to be about the interface. The three rules are what actually set the size of the number the interface is going to bill you.

The single most useful line for a mid-market buyer is one Relativity puts on its own law firms page as a boast: "Relativity powers 198 of the Am Law 200. Preferred by the largest global firms, yet flexible enough for any size matter." The same page says "From large law firms to litigation boutiques", so the company is not claiming to be enterprise-only. But read the sentence as a fact about the reference class. A product whose reference customers are the two hundred highest-grossing firms in the country is a product built, priced and supported for organisations with a litigation support department, a director of practice technology, and enough matter flow to keep a platform administrator busy. If your firm has one paralegal who has become the accidental owner of e-discovery, you are not the reference customer, whatever the flexibility line says.

Relativity also runs a substantial partner channel: its own site carries service provider partner pages, a partner directory, a partner learning portal and a "Become a Partner" path. That channel is how a firm without its own litigation support department typically reaches the platform: a service provider holds the licence and resells hosting and project management. That is not a criticism; for a lot of firms it is the right answer, because it converts a platform decision into a per matter service purchase. It does mean that a shortlist which puts Relativity next to a self-serve product is comparing two different transactions.

Everlaw is the vendor on this list that most visibly sells down into the band, and the honest reading of its own site says so. It publishes a dedicated Everlaw for Boutique Firms page and a dedicated Everlaw for Plaintiffs' Firms page, and the law firm customers it puts on its own law firms page are United States firms rather than global ones: Tonkon Torp, Adams and Reese, Sher Edling, and Ferguson Case Orr Patterson. Its own pricing copy leans on the same point, promising to "Control costs and minimize legal spend with Everlaw's transparent pricing model for firms of all sizes" and naming no additional costs for users, training, onboarding, data migration or support. If a mid-market firm is going to host review itself on a major platform, this is the one whose own marketing is aimed at that firm.

Casepoint's own site does not list law firms as a market at all. Its markets are Government (federal civilian, federal defense, state, local and education, Canada) and Corporate, and its teams are corporate legal, government legal, FOIA and public records, government audit and investigations, and IT and information security. Its differentiators are FedRAMP High and Department of Defense IL5 and IL6 authorisations, with the claim that only six software-as-a-service companies in the world hold IL6. That is a serious product with a serious moat, aimed at a buyer who is not you. If it has arrived on your shortlist anyway, it should usually come off again unless your practice is substantially government-facing.

Nuix and DISCO sit either side of that line. Nuix's own industry list is regulators, financial services, government, telecommunications, pharmaceutical and energy, with pricing described as built for enterprise scale; law firms are a use case rather than an industry. DISCO does list law firms first among its solutions and separately sells professional services, document review, forensic services and collections, and enterprise managed services, which is the shape of a vendor comfortable doing the work for you rather than renting you the tool.

One more fact about DISCO that no other vendor on this list offers, and it belongs in a buying decision: it is the only one you can read the audited financials of. CS Disco, Inc. is an SEC registrant, central index key 0001625641, filing under file number 001-40624, and its most recent annual report on Form 10-K was filed on February 25, 2026. If the durability of your platform vendor matters to you, and on a matter that has to be retrievable years after it closes it might, one of these companies has to tell you the truth about its finances once a year in a document you can download for nothing.

Almost everything else on a firm's technology budget is overhead. Research subscriptions, practice management, drafting tools: the firm absorbs them and hopes for leverage. E-discovery is different, and Relativity says the quiet part on its own law firms page under the heading "Invest in clients and recover costs": "Relativity makes cost recovery a no-brainer. Your clients gain a competitive edge in the courtroom while you grow revenue for your firm."

Strip the marketing and that is an accurate description of how the category works. Hosting and review costs are commonly passed through to the client as a disbursement, sometimes at cost, sometimes at a marked-up rate, depending on the engagement letter and on what the client's outside counsel guidelines permit. Which means the buying question a mid-market firm should be asking is not the one it usually asks.

The usual question is "can we afford this platform". The real questions are four, and they are arithmetic:

• Can you bill it? Read your own engagement letters and your largest clients' outside counsel guidelines before you read another feature list. Some guidelines cap or forbid technology mark-ups outright. Some require the vendor invoice to be attached. If your five biggest clients will not accept a hosting line item, hosting is overhead and should be evaluated as overhead.
• At what margin, and is the margin defensible? A pass-through at cost is easy to justify and produces no revenue. A mark-up produces revenue and invites a fee dispute at exactly the moment a matter goes badly. Whichever you choose, choose it once, write it down, and apply it the same way on every matter, because the version that causes trouble is the one that varies by who quoted it.
• Who eats the gap? The failure mode is specific: the firm commits to an annual platform contract sized to expected volume, the volume does not arrive, and the unbilled subscription sits on the firm's own profit and loss with nothing to pass through. Per matter pricing pushes that risk back onto the vendor and costs more per matter in exchange. That trade is the actual decision.
• Can you produce the record twelve months later? When a client questions a disbursement, or a court asks about the burden you claimed under Rule 26(b)(1), somebody has to reconstruct what was hosted, for how long, at what rate, on whose instruction, and what was billed against it. If that reconstruction would currently take a week of somebody's time, the record is the thing to fix before the platform is.

That last one is where the interesting work is, and we will come back to it, because it is the only part of this problem we would take money to solve.

We will correct ourselves in public rather than let this sit. Our page on Harvey alternatives for mid-market law firms tells a firm whose biggest leverage point is document review at scale that Harvey is worth a real evaluation. That sentence is right for one meaning of the phrase and wrong for the other, and the difference is worth a section.

Meaning one is transactional and knowledge work. Reading a data room during due diligence. Reviewing a stack of contracts for a change-of-control clause. Pulling the argument out of a hundred prior briefs. These are documents you chose to look at, in volumes a person could theoretically read, where the value is analysis. Legal AI research and drafting products are built for this, and that is the sense in which our Harvey page is correct.

Meaning two is litigation discovery. Hundreds of thousands of documents arrive because a court ordered them to. You must review all of them for responsiveness and for privilege, log the privileged ones, redact what needs redacting, and produce the rest in a specified format with a defensible record of how you decided. Nobody chose this corpus and nobody is reading it for insight. The product category built for this is the one on this page, and its features exist because of the obligations: processing and imaging, near-duplicate detection, predictive coding and active learning, redaction tooling, production sets, legal hold and custodian tracking, audit logs.

The two categories have converged in marketing language and have not converged in product. A drafting or research tool has no production module, no privilege log, no chain of custody and no hold. A review platform is a poor place to draft a brief. A firm that buys the first because a shortlist said document review, and then receives a production, has bought the wrong instrument.

So: if your document burden is diligence, contracts and knowledge retrieval, read the Harvey alternatives page, CoCounsel against Harvey, and Spellbook against a custom build. If your burden is opposing productions, stay here. If it is genuinely both, they are two budgets, and pretending otherwise is how one of them ends up unfunded.

The shortlist is usually written as a list of platforms. It is more usefully written as a list of transactions, because the platforms cluster into five quite different arrangements, and the arrangement decides more than the software does.

One. Buy the platform your opponents use, through a service provider. This is the usual route to Relativity for a firm that is not in the Am Law 200. You do not hold the licence; a service provider does, and sells you hosting, processing and project management per matter. You get the format everybody else in the case already speaks, an experienced human running the technical side, and no fixed commitment. You pay a margin on top of the platform cost, and you do not own the relationship or the pricing. Right answer when litigation is episodic and each matter is large enough to justify a project manager.

Two. Hold the platform yourself and run review in-house. This is what Everlaw's boutique and plaintiffs' firm pages are selling, and what DISCO and Casepoint sell to firms with real volume. Unlimited seats mean you can put the whole case team in the tool at no marginal cost, which genuinely changes how a small team works. You take on the annual commitment, the administration and the training. Right answer when your matter flow is steady enough that the commitment gets consumed, and when somebody in the building actually wants to own the platform.

Three. Buy the small, priced, self-serve tool. Digital WarRoom is the clearest example, at a published 500 dollars per month for a single hosted matter with 50 gigabytes included, or 2,000 dollars per month for unlimited matters with 500 gigabytes. For a firm whose matters are measured in tens of gigabytes rather than terabytes, the arithmetic here is not close, and the fact that you can do the arithmetic at all is the point. What you give up is the analytics depth and the AI tooling of the larger platforms, and the reassurance of using what the other side uses.

Four. Do not host it at all; buy managed review per matter. DISCO sells document review, forensic services and collections, and enterprise managed services alongside its platform, and the independent service provider market sells the same arrangement. The cost is fully variable and fully passable through to the client, there is no commitment, and there is nothing to administer. You have the least control over quality and the least ability to compress the timeline yourself.

Five. The hybrid almost nobody puts on the shortlist, and which is often correct. Buy the small tool or a per matter arrangement for the routine work, and reserve the large platform, through a provider, for the two or three matters a year that genuinely need it. Firms resist this because it looks like running two systems. It is running one system for the matters that pay for it and another for the matters that do not, which is what a well-run department does with any expensive capacity.

Notice what is not on this list: commissioning a custom review platform. That is covered below, and the answer is no.

Four measurements, all of which exist somewhere in your billing system or your last four matter files, and all of which take an afternoon to pull. Do them before the demos rather than after, because after the demos you will be comparing feature lists.

Measure one: gigabytes per matter, for your last eight litigated matters. Not an estimate, the actual figure from your provider's invoices. If the median is in the tens of gigabytes, options three and five deserve a serious look and the enterprise platforms are being sold to you on capabilities you will not use. If the median is in the hundreds, or if a single matter has run into terabytes, you are in platform territory and the conversation is about which one.

Measure two: how many matters a year, and how lumpy. An annual platform commitment is a bet on volume. Count the litigated matters that produced any e-discovery spend in each of the last three years. If the count swings by more than about half between years, a fixed commitment is a bet you are likely to lose in the low year, and the per matter arrangements exist precisely for that shape.

Measure three: what proportion of e-discovery spend you actually recovered. Pull the disbursement lines and the write-offs. This is the measurement that decides whether e-discovery is a revenue line or an overhead line for your firm, and it is the one most likely to differ from what people assume in the meeting. If recovery is high and consistent, per matter pricing is cheap because the client is paying. If a meaningful share is being written off in fee negotiations, every dollar of platform cost is coming out of partner profit and the calculus changes completely.

Measure four: who owns it today, and what happens when they leave. In a firm of this size the answer is often one paralegal or one litigation support person who learned it on a bad case. That is a single point of failure on an obligation with a court date attached. If the answer to "who runs this if that person is on holiday" is silence, that is an argument for options one or four regardless of what the volumes say, because you are buying an institution rather than a tool.

If those four measurements point in different directions, the tie-breaker is the second one. A commitment sized to a good year is the hardest of these mistakes to unwind, because you carry it for twelve months whether or not the matters arrive.

We commission custom AI systems for mid-market operators, so the expected move at this point in the page is to argue that a firm should own its review platform instead of renting one. We are not going to, because it would be wrong.

Do not commission a review platform. Start with the processing and imaging layer, which has to ingest whatever file formats arrive in whatever state they arrive in, and which every vendor on this list treats as a headline capability rather than a detail. Above it sit near-duplicate detection, threading, predictive coding with a validation protocol that has to hold up when challenged, production sets in the formats opposing counsel will specify, redaction, and an audit trail built to survive someone whose job is to attack it. This is the clearest example we can point to of a category where the commodity is genuinely better than anything a custom build would produce, and where the failure would be discovered in front of a judge. The whole point of our what we do not build list is to say so before the invoice, and this belongs on it.

What is worth owning is the layer above the platform, and no vendor on this page sells it, because it is not a product. It is the firm's own decision record. Concretely, four things, none of which is a review engine:

• Matter triage against your own numbers. A recorded, consistent decision at intake about which arrangement a new matter goes into, driven by the measurements above rather than by whoever picks up the phone. The output is a routing decision and a written reason, on every matter, retrievable later.
• Legal hold tracking against your actual matter list. Not a hold product, which the platforms sell perfectly well, but the reconciliation: which open matters have a hold, which custodians were noticed, when the reminder went, and which holds should have been released and were not. Rule 37(e) turns on reasonable steps, and reasonable steps are a record.
• Cost recovery reconciliation. Vendor invoice to matter to client bill, with the gap visible while it is still recoverable rather than at year end. This is where the money is, and it usually lives in a spreadsheet whose accuracy depends on one person remembering to update it.
• The proportionality file. Your own per gigabyte cost, your own review rate, and your own volume history, assembled so that a proportionality argument under Rule 26(b)(1) can be made with figures rather than adjectives. The rule makes burden and expense an express factor. Very few firms can produce their own numbers when asked.

That work sits on the systems you already run rather than replacing them, which for a litigation practice usually means iManage or NetDocuments on the document side and Clio or Litify on the matter side. It is a focused single-system build in our terms: 45,000 to 65,000 dollars over four to five weeks, following a working prototype on your firm's real data inside 7 to 10 days, before any fee is owed. An end-to-end workflow rebuild across two or three systems runs 75,000 to 120,000 dollars over six to eight weeks, inside the same 45,000 to 180,000 dollar fixed-fee band the rest of our legal work is priced in. The code is yours at handoff.

We have not commissioned an e-discovery system for a law firm. Our legal work has been on intake, conflicts, knowledge retrieval and document workflow, and this page is research into what the vendors publish and what the rules require, not a case study. If you read all of the above and conclude that the right move is to call two service providers and keep doing what you are doing, that is a legitimate outcome of this page and it costs you nothing. The law firm resource and its twelve-question diagnostic are the shorter route if you want a number on the rest of your workflow first.

Every vendor statement on this page was read on that vendor's own live site on August 22, 2026, fetched over public HTTP with a declared research user agent. Every status code in the table is one we recorded in the same pass. The three rules are quoted from the rule text, not from a secondary summary.

What we did not do, stated plainly so nobody over-reads the page. We did not obtain a quotation from any of these vendors, so we cannot tell you what any of them charges per gigabyte; the finding is that they do not publish it, which is a different and smaller claim. We did not test any of the products. Vendor performance claims are reported as vendor claims and are not independently verified here, including Everlaw's statement that its users reduce documents promoted to active review by 74 percent with early case assessment, its claim that Coding Suggestions achieves recall and precision equivalent to or surpassing a first-pass human review, DISCO's claim that Auto Review reviews up to 32,000 documents per hour with precision and recall 10 to 20 percent above human reviewers, and Casepoint's claim that only six software-as-a-service companies in the world hold Department of Defense IL6 authorisation. Each of those may well be true. None of them was measured by us.

One relationship is reported from primary evidence rather than an announcement: logikcull.com's own navigation and footer route Newsroom, Partner Program, Careers and Support to revealdata.com, and its support documentation lives at docs.revealdata.com. We report what the site does rather than characterising a corporate transaction. The CS Disco filing details are read from the company's own SEC EDGAR record.

Related reading.

On the rest of the legal stack: what we would build for a law firm, the consultants mid-market firms actually shortlist, and the wider list of AI consultants for law firms. On the research and drafting layer this page deliberately hands off to, Harvey against a custom build, Legora against Harvey, and the citation checkers and what each one actually checks. On the litigation neighbours, EvenUp and demand drafting, Filevine and case management, and Smith.ai and the intake layer. On the systems the work above would sit on, where Clio's own AI stops. On governance, which is where a hold-tracking record actually lives, bar rules and malpractice and answering client AI questions in outside counsel RFPs. On the buy-or-build arithmetic underneath all of it, build, buy or commission, renting AI against owning it, and hiring against commissioning. On process and cost, how to choose an AI consultant for law, how we work and our own fee bands. Our forward view is in the 2027 law firm benchmark, the maturity assessment places your firm before any of it, and the rest of the series sits on the comparisons hub. If you would rather talk it through than read another shortlist, talk to us.