A reference table of realistic all-in eDiscovery costs at five matter sizes — from a 5 GB employment dispute to a 500 GB commercial litigation — under traditional per-GB/per-document pricing and AI-augmented all-inclusive pricing.
The benchmarks below model five representative matter sizes at mid-market US rates as of early 2026. Each size reflects a distinct litigation context: a 5 GB employment dispute, a 20 GB regulatory inquiry, a 50 GB single-plaintiff commercial matter, a 100 GB mid-sized commercial dispute, and a 500 GB large commercial litigation with multiple custodians and claim types.
The assumptions held constant across all five sizes are as follows. Data collected equals 100% of the stated GB figure — no additional buffer for collection overhead. Documents after deduplication scale at approximately 2,500 documents per GB, which reflects typical enterprise email and document archives; data-intensive sources (large attachments, image-heavy files) would produce fewer documents per GB. Responsiveness rate is 6% of documents after first-pass review. Privilege rate is 13% of responsive documents, requiring privilege review and log entries. The review window is 6 months for all matter sizes.
The traditional vendor column uses: processing at $100/GB; hosting at $25/GB/month; first-pass review at $1.50 per document; privilege review at $6 per document; privilege log at $10 per log entry; production at $8/GB; and project management at $250/hr, scaled by matter size (approximately 8 hours for a 5 GB matter, increasing to 120+ hours for a 500 GB matter to account for team coordination overhead). The AI-augmented column uses DecoverAI's published rate of $60/GB/month, all-in, for 6 months. All figures are approximations. Actual costs vary with vendor negotiation, data complexity, review team composition, and matter-specific privilege volume.
The table below presents the all-in cost comparison across five matter sizes. Figures are rounded to the nearest thousand for readability. The cost reduction column reflects (traditional − AI-augmented) ÷ traditional.
| Matter size | Documents (est.) | Responsive (est.) | Privilege (est.) | Traditional all-in | AI-augmented all-in | Cost reduction |
|---|---|---|---|---|---|---|
| 5 GB | 12,500 | 750 | 98 | ~$31,000 | ~$1,800 | 94% |
| 20 GB | 50,000 | 3,000 | 390 | ~$103,000 | ~$7,200 | 93% |
| 50 GB | 125,000 | 7,500 | 975 | ~$228,000 | ~$18,000 | 92% |
| 100 GB | 250,000 | 15,000 | 2,000 | ~$460,000 | ~$36,000 | 92% |
| 500 GB | 1,250,000 | 75,000 | 9,750 | ~$2,100,000 | ~$180,000 | 91% |
Note: Traditional costs for 5 GB include fixed minimums — processing setup, project management onboarding, and platform configuration — that make per-GB costs disproportionately high at small volumes even before the per-document review charges apply. Traditional costs for 500 GB include team-scaling overhead on top of per-document rates: the review team required for 1.25 million documents runs 40–80 contract attorneys, with supervisory and QC layers that grow with team size rather than data size.
At small data volumes — 5 to 20 GB — traditional eDiscovery pricing is most difficult to justify relative to the work actually performed. A 5 GB matter with 12,500 documents typically involves an employment dispute or a small regulatory inquiry. These are matters where the amount in controversy is commonly $500,000 to $2,000,000. A party facing a traditional vendor quote of approximately $31,000 for eDiscovery infrastructure has spent 1.5% to 6% of the amount in controversy on document review overhead alone — before a single hour of outside counsel's time has been billed.
The AI-augmented cost of approximately $1,800 for the same 5 GB matter represents 0.09% to 0.36% of a typical amount in controversy. The difference is not a rounding error; it is a structural consequence of whether the pricing model includes a per-document review charge. At 5 GB with 12,500 documents, first-pass review at $1.50 per document is $18,750 — already the dominant cost component in the traditional model, and already larger than the entire AI-augmented budget for the matter.
The fixed minimum problem compounds this at small scale. Traditional vendors impose minimum charges for platform setup, project management onboarding, and processing infrastructure that are not negotiable below a certain floor. A vendor with a $5,000 platform minimum and a $2,000 PM onboarding minimum has already added $7,000 to the cost of a 5 GB matter before a single document is reviewed. These minimums are economically rational for the vendor — their fixed costs are real — but they reveal something important about the model: the rate card does not reflect the actual economics of the matter at small scale. Some vendors waive minimums to win small matters; if they do, the rate card was not accurate to begin with.
At large data volumes — 100 to 500 GB — the per-document model becomes exponentially dangerous for a different reason: review team coordination overhead grows faster than document volume, and the cost model does not make this visible until the matter is already underway.
A 500 GB matter with 1,250,000 documents after deduplication requires review teams of 40 to 80 contract attorneys running in parallel shifts to meet any reasonable review deadline. A review team of that size does not manage itself. It requires multiple team leads (typically one per 10–12 reviewers), a dedicated project management infrastructure, QC layers that must sample from a population an order of magnitude larger than in a 50 GB matter, and privilege review that requires senior-level resources on a volume — 9,750 privilege-designated documents — that cannot be processed by a small privilege team in a standard review window.
The $2,100,000 traditional estimate for a 500 GB matter in the benchmark table is, if anything, conservative for complex litigation. In matters with a high proportion of privileged communications — executive-level email chains in M&A disputes, for example, or in-house counsel correspondence in regulatory investigations — the privilege review line item can significantly exceed the 13% rate assumed in the model. A 20% privilege rate on 75,000 responsive documents is 15,000 privilege reviews at $6 each plus 15,000 log entries at $10 each — $240,000 on the privilege line item alone, versus the $117,000 assumed in the benchmark.
The AI-augmented cost for a 500 GB matter is $180,000: 500 GB × $60/GB/month × 6 months. That cost is the same regardless of whether the privilege rate is 13% or 25%, regardless of the size of the review team, and regardless of how many QC passes are required to validate the AI classifications. The per-GB rate scales exactly with data volume and does not carry a hidden escalation clause for complexity.
The benchmark table above has a practical application that goes beyond vendor selection: it gives litigators a concrete reference for FRCP Rule 26(b)(1) proportionality arguments. The 2015 amendments to Rule 26(b)(1) made proportionality a threshold test for permissible discovery, not merely an affirmative defense. A party whose review estimate is disproportionate to the needs of the case faces a court that is now formally empowered — and obligated — to weigh the burden of production against its likely benefit.
When a court asks whether a $460,000 eDiscovery estimate is proportional to the needs of a $1.2 million commercial dispute, the producing party who can present an alternative all-in quote of $36,000 for the same work has given the court something concrete to weigh. The party who cannot present an alternative has left the court with only the traditional vendor's estimate to evaluate — which means the disproportionality argument has no denominator. Courts are not required to accept that a cheaper alternative exists in theory; they respond to evidence that a cheaper alternative exists in practice and is available to the party making the argument.
This is the operational reason why published pricing matters. A vendor with a publicly available rate of $60/GB/month, all-in, gives a producing party's counsel a cite-able number that does not require an expert declaration to establish. A vendor whose pricing is disclosed only after an SOW negotiation cannot provide the same foundation for a proportionality brief. The practical value of price transparency in eDiscovery is not primarily competitive — it is evidentiary. For a fuller analysis of how the post-2015 proportionality case law applies to eDiscovery cost arguments, see Proportionality in eDiscovery: How Courts Are Redefining "Reasonable."
The key insight from the benchmark table: at every matter size from 5 GB to 500 GB, the cost differential between traditional and AI-augmented pricing is 91–94%. The percentage is stable across the full range because the gap is structural, not size-dependent — it comes from the per-document first-pass review line item that AI eliminates at every scale. The fixed minimums slightly compress the AI advantage at 5 GB; team-scaling overhead slightly compresses it at 500 GB. Neither effect is large enough to materially change the comparison.