DISCO eDiscovery: the per-document model explained

DISCO’s pricing page references a per-document pricing unit but publishes no price figure. It directs prospective buyers to contact the company, so any cost estimate requires a quote tied to a specific matter profile. This page was checked in November 2026.

The central point is narrow but consequential: DISCO does not publish a price on the page reviewed. The published unit indicates how a quote may be structured, not what it will cost. Verify current pricing and capability with the vendor because pricing pages and product documentation change.

This is not a review, and nothing was tested. The discussion below explains the per-document model, identifies the variables that affect a document count, and sets out questions buyers should resolve with DISCO before comparing quotes.

What the per-document model means#

A per-document model charges according to the number of documents in the population being priced. In the model described here, the relevant population is the set of documents after processing. That differs from a per-GB model, which starts from the volume of data ingested or processed before document-level reductions have an effect.

The distinction changes which controls matter. Under a raw-volume model, reducing file size, excluding unnecessary files, or narrowing the source set may have an immediate effect. Under a per-document model, the more direct levers are document-level: how many documents survive processing, how duplicates are handled, and whether related documents are grouped or counted individually.

De-duplication can reduce the bill because documents identified as duplicates may not remain in the final count. Threading can have a similar effect when messages in a conversation are treated as a family rather than as separate documents. The effect depends on how the vendor defines the priced population and applies those functions; the pricing page alone does not establish the precise counting rules.

Date and custodian scoping can reduce both the volume processed and the number of documents counted. A narrower search for custodians, sources, or date ranges may remove whole collections before processing. It may also reduce the number of responsive or produced documents later in the review, but those are separate populations that should not be assumed to drive a quote.

That sequence is the key to understanding the model. A common workflow moves from collection and processing to de-duplication, search, review, and production. A per-document quote generally follows the count at the point defined by the contract or quote, not automatically the number of files collected, the number of pages reviewed, or the number of documents produced.

Why culling matters differently#

Consider a collection containing 1 million documents, including extensive exact and near-duplicate families. If processing and de-duplication reduce that set to 300,000 documents, a per-document quote based on the processed population would use 300,000 units rather than the original million. A per-GB quote would still depend on the volume handled and on the vendor’s treatment of that volume.

Threading presents a similar question. If 50,000 email messages are grouped into 10,000 conversations, the document count may fall by 40,000 if families count once. If each message counts separately, the reduction is smaller or absent. Attachments, embedded files, and image-only records create further questions because the same content can appear in several forms.

Date and custodian scoping works through the workflow rather than merely filtering the final set. Removing irrelevant custodians before processing can reduce ingestion, processing, de-duplication, and review. Applying date filters after processing may remove documents from search results but leave them within a population that the vendor prices.

The practical implication is that buyers should establish when the count occurs. “Per document” is not a complete pricing definition. A quote may reflect documents loaded, documents retained after processing, unique documents after de-duplication, or another expressly defined population. Each basis creates a different response to data shaping.

Where the model tends to fit#

A per-document model can fit a matter with a relatively predictable document population. The unit may be easier to frame when the collection size, source types, duplication rate, and expected processing scope can be estimated in advance. It also gives a defined question for scoping: how many documents will count, and at what point?

Matters dominated by structured custodial data may behave differently from matters with heavy duplication or expanding processing. A small percentage change in the number of documents can produce a larger change in the total than the same percentage change in raw gigabytes, particularly when large files and many small records coexist in the same collection.

The model becomes less stable when processing expands. New custodians may require additional sources. Family relationships may reveal additional messages or attachments. De-duplication may identify overlap with a larger collection rather than with a discrete data set. If the priced document count rises after each expansion, the buyer needs a mechanism for revising the quote and approving the change.

This is not a categorical judgment about suitability. It is a matter-profile distinction. Predictability favors a count-based model; large or uncertain expansion makes the definition and timing of the count especially important. Data volume alone does not answer either question.

Why per-document and per-GB quotes are not directly comparable#

A per-document quote and a per-GB quote use different denominators. Comparing them requires a data profile that states both the volume of data and the number of documents in the same population. Without both figures, a buyer cannot determine whether one proposal reflects a lower price or simply a different measure.

The profile should identify the source types, approximate file volume, document count before and after de-duplication, number of families, treatment of attachments, planned custodians, and date scope. It should also separate the initial processing population from later review or production populations. Those details matter because a vendor may count documents at a different stage than another vendor uses for its quote.

Buyers should ask each vendor to price the same profile. That means applying the same custodians, date ranges, data sources, processing steps, and document-counting assumptions to every proposal. If a vendor cannot explain the unit or its application to the profile, the quote is not meaningfully comparable even if a numerical figure appears in the response.

Ask for a written quote with the priced population, included services, usage assumptions, and expansion terms identified. The central verified finding across the major pricing pages reviewed in November 2026 was that none publishes a price figure and every one routes the reader to a sales conversation. A quote is therefore part of the pricing information, not a substitute for a published price that buyers can inspect in advance.

Questions to ask about a document count#

The first question is simple: what counts as a document? Possible definitions include an electronic file, an email message, a page, a family, an attachment, or a record created during processing. Each definition can produce a different total from the same source material.

Buyers should ask whether email messages and their attachments are counted separately. A PDF stored as an attachment may be one document, one attachment associated with another document, or another distinct item in the count, depending on the vendor’s rules and the processing configuration. Native files, converted files, image-only documents, and documents with embedded objects may also be treated differently.

Family treatment requires equally specific answers. Ask whether a family counts as one unit, whether each member counts, and how attachments are associated with family members. If a family contains multiple attachments with the same content across several messages, the vendor should explain whether those attachments are consolidated, de-duplicated, or counted separately.

De-duplication requires attention to both scope and method. Exact duplicates and near duplicates may receive different treatment. The count may change depending on whether comparison occurs within one custodian, across custodians, within a collection, or across the entire matter. Global de-duplication can materially reduce a count compared with collection-level treatment.

Ask whether the count shown before de-duplication is the same count used for pricing. Also ask what happens to documents removed as duplicates, how threaded messages are represented, and whether the count includes empty files, system artifacts, processing derivatives, or documents outside the search terms.

Finally, define the trigger for a change. A price based on a processed document population may change after a new source is added or an expanded collection reveals additional families. The contract should state whether an expansion receives a new quote, what notice the vendor must provide, and how the revised population is measured.

Making the quote operational#

A defensible comparison begins with a common data profile rather than a list of vendor rates. The profile should preserve the distinction between raw volume, processed volume, unique documents, families, and produced documents. It should state the proposed processing scope and the point at which each measure is taken.

For DISCO, the relevant published fact is limited: its pricing page references a per-document unit, provides no figure, and directs readers to contact the company. Buyers using DISCO in a bake-off should obtain a quote that applies the same profile used for per-GB proposals and document how DISCO derives the priced count.

Verify current pricing and capability with the vendor because pricing pages and documentation change. The practical objective is not merely to compare unit labels. It is to ensure that each proposal prices the same material, at the same processing stage, with the same treatment of duplicates, families, attachments, and scope expansion.