Early case assessment: deciding what a matter will cost before committing to it

Early case assessment, or ECA, is the work performed on collected or sampled data before full processing and review. Its purpose is to establish what the available information appears to contain, how much work a complete response may require, and where legal or operational exposure may be concentrated.

ECA is often compressed when a matter arrives with immediate pressure to preserve data, identify custodians, and begin review. That sequence may satisfy urgent procedural needs, but it can leave the producing party without a reliable basis for estimating cost, testing search terms, or making informed proposals at the Rule 26(f) conference.

ECA is not early data assessment#

ECA and early data assessment can overlap, but they answer different questions. Early data assessment concentrates on narrowing the data universe: identifying a defensible sample, excluding sources that are unlikely to add value, and defining the collection that can support an early assessment.

ECA begins where that narrowing has produced at least some usable material. It characterizes the data to estimate scope, cost, and exposure before the matter commits to full processing, document-level review, and production. The distinction matters because reducing data is a collection decision, while understanding the consequences of that data is an assessment decision.

Neither label changes the need for legal judgment. A small sample can support a useful estimate only when the collection method and its limitations are understood. A large collection does not create confidence if duplication, missing sources, or an unrepresentative cut makes the resulting counts misleading.

What ECA produces#

A useful ECA report translates a collection into a set of concrete working assumptions. It should explain what was sampled, what was excluded, which processing steps were applied, and what the figures mean. The following outputs give counsel and litigation support a shared factual basis for early decisions.

Volume estimates connect the initial sample to the possible size of the complete population. They also expose the effects of expansion: additional custodians, date ranges, or data sources can change the review population after processing, de-duplication, and email threading. The estimate is therefore a planning tool rather than a substitute for reconciliation against the final collection.

Custodian and date distributions show where information is concentrated. A domain profile identifies the systems represented, while a file-type profile shows whether the collection consists largely of email, images, structured files, or another format. These distributions help practitioners distinguish a genuinely narrow issue from a broad collection that merely appears manageable at the outset.

Candidate-term hit reports provide another layer. Counts can show whether a proposed term appears frequently, whether its result set is dominated by a small group of custodians, and whether related variants retrieve different material. Hit counts do not establish relevance, but they give both sides concrete information to discuss instead of exchanging unsupported assertions about search adequacy.

The early read is a bounded review of selected documents, not a substitute for issue-by-issue review. It can reveal recurring factual themes, the prominence of particular custodians, and the presence of documents that may raise privilege, work-product, confidentiality, or redaction concerns. It can also show where the documents do not support the assumptions with which a party entered the dispute.

The decisions ECA informs#

ECA earns its cost by changing decisions, not merely by producing another report. The commercial value lies in replacing generalized warnings about scope with information that can be tested, challenged, and revised.

One decision is whether early settlement deserves further consideration. An early read may clarify the strength of the documents, the concentration of damaging material, and the gaps created by incomplete collection. ECA does not determine whether a case should settle. It provides a more informed basis for deciding whether the early read is sufficient to support that decision.

Another is what to propose at the Rule 26(f) conference. Rule 26(f) requires the parties to confer before the scheduling conference, and Rule 16(b) governs the scheduling order. ECA can give counsel concrete material for that exchange: likely custodians, search-term performance, estimated review volume, and the relative access that Rule 26(b)(1) includes among the proportionality factors.

Search terms are a particularly useful subject for early testing. A term that appears productive in one collection may generate noise in another because of differences in vocabulary, document types, or custodian behavior. Iterative hit reports allow the parties to identify narrow, overbroad, or redundant terms before committing to production review.

ECA also informs proportional custodian scoping under Rule 26(b)(1). The amendment placed the proportionality factors within the definition of the scope of discovery itself and added the parties’ relative access to relevant information. Custodian and communication-pattern analysis can turn that abstraction into a source-specific discussion. The relevant comparison is between information available in one party’s collection and information accessible to another party, not a generic assumption that all custodians are equally necessary.

Finally, ECA supports the budget decision. A budget based on an assumed document population is less useful than one tied to sampled volume, anticipated expansion, processing requirements, and realistic review patterns. The estimate will still change as facts develop, but it gives the person approving expenditure a basis for identifying which assumptions carry the most risk.

How ECA is performed in practice#

Sampling and baseline characterization#

The assessment normally starts with a sample from the collected data. The sampling design should connect to the question being evaluated: custodians, date ranges, sources, file types, or known search terms. A useful report records the selection method and any limits so that later reviewers can distinguish a property of the evidence from an artifact of the sample.

Processing should be sufficient to make the sample meaningful. De-duplication can reveal how much content consists of repeated documents, while email threading can expose the relationship between a message and its surrounding exchange. Near-duplicate detection can identify material that is not identical but belongs to the same document family. Without a defined processing baseline, raw file counts may overstate the number of distinct documents requiring review.

The sample is then profiled by custodian, date, domain, and file type. This baseline may already alter the collection plan. A source that appears to contain few responsive documents may still be important because it is difficult to access elsewhere, while a large source may contain material substantially duplicated in another system. Relative access is therefore a collection question as well as a proportionality question.

Iterative term testing#

Candidate search terms are tested against the sample and refined using hit reports. The objective is not to declare terms complete during ECA. It is to learn which terms retrieve distinct material, which retrieve the same documents repeatedly, and which are too broad to support an efficient review.

Each iteration can compare the proposed term with variants, related concepts, and existing source metadata. The hit report should show enough distribution information to reveal whether a term’s results are concentrated, repetitive, or disconnected from the issues under examination. Counsel can then evaluate the term against the documents retrieved rather than relying on the count alone.

This process also creates a record for later negotiation. If a term is revised, the parties can see what prompted the change and whether the revision narrowed the result set. That record is more useful than a final list presented without an explanation of how it was developed.

Concept clustering#

Concept clustering examines groups of documents that appear to address the same subject. It can expose terminology that did not appear in the initial issue list, distinguish competing descriptions of an event, and identify factual themes that warrant targeted searches.

Clustering complements, rather than replaces, Boolean or keyword search. Keyword search depends on the vocabulary selected in advance. Concept analysis can reveal that documents describing the same activity use different terms, allowing the search plan to incorporate those expressions before full review.

The clusters should also be checked against known documents and custodian distributions. A cluster may represent a genuine issue, a recurring administrative phrase, or an artifact of similar document structure. Those possibilities cannot be separated by an automated label alone; the early read provides the context needed to interpret it.

Communication-pattern analysis#

Communication-pattern analysis compares relationships among custodians and the data they generated. In ECA, that comparison can identify concentrated exchanges, custodians connected through several channels, and sources whose apparent volume does not match the number of distinct conversations.

The analysis helps test custodian-scoping assumptions. If one custodian appears central to multiple channels, excluding that source may carry more risk than its document count suggests. If a custodian’s material largely duplicates material held elsewhere, narrowing the custodian list may have a different effect on burden and completeness.

Patterns also help frame the early read. A document’s relationship to other communications may clarify its context without making the document irrelevant by itself. The practitioner still needs to distinguish a useful communication path from a mere coincidence in the metadata.

The honest limits of ECA#

ECA is an estimate based on an evidentiary snapshot, not a warranty about the complete collection. An incomplete collection can mislead in either direction. Missing sources may contain the most relevant material, while included sources may produce counts that do not generalize to custodians or date ranges outside the sample.

The report should therefore state what remains unknown. Collection gaps, omitted sources, processing choices, and unresolved questions about search terms belong alongside the estimates. If those limitations disappear between the ECA report and later reporting, the apparent precision of the numbers becomes misleading.

Hit counts are not relevance counts. A low count may reflect an uncommon but decisive term; a high count may reflect repeated boilerplate. Counts become useful when connected to document content, custodian distribution, and the candidate term’s role in the matter.

The early read carries a separate risk: an interpretation can harden into an assumption. Once a team concludes that the sampled documents establish a fact, later reviewers may test new material against that conclusion instead of the original issue. The read should remain provisional, tied to the sample that produced it, and revisited when the collection or the legal issues change.

ECA should also avoid becoming an informal substitute for review. A selected read can identify themes and risk areas, but it cannot supply a defensible basis for calling every responsive document responsive or nonresponsive. The objective is to improve the decisions that precede full review, not to create confidence that full review is unnecessary.

Negotiation value#

A party that understands its own data arrives at the meet-and-confer able to discuss proportionality with specifics rather than adjectives. It can identify the custodians that matter, explain the search terms being proposed, quantify the expected review population, and distinguish meaningful sources from duplicated or marginal material.

That position supports negotiation because each assertion can be examined and tested. The other side can evaluate the sampled basis, question the expansion assumptions, propose narrower terms, or suggest a source that changes the relative access analysis. ECA does not remove disagreement; it gives the parties better material for resolving it.

The strongest ECA deliverable is therefore not a single forecast. It is an explicit, revisable set of findings about volume, distribution, search performance, document content, and uncertainty. When those findings are tied to the collection and communicated clearly, they allow scope, cost, and exposure decisions to be made before the matter becomes committed to a processing and review path.