First-pass document review is a relatively recent phenomenon that has become increasingly common (and expensive) as the volume of email stored on corporate servers has exploded. The purpose of first-pass review is to use a less expensive labor force to take an initial look at the documents collected in order to identify those that must be looked over by the senior attorneys on the case team. No one is asking the first-pass team to make nuanced legal judgments; the goal is simply to save the case team from looking at documents that couldn’t possibly be relevant.

Even though labor rates for first-pass reviewers are relatively affordable, the large volume of documents has driven annual costs for first-pass review into the millions of dollars for corporations with active litigation portfolios. Can anything be done to control these expenditures?

The only significant way to directly control first-pass review costs is to limit the number of documents that are actually read by the reviewers. One current approach is to replace first-pass reviewers with predictive coding algorithms. There are pluses and minuses to this approach, but a review of those is not the subject of this article. Rather, we offer an alternative approach that will reduce the cost of first-pass review by 75 percent without using any AI algorithms.

As stated previously, to significantly reduce the cost of first-pass review, fewer documents must be run through the process. There are two strategies that can be employed to achieve this result; both require the same subtle shift in perspective. Instead of thinking of the task as document review, think about the task as language review.

Actually, this shift is just a more accurate statement of the process. When we review a document, what are we actually doing? We are looking to see if the document contains language that causes it to be relevant to a case issue. From this perspective, document review is the hunt for relevant language. For a document to be relevant, it must contain “strings” of language that are relevant.

Now that we have adopted this frame of mind, there are two workflow changes we can make to reduce the cost of first-pass review:

  1. Identify documents that must be relevant or could not possibly be relevant without actually reading them.
  2. Read only one copy of each relevant language string.

Identifying documents that either must go to the case team or that couldn’t possibly be relevant (and therefore can be ignored) is very much like a keyword exercise, only without the pressure to get the hard choices right.

We know at the outset of every review that if certain words or strings of words appear in a document, the case team will have to look at it. There is no need to send those documents through first-pass review. Similarly, for a document to even potentially be about a given topic, there are concepts it must contain, and those concepts are communicated by words. For example: If you are looking for documents about a certain kind of meeting, then for a document to possibly be about that topic it must contain at least one word that can communicate the concept of a meeting. If it contains no such word, then the document cannot be relevant and need not be reviewed. The existence of the word doesn’t mean the document is relevant, but it is worth reviewing.

Engaging in this simple logical “what words are a must” exercise can greatly reduce the number of documents that are subject to first-pass review. Our experience is that several hours of logical thinking will cut the number of documents subject to first-pass review by 60 percent or more — an astounding return on the effort.

Let’s turn to the second strategy by asking a question: In theory, how many different ways are there to talk about the same subject? Given the number of synonyms and euphemisms in English (and most other languages), the answer is effectively incalculable. But let’s ask a different question: In a given document collection, how many different ways do people actually talk about the same topic? The answer is never more than 30 and usually a dozen at most. Language is a social phenomenon, and homogenous populations share their language choices. Relevant language is dramatically redundant in any document collection, and you can take advantage of that redundancy to lower costs.

When a first-pass reviewer comes across language in a document that he/she believes must be seen by the case team, it is fair to say that every other document containing that same language must also go to the second pass. In a large collection there will be dozens of documents that contain the very same relevant language the reviewer is seeing. A simple Boolean search engine can easily identify those documents. Run the search, bundle the documents and send them to the second pass with the relevant language highlighted. Our experience is that by working this way, you will cut the cost of first-pass review by at least 15 percent.

To summarize, first-pass review has become both prevalent and expensive as the volume of email stored on corporate servers has soared. Shifting one’s perspective of this process, from examining documents to hunting for relevant language, will significantly reduce time and costs. Savings on the order of 75 percent are readily achievable by making the two proposed changes to the first-pass workflow.