Introduction

E-discovery costs are spiraling ever higher, posing a significant challenge for companies faced with litigation and regulatory investigations that require extensive data collection and review. Although market forces have driven data processing costs down as much as 90 percent since 2003,^[2] little has been done to address the root cause of the overall cost burden: the collection, processing, and review of too much data, much of it irrelevant or nonresponsive.

C-level executives, corporate counsel, and even judges recognize that e-discovery can no longer be conducted outside a framework of cost controls, given the excessive amounts being spent on inefficient data processing and review. Fortunately, the Federal Rules of Civil Procedure now require parties to meet and confer regarding e-discovery protocols. This further encourages cost-conscious litigants to better understand their records management protocols, including backup schedules, IT systems, and document repositories. Some corporate respondents have begun to use statistical sampling and other documentable metrics to estimate the total cost of requested productions and support arguments for excessive burden and cost shifting.

The problem of too much irrelevant or nonresponsive data should be addressed first at its source – failed records and information management – and, second, by rethinking and reengineering the e-discovery response. The challenge associated with records and information management took decades to develop and could take some time to solve. This paper provides suggestions for approaching this challenge and its impact on discovery.

This paper also addresses the complexity and severity of the cost problem and suggests that a fundamental shift needs to occur in how e-discovery is executed. We describe an approach we call Smart Evidence and Discovery Management (Smart EDM), which regards e-discovery as a business process and draws on such disciplines as Lean Manufacturing (Lean) and Six Sigma^TM,^{[3] [4]} to drive down costs.

Fundamental to Smart EDM is an iterative rather than linear approach to e-discovery processes that seeks to identify responsive data as close to its source as possible. What makes it smart is that on any given discovery project, these processes support learning. Using iterative techniques, such as sampling, data stratification, prioritization of custodians and data sources, and targeted collections, counsel can make adjustments as they learn more about the people, terminology, and related issues. The potential for companies is a smaller corpus of collected data, improved relevancy, reduced review time, and, ultimately, lower costs.

The Problem: Too Much Data

For many years, digital storage costs have been declining at an accelerating pace. Online storage that used to cost hundreds of thousands of dollars to purchase 25 years ago now costs but a few pennies.This precipitous drop has led companies to keep buying more storage without facing the time-consuming and expensive process of developing and implementing an effective records and information management (RIM) program. companies should be assessing their long-term return on investment of a fully implemented RIM program compared to the higher costs stemming from litigation or an investigation when no RIM program exists. These costs include storage as well as searching, collecting, reviewing, and producing potentially responsive data.

Records and Information Management: What Should Be Considered

Companies that lack an organized records management system must figure out what corporate documents exist, whether they are potentially responsive to a specific litigation or investigative trigger, and where they are located.

For many years, the save-it-all approach has led to extraordinary volumes of document stores. This has been further exacerbated by a lack of RIM programs to guide the document life cycle from creation through destruction. The lack of a proper RIM infrastructure increases the possibility of deleting a valuable record, or one subject to a hold order.

Digital information – including customer and financial databases, regulatory archives, Microsoft Office files on file shares, image files, and e-mail – is collectively referred to as electronically stored information (ESI). ESI is not paper in digital form. It represents a new paradigm with a set of technical and operational challenges that did not previously exist. A corporate employee involved in a litigation or investigation as a witness might have ESI on a work computer, home computer, thumb drive, cell phone, or PDA, as well as on an individual file share on the corporate network, departmental network, or shared work site. The challenge of determining where this ESI is located, how it can be accessed, who manages it, how it is backed up, how relevant it is, and how it can be identified and preserved requires legal, technical, and forensic competence.

ESI is far more voluminous than paper and continues to proliferate. For example, a small thumb drive can hold the equivalent of 250 banker’s boxes of paper. For corporate data, however, the notion of cheap storage is anything but when one takes into account the combined costs of power consumption, requisite redundancies, disaster recovery and business continuity, data availability, and staffing. These costs increase even further due to the inefficiencies of searching and retrieving specific documents and data from voluminous and uncontrolled data stores.

Few organizations have implemented RIM policies and procedures in a way that addresses the complexity and risks of the digital world. Ineffective policies leave content – that could have been systematically and legitimately deleted – as fair game for regulatory or internal investigations or discovery-intensive litigation. E-mail produces a wide range of sensitivities, retention requirements, and risk, but there is often no methodology, protocol, or systematic support for managing it.

As a result, corporations must seek proactive ways to create new or enforce existing content creation policies and document retention schedules (thereby allowing the permissible destruction of data) as well as seek ways to eliminate outdated and unnecessary legacy or archival data.

The Duty to Preserve and Litigation Hold

The duty to preserve data for potential litigation is a contentious legal area and produces risk for corporate defendants. Precedents suggest that once a party reasonably anticipates litigation, it must suspend its routine document retention and disposition processes and establish a litigation hold. Still, the law recognizes there are limits to what an organization can reasonably preserve.^[5]

The best time to consider retention-destruction tradeoffs and litigation hold execution is when developing information life-cycle policies and procedures, not after the duty to preserve takes effect. A sound RIM policy can mitigate potential risks from inadvertent data deletion, since, per Federal Rules of Civil Procedure,^[6] the court may not impose sanctions when a party deletes or otherwise destroys ESI “as a result of the routine, good-faith operation of an electronic information system.” However, organizations and their counsel are expected to have a high level of competency in these matters. The prerequisites to obtaining such protection include clear and enforceable policies and controls relating to litigation holds and an understanding of the technology drivers.

The Demand for a Better Way

While the duty to preserve may appear to be broad, there is no reason the collection, processing, and review of data need to be equally broad. Nevertheless, these activities have historically had an unnecessarily high impact on the total cost of litigation. Having saved all its data, the organization preserves it all, the e-discovery service provider collects and processes it all, and the reviewers (billing by the hour, document or page) review it all – even though much of the material is irrelevant or nonresponsive. Over the past several years, KPMG’s analyses of selected databases containing millions of documents reviewed by attorneys have revealed that as much as 90 percent of the documents reviewed were non-responsive. In other words, the armies of contract attorneys that companies deploy for reviews are reviewing predominantly irrelevant or nonresponsive documents. Compounding this situation is the demand by those managing the review to provide the reviewers with as many documents as possible so they are never idle. This is not only inconsistent with the concept of pull from Lean manufacturing, but perhaps the epitome of what Lean and Six Sigma refer to as waste and is a key driver to unnecessarily increased cost.

In Lean parlance, the traditional e-discovery method is one based on push, where the producing party receives a request and proceeds to collect, process, review, and produce the data in a linear fashion by pushing it through the system. The opposite and more desirable method is pull, where the activity is tightly connected with demand. Creating an e-discovery process that leverages the Lean concept of pull is a challenge, and Smart EDM focuses on providing increased efficiency and productivity to reduce the risk and cost of the discovery process.

Smart EDM

Consistent with Lean manufacturing processes, the practical implications of Smart EDM are fundamental. Organizations cannot rely solely on unit cost reduction to achieve necessary savings but must instead drive savings through the smart use of process and technology to collect less data, process less data, and review less data. We discuss a number of key principles below.

EDM as a Business Process

Given the significant potential expense of e-discovery, organizations cannot allow service providers to operate carte blanche. In-house and outside counsel must understand the e-discovery process in the context of legal compliance and the current matter, and they also must be answerable to the CFO. In short, e-discovery must be comprehended operationally and financially and be subject to standards of accountability, efficiency, and effectiveness like any critical business process.

Those who are driving the discovery process still often believe that more is better, i.e. the more data, the greater the potential for discovery. However, the volumes of data kept by companies creates the potential to over-collect and consequently over-review and produce, leading both the respondent and complainant to look for a better way. A sound defensible discovery response strategy, supported by a repeatable, reproducible, and well-documented work flow, can help manage the resulting tension and costs.

Iterative Approach

As with most legal issues, a one-size-fits-all approach is ineffective. The best discovery process uses a consistent approach that considers the circumstances of the specific matter and documents the decisions involved in searching and reviewing relevant and potentially responsive documents. The choice and efficacy of particular search terms must be well thought out and tested. Failure to test search terms or sample the relevance of retrieved documents may lead to the collect too much, review too much syndrome. In one notable case for EDM, the judge makes a pointed reference to this very topic:

[C]ommon sense suggests that even a properly designed and executed keyword search may prove to be over-inclusive or under-inclusive, resulting in the identification of documents as privileged which are not, and not-privileged which, in fact, are. The only prudent way to test the reliability of the keyword search is to perform some appropriate sampling of the documents determined to be privileged and those determined not to be in order to arrive at a comfort level that the categories are neither over-inclusive nor under-inclusive.^[7]

We suggest that sampling is prudent and uses common sense throughout the process continuum we call e-discovery. Companies that use trial and adjustment, testing, measurement, and process improvement can help meet the cost and procedural challenges of e-discovery, regardless of whether it is devising litigation hold and preservation protocols or developing collection and culling strategies.

Such testing is at the core of Lean and Six Sigma. The Six Sigma DMAIC methodology (the acronym stands for define, measure, analyze, improve, and control) emphasizes a learning, iterative approach that we believe is readily adaptable to KPMG’s e-discovery process:

• Define the problem, the scope, and key issues, risks, deadlines, and [litigant] needs for the discovery review and how the legal team plans to address them
• Assess what measurement tools will be used on the project and how data will be gathered
• Analyze the data to identify potential root causes and track “as is” performance
• Develop improvement plans to address key issues
• Develop a control plan for maintaining the improved results over time.^[8]

Likewise, Lean considers the use of resources for any purpose other than the creation of value to be wasteful and, thus, a target for elimination. Lean requires that processes be designed to be as efficient as possible – to reduce cycle times and eliminate wasted efforts. “Waste” is defined as any activities that are not required and do not add value to the desired result. In short, Lean is about process efficiency, and Six Sigma is about process effectiveness – both useful concepts in managing the costs of e-discovery.

The Seven Wastes of Lean

Among the Seven Wastes cited by Lean, these three are particularly applicable to e-discovery:

Waste of Inventory: Redundant data sources or multiple instances of a single document add risk and cost but not value and are contrary to sound RIM policy.

Example: Multiple copies of documents on multiple systems

Anecdotally: Large organizations store a single fact of data more than 10 times.

Waste of Overproduction: Overproduction comes in three forms: redundant systems, duplicate records, and hidden information factories, also contrary to a sound RIM policy.

Example: Redundant systems – more than one system that captures customer information

Anecdotally: A typical mid-sized organization has 16,000 hidden information factories.

Waste of Defects: Excessive documents identified as potentially responsive by poorly defined search criteria

Examples: Files that contain keywords but aren’t relevant or files that are resistant to processing that are discovered during processing because of inadequate data cleansing or correction.

Perhaps mistaking simplicity for speed, the traditional linear approach to e-discovery tends to define the entire EDM process at the outset before sufficient learning about the matter has occurred. Once the process for the matter is defined, teams move through their assigned steps in full batches, without the benefit of being able to adjust upstream processes according to downstream results.

For example, data preservation and collection are often completed before a full multi-level review has been performed for any sampling of the data (or custodians). Logically, how a company preserves data should affect how it collects that data, which should affect how it processes the data, and so on. Yet traditional e-discovery approaches provide limited or no opportunity for upstream adjustments that could affect downstream processes. Lean engineering refers to this as “Waste of Defects.”

An iterative approach tactically leverages technology within a strategy of smarter process. For example, a targeted collection may at first be confined to key systems and personnel, whose data is then filtered and sampled for efficacy of search terms, dates, and issues before moving on to the wholesale collection. Low relevancy rates may result in recalibrating search terms and/or reconsidering relevant custodians and the value of data from certain systems.

The approach is continuously refined from the outset. Perhaps new custodians are identified, new systems targeted, or simply more effective search criteria are identified, and then the collection process continues applying what is learned, creating a controlled process that becomes continually smarter.

The use of sampling in this manner can directly help control costs. This is simply the “measure twice, cut once” carpenter’s admonition and is a significant departure from the collect-it-all, process-it-all, review-it-all-three-times-for-privilege constraints of traditional e-discovery workflow.

Forensic Imaging versus Targeted Collection

Traditional data collection entails full bit level imaging of custodians’ hard drives. Rather than a file copy, this method creates an exact image of the entire hard drive, including empty space, file fragments,^[9] systems files, and programs. In other words, if a company has a witness whose laptop contains a 60 gigabyte hard drive, 60 gigabytes are collected.