To say an overwhelming portion of documents involved in litigation or regulatory matters today are stored electronically is nearly becoming rote. How many documents are involved, how they're processed, in what format they're produced -- it's all old news.
With recent cases cropping up involving big-name companies, a more interesting conversation surrounds what mistakes are being made in the electronic discovery process and what happens with data before it is handed over to the other side.
What mistakes are we talking about, really? In the process of collecting, preserving, de-duplicating, filtering, culling, and reviewing ESI, opportunities for errors range from entering incorrect date ranges to inaccurately entering specific format requirements for tools used in downstream stages of the process. Not surprisingly, quality is a huge challenge for a discovery process that involves ever-growing volumes of data. The opportunity for mistakes, oversight, or simple carelessness comes from imperfect technology and the fallible nature of people who just can't guarantee 100 percent focus and attention to massive quantities of information.
Several high-visibility cases involving McDermott Will & Emery, Google, and Duane Reade, Inc. have drawn attention to quality control issues in the discovery process and highlighted how slip-ups can result in waiver of privilege and ultimately impact the outcome of the case. These cases reinforce the critical importance of quality control at the tail end of discovery before production to opposing counsel.
Inadvertent waiver of privilege is huge and costly in terms of damages paid and reputation tainted. But what are the other mistakes that law firms, e-discovery vendors and in-house counsel are making -- and how can they be readily fixed? The top 10 mistakes include:
1. Ineffective Redactions
Applying redactions can be tricky if you don't understand the process. Mistakes include failing to "burn-in" the redaction on the image, not updating or re-OCRing the text files to match, providing un-redacted native files, and failing to redact certain metadata. To remedy, re-review documents on your final production media to make sure the redactions are permanent. Compare the images along with the text. Check if the native is being produced -- and update it to reflect the redaction or remove it from the production. Inspect the metadata fields -- and ensure no privilege or confidential information is found.
2. Outdated Review Coding
A document may have been marked 'Privilege' by the review team, but if a request for that document to be produced occurred before a change in review coding, who knows about it? It's important to communicate changes, have a process to lock documents in the review system, and recheck the review coding against the final production media before sending it out the door.
3. Extra and/or Orphaned Files
Sometimes documents need to be manually removed from a production before it's sent off to opposing counsel. It's not enough to just remove the references from the load files. Have a process in place that checks all the file references in your load files and verifies those supporting files -- including images, text, and native -- can be matched against your load files. If you find any files that can't be matched, they may not belong in the production.
4. Wrong Starting Number
Your first two productions went smoothly, but things started to go wrong with the third production -- the "next bates number" for the third production was already used in the second -- and no one noticed until opposing counsel complained. Now you have documents with duplicate bates numbers, and every production from where the error started needs to be re-produced. Keeping accurate logs for tracking productions and using e-discovery software to compare bates numbers in past productions to new productions can flag duplicates or gaps before it's too late.
5. Missing Confidentiality
The production request has been made to the litigation support team -- and some documents have been designated 'C' for Confidential and 'HC' for Highly Confidential. Transferring the confidential coding and branding it on the images is another step in the process. Unfortunately, the team 'forgot' to do the branding step. By spot-checking the final production media -- and comparing the coding field that contains the confidentiality status against the production image -- the crisis could be adverted.
6. Wrong Production Specifications
Is everyone on the same page regarding the production specifications? Is it supposed to be single-page or multi-page images; TIFF or PDF; native files for all or just the spreadsheets; extracted text along with OCR; metadata fields included in the appropriate load files; and correct load files? If final productions meet the agreed upon production format, you will save time, money and embarrassment by not having to fix mistakes.
7. Not Producing Enough
If documents are being reviewed natively and converted to image, such as TIFF or PDF files, ensure all that can be produced is being produced. For example, spreadsheets may include hidden rows, columns, or sheets, or print areas might be set to exclude data. PowerPoint files may contain notes, or CAD files may have missing supporting files. It's possible the review team did not review everything until a document has been imaged -- or even missed relevant data that should have been deemed responsive. Your process should check the results of native files after they are converted to image.
8. Load Files Poorly Constructed
Transferring documents from one e-discovery platform to another generally involves the use of load files. Load files move a document's file references -- including any image, text, and/or native files -- along with metadata. But with no single format commonly used or requested, knowing the intricacies of each load file requires experience. Simulating and testing the loading process -- whether it's using the target platform or a tool designed for reading load files -- can quickly identify these mistakes.
9. Failure to Communicate
Communication among all groups involved can be paramount to an inadvertent production of privileged documents. For example, the final production search may include privilege documents that need to be numbered inline and then removed from the export. But if no one checks to ensure these documents were removed after being numbered, you may end up producing privilege. The team requesting the production should always check the contents of the final production to ensure the correct number of documents and search filters have been applied before sending to opposing counsel.
10. Assuming it's Right
Quality control measures should be in place among all the teams handling a document production, whether it's the operator running the production or the attorney signing off on the production before it's sent to opposing counsel. Assuming a production doesn't inadvertently contain privilege documents -- when you had the opportunity to inspect it beforehand -- can be a costly mistake. Utilizing quality control tools designed for e-discovery, especially for productions, can allow litigation support, attorneys and paralegals to effectively and efficiently check their productions.
Justin Blessing is the Director of Product Development for Compiled Services, makers of e-discovery software ReadySuite. He can be reached at Justin@compiledservices.com. More information can be found at www.compiledservices.com. The company can be followed on Twitter @CompiledSrvcs or on Facebook at www.facebook.com/CompiledServices.
In the case of predictive coding, its advocates have touted the efficacy of their approaches in white papers, blogs, and social media postings, and have practically created a separate industry to host conferences promoting the wonders of predictive coding. Few people want to ruin the moment or buck the trend by pointing out what is obvious when one considers the technology underlying predictive coding - it is completely dependent on having text to analyze. It will absolutely fail to analyze documents for which there is no text, and will do a miserable job where the text is of poor quality.
In practically all industries it is common practice to develop documents in one application like Word and then, once finalized, distribute them as image-only PDF so they can be viewed on a variety of devices and so recipients can't easily change the content. In one collection we analyzed, only 20% of the PDFs had associated text. Even if predictive coding were 100% effective, the most it could classify would be 20% because it literally cannot "see" the 80% without text. If in fact predictive coding has a recall rate of 70-80% of what it can see, that would mean that predictive coding would have identified 14 to 16% of the total PDFs (70% x 20% = 14% or 80% of 20% = 16%). By contrast, BR's visual classification technology classified 100% of them.
