A sexual harassment case in Atlanta recently used the two-tiered analysis of Rule 26(b)(2)(B) to reject plaintiffs’ over broad discovery request. Petcou v. C.H. Robinson Worldwide, Inc. 2008WL 542684 (N.D. Ga., Feb. 25, 2008). When subsequent discovery about discovery under Rule 26(b)(2)(B) revealed that email was destroyed after the EEOC charge was filed, the plaintiffs moved for sanctions. Sanctions were denied using the routine operations analysis of Rule 37(f), although the rule itself was never mentioned.
In this case four female plaintiffs alleged harassment “based on offensive comments and images on coworkers’ computer screens, as opposed to e-mails the Plaintiffs themselves received.” Id. at *3. Plaintiffs’ counsel served a request for production on the defendant, C.H. Robinson Worldwide, Inc., a transportation company with over 5,300 employees, for “any documents relating to or evidencing the presence of pornography . . . , including but not limited to email. . . .” The request initially had no date range, but was later clarified to include an eight year time span from 1998 to 2006.
This was, in my view, a vague and poorly worded request to produce. As such, it not only violated a legion of case law, but also the fourth principle of the Sedona Principles (all 14 Sedona Principles are listed on the right hand side of this blog):
4. Discovery requests for electronically stored information should be as clear as possible, while responses and objections to discovery should disclose the scope and limits of the production.
The request here literally required the responding party to search through all of its electronic files (and paper records), not only to look for pornographic materials, but also for records that might somehow “relate to or evidence the presence of pornography.” First of all, even the Supreme Court cannot define pornography, nor tell what pornography is unless and until they actually see it (one person’s porn may be another’s art photo). Thus the request literally requires a search of all records in the company, including all email, to try and determine if any of these records meet this elusive criterion.
Magistrate Judge Gerrilyn G. Brill, who is very experienced in discovery issues, understood this flaw well, and held:
As an initial matter, the Court finds that Plaintiffs’ discovery requests were extremely broad. The original requests were not limited by time, sender, or recipient. Even as later revised by Plaintiffs, the requests would require a search of all e-mails of all employees in the Atlanta North and Atlanta South branches for a six-year period. Moreover, it is unclear how Defendant would determine whether e-mails were “relating to or evidencing the presence of pornography” or were “of a sexual or gender derogatory nature” without examining the content of each and every e-mail and without making judgments about what constitutes pornography.
Petcou v. C.H. Robinson Worldwide, Inc., supra at *2.
Plaintiffs’ counsel should have known that defendant would object to such a vague and over broad request. In my experience, such obviously objectionable requests are either made out of naive error, or in the hope that defense counsel will err and not make a proper, timely objection, or the court will err and not sustain the objection. Should that happen, and believe me it occasionally does, the requesting party (here plaintiffs’ counsel) then has the responding party over a barrel, as it would obviously cost any large company a small fortune to comply with such broad and vague requests.
In this case, defense counsel did their job. Defendant timely objected to the request, including I assume the standard litany that the request was vague, over broad, and designed more to harass with burdensome discovery, than to uncover relevant facts. The defendant also objected on the grounds that much of the ESI requested was not reasonably accessible under Rule 26(b)(2)(B). Still, the plaintiffs moved to compel, hoping they would have better luck with the judge. In fairness, plaintiffs were anyway forced to file a motion on another document category requested, where defendant objected to “computer-generated reports of attempts by its employees to access adult website.” Id. at *1.
Judge Brill granted plaintiffs’ motion on the easily accessible reports, but denied it as to the pornography related request. Id. She did, however, tell plaintiffs they could renew the motion, as to emails only, if later discovery uncovered new grounds to undercut defendant’s claim that such discovery was overly burdensome. Plaintiffs then initiated what is called “discovery about discovery.” They took depositions concerning defendant’s computer system and document retention policy to try and disprove defendant’s Rule 26 claim that the email sought was not reasonably accessible.
If you are ready for a humorous interlude at this point, see the video below of John Cleese providing his in depth analysis of Rule 26 (thanks to his sponsor, Iron Mountain, and their funny website: friendlyadvicemachine.com).
John Cleese on Rule 26
What Plaintiffs’ Rule 26(b)(2)(B) Discovery Uncovered
The depositions taken by plaintiffs did not really undercut defendant’s inaccessibility position under Rule 26, but did give plaintiffs grounds to argue spoliation and Rule 37(f) (renumbered as Rule 37(e)).
The depositions established that defendant has an Outlook storage system where email is initially kept on exchange servers. Employees may freely delete any of their own emails they choose, and deleted emails can only be easily recovered for eight days thereafter. (I assume that the deleted files folder is automatically emptied after a file resides there for eight days, or some equivalent system.) Further, after an employee leaves the company, all of their email is deleted from the servers in ten days.
It is, of course, not difficult to access email located live on an exchange server. But for reasons not fully explained in the opinion (presumptively most of the key players to this litigation had previously left the company), the court concluded it was unlikely any such emails still existed on the servers:
E-mails of Defendant’s current employees that have not been deleted by those employees may still be on Defendants’ server and may be retrieved without great expense. However, it is very unlikely that e-mails from the period relevant to this lawsuit, sent to or from individuals who were employed during that period, are still on Defendants’ servers.
The depositions also established the email exchange servers are backed-up onto tapes every three to five days as part of the company’s disaster recovery plan. All of the different servers are included together, and the various emails servers are not segregated onto different backup tapes. The opinion implies that these daily tapes are not recycled on a weekly or monthly basis, but are kept for at least two years. The depositions established that the cost to simply retrieve email from a single backup tape was from between $325 to $365. (This is a low per-tape retrieval cost, and was probably predicted restoring a large number of tapes, here approximately 229 tapes.) The testimony established that it would cost approximately $79,000 just to recover emails for one employee from two years worth of tape. Id. (The opinion’s reference here to two years is why I assume the defendants did not have tapes older than that. At an average of $345 per tape, the defendant thus had approximately 229 tapes to restore.)
Rule 26 Applied to Prohibit Discovery
Judge Brill held that the testimony as to the $79,000 restoration cost met the defendant’s burden of proof under Rule 26(b)(2)(B) that “deleted e-mails from the period relevant to this lawsuit are not reasonably accessible because of undue burden and cost.” The next step in analysis under the rule to see whether there is good cause to have the ESI produced anyway, in spite of the undue burden and cost. Here the burden of proof shifts to the requesting party to establish good cause under the dictates of Rule 26(b)(2)(C). As Judge Brill explains:
Rule 26(b)(2)(C) directs the Court to consider, inter alia, whether “the burden or expense of the proposed discovery outweighs its likely benefit, considering the needs of the case, the amount in controversy, the parties’ resources, the importance of the issues at stake in the action, and the importance of the discovery in resolving the issues.”
Magistrate Judge Brill then goes on to conclude that good cause has not been established, primarily because the over broad nature of the request would require defendants to examine every email in the system. The volume of email existing over the two-year span of the backup tapes is not stated, but, for a company such as this with 5,300 employees, there would typically be millions of emails. The cost to review this much email would be astronomical. Of course, plaintiffs could have tried to avoid all of this with a clear and very focused request to produce.
In addition, plaintiffs could have focused the search somewhat by requesting production of specific pornographic images with known hash values. See Krause v. Texas, 243 S.W. 3d 95 (1st DCA, Oct. 31, 2007). This is the way police searches are normally conducted on computers to locate pornography, especially child pornography. This kind of hash search can be automated for known files, and thus might have slightly lessened the burden on defendant to search for pornography, especially if the requests were limited to known pornographic images. But, a hash search technique would still require restoration from backup tapes, and then breakdown of the attached files from the exchange packaging files so that all attachments could then be searched as individual files. You cannot hash search backup tapes, nor all emails packed together in an exchange format. They can only be searched after the individual emails and attachments have been unpacked. The limits of hash search in this kind of situation are very well described in Craig Ball’s critique of the original version of this blog, where I overstated the powers of hash. My thanks to Craig Ball for catching and pointing out this mistake to me in such a complementary manner. FN 1.
FN 1. I am really appreciate Craig taking the time to point out my error. It is exactly the kind of thing I hoped would happen when I started this writing exercise some 18 months ago. No one is perfect, despite our best efforts (and I put about 12 hours into researching and writing the original version of this essay). We all make mistakes, including people with far more expertise than me. Online writing in a blog format like this facilitates immediate peer review. The only alternative is the academic law review process. This traditional peer review model takes months, if not years, to hash out the truth. This kind of delay is unacceptable in today’s world, especially in a subject like e-discovery which changes so rapidly. Here Craig pointed out my hash error in his own online publication in just two days! What a wonderful thing. This is exactly why I write my e-discovery books online, in 2,500 to 3,000 chunks at a time, every weekend. When my blogs are later reworked and converted into my next book, the writings will already have had the benefit of the review and insights of my peers in the discovery world, including Craig Ball’s recent review. So dear reader, should you note an error, or just disagree with one of my many opinions, please let me know about it. Leave a comment below or write about it in your own blog as Craig has done. This will allow my next book, which, by the way, is now more that half way completed, to be far better than any solo, non-peer reviewed effort. End Footnote.
In addition to the disproportionate cost of the over broad search requested, the Judge concluded that the email information sought was indirect and cumulative. The plaintiffs did not claim harassment because they were emailed sexually explicit materials, but because they saw these images on other’s computers. For that reason, proof of pornographic emails was not important enough to the issues in the case to justify the burden. In Judge Brill’s words:
Finally, although the e-mails would corroborate Plaintiffs’ testimony regarding the prevalence of pornography in the workplace, they would provide little, if any, relevant information that Plaintiffs themselves have not already provided.
Id. at *2.
Judge Brill employed the analysis of Rule 26(b)(2)(B) and concluded that “the burden or expense of the proposed discovery as a whole outweighs its likely benefit.” In fact, her analysis relied primarily upon the proportionality factors stated in Rule 26(b)(2)(C). As other commentators have noted, the good cause provisions of new Rule 26(b)(2)(B) do not really change discovery analysis. Thomas Y. Allman, The “Two-Tiered” Approach to E-Discovery: Has Rule 26(b)(2)(B) Fullfilled Its Promise? 14 Richmond J.L. & Tech. 7, at pgs. 31-33. But still, the two tiered analysis does focus the parties, and the court, on the unique properties of ESI that impact traditional burden and cost analysis, and the problems of finding and retrieving certain types of ESI. Id.
This case shows how this kind of two-tiered accessibility road map can assist in attaining a just result. Judge Brill considered the relative inaccessibility of the email on the backup tapes and denied the motion to compel as to those emails. But the discovery on accessibility had shown that one witness still had active data on the exchange servers. The email of this one witness was easy to access. For that reason, the motion to compel was granted for that one witness only and defendant was ordered to produce:
… undeleted e-mails (if any), sent prior to 2007, by a current employee who has been specifically named by Plaintiffs as having had sexually explicit material on his computer;
Thus, this case provides an example of how the focused good cause analysis of 26(b)(2)(B) can lead the parties and the court to a just result. Under a general Rule 26(b)(2)(C) proportionality analysis this live data exception for one witness might have been missed, and all email discovery might have been prohibited.
The Rule 26 discovery uncovered the fact that defendant did not initiate a litigation hold at the time the original EEOC complaint was filed in 2001. Instead, defendant waited to place a hold on its normal document destruction policies until after suit was filed in 2007. Thus between 2001 and 2007 defendant continued to allow current employees to delete any emails they wished, even emails that could have been relevant to the EEOC claims. The defendants also continued its usual policy of deleting all emails from employees who left the company during that time.
Plaintiffs now argue that defendant should have suspended its usual practice when the EEOC complaint was filed, and the destruction of emails thereafter constitutes spoliation for which sanctions should be entered. This is in accord with the holding of Judge Scheindlin in Zubulake IV that the duty to preserve was triggered when an EEOC charge was filed, and perhaps even before that when Zubulake’s supervisors were convinced she would sue. Zubulake v. UBS Warburg LLC, 220 F.R.D. 212, 218 (S.D.N.Y. 2003).
The problem with plaintiffs’ argument is that they never asked defendant to save emails, and almost six years passed before they filed suit. Instead, they were silent for years before filing the motion for sanctions in 2007. This fact, coupled with the undisputed evidence that all of the email destruction was made in accordance with normal and usual procedures, caused Judge Brill to deny their motion:
In this case, Defendant deleted its employee’s emails in accordance with its normal retention and destruction schedule even after an EEOC complaint alleging company-wide sexual harassment had been filed in June of 2001. However, the plaintiffs in that case did not request company-wide preservation of e-mails, nor did they provide Defendant with the names of individuals in Atlanta whose e-mails should be preserved. It does not appear that Defendant acted in bad faith in following its established policy for retention and destruction of emails.
Although the opinion does not cite new Rule 37(e), it follows the analysis of the rule, and provides a good example of its operation. Here is the language of the new rule:
New Rule 37: Failure to Make Disclosures or Cooperate in Discovery; Sanctions.
(f)* Electronically stored information. Absent exceptional circumstances, a court may not impose sanctions under these rules on a party for failing to provide electronically stored information lost as a result of the routine, good faith operation of an electronic information system.
*Note, after 2008 rules renumbering 37(f) has now become 37(e).
The court found that the emails were destroyed according to the company’s normal retention and destruction policies. In other words, the information was lost through the routine operation of an electronic information system. Judge Brill found the destruction was in good faith because there was no evidence to the contrary, and because plaintiffs never sent a preservation demand letter.
Concluding Thoughts on Rule 37 Safe Harbor
This case would have presented a closer question if the plaintiffs had sent a demand letter upon filing the EEOC complaint. Then there might be a question as to whether the failure to depart from routine operations was in good faith. For instance, is it good faith to refuse to stop destroying emails of departing employees after receiving a demand notice? If you do place a hold on a routine procedure, how long should you continue the hold? Is it reasonable to expect an employer to incur additional expenses for six years to keep old emails on the off chance a law suit might someday be filed?
These are difficult questions, the answer to which depends on a host of facts and circumstances. Certainly the mere sending of a demand letter in and of itself should not trigger a hold obligation. In most circumstances the letter should, however, trigger a careful legal evaluation of duties. Is the threat credible enough to warrant a hold notice, and if so, what computer files should be preserved, and what routines should be stayed? A threat might be just credible enough to barely trigger a hold. In those circumstances it might be reasonable to limit the extent of the hold and construe it narrowly.
It is difficult enough to know what information may be discoverable based on the wording of a complaint. It is nearly impossible to get it right before a suit has even been filed. You might expect one kind of law suit, and end up getting hit by another. This is by no means an exact science. Unless there is a very specific and targeted preservation demand letter, the whole exercise of predicting which ESI might become discoverable is inherently error prone.
We all need as much help in this difficult guessing game as possible. Fortunately, the Sedona Conference provides us with guidance in its publication The Sedona Conference Commentary on Legal Holds: The Trigger & The Process. It lists multiple factors to consider in order to make a reasoned credible threat analysis. If a company considers these factors in good faith, and makes a reasoned credible threat assessment, then it should receive the protection of Rule 37 from sanctions when its projections later prove incorrect. For more information and discussion about the Sedona Guideline on holds see my previous blog The Sedona Conference Releases Two New Must-Read Commentaries on “Email Management” and “Legal Holds”.
It is important to remember that a Rule 37 safe harbor analysis is only needed in circumstances where the threat evaluation was wrong, or the guesstimate on scope of hold was wrong. The company did not think they would sue, but in fact they did. Or they thought they would sue, but guessed wrong as to what computer evidence would be discoverable. As long as the guess was reasonable and made in good faith, the protection of the rule should remain.
Predicting the future is always hazardous and no one is right all of the time. This is especially true in the emotional world of litigation, where it is common for tempers to flare and demand letters to fly, only to subside later with time. No one ever knows for sure whether another will later file suit or not. Further, the law is extremely complex and legal theories change constantly. Even after a complaint is filed, it is very common for amended complaints to be filed, and new and different theories to arise.
This means that in order for Rule 37 to work, there has to be plenty of room for error, so long as a reasoned, credible process is followed. When it comes to credible threat assessments and the application of Rule 37, courts will have to focus on reasonability of process, not accuracy of crystal ball. Companies threatened with litigation (and these days, who isn’t?) can help courts to do that, and thus help themselves stay within the protection of the rule, by documenting any assessments they make that a threat is not credible. On occasion, when large sums of money or important issues might be involved, a third party review of such decisions may be appropriate. This way they can confirm the rationality of the assessment, and make sure it is not a delusional exercise in wishful thinking.
Where a threat of litigation is accompanied with a specific and focused preservation demand, the rationale for an incredible evaluation should be especially strong. Further, if there is a specific preservation demand, and the decision is made to implement a hold, but to preserve less that is demanded, extraordinary care is again advised. In all close questions such as this you should communicate with opposing counsel and attempt to reach a compromise. If they do not respond, or respond but are unwilling to reach a reasonable compromise, then at least document your efforts and the parties’ rationales. That will go a long way in providing protection. If all else fails, and the potential dollar exposure justifies the expense, a preemptive suit for a declaratory judgment may even be appropriate. You will be breaking new ground to be sure, but as the landscape of litigation changes, new strategies such as this should be considered to cope with the new challenges these changes present.