On May 13, 2003, I ordered defendants UBS Warburg LLC, UBS Warburg, and UBS AG (collectively “UBS”) to restore and produce certain e-mails from a small group of backup tapes. Having reviewed the results of this sample restoration, Laura Zubulake now moves for an order compelling UBS to produce all remaining backup e-mails at its expense. UBS argues that based on the sampling, the costs should be shifted to Zubulake. For the reasons fully explained below, Zubulake must share in the costs of restoration, although UBS must bear the bulk of that expense. In addition, UBS must pay for any costs incurred in reviewing the restored documents for privilege.

 

I.                   BACKGROUND

 

The background of this lawsuit and the instant discovery dispute are recounted in two prior opinions, familiarity with which is presumed.1 In brief, Zubulake, an equities trader who earned approximately $650,000 a year with UBS,2 is now suing UBS for gender discrimination, failure to promote, and retaliation under federal, state, and city law. To support her claim, Zubulake seeks evidence stored on UBS’s backup tapes that is only accessible through costly and time-consuming data retrieval. In particular, Zubulake seeks e-mails relating to her that were sent to or from five UBS employees: Matthew Chapin (Zubulake’s immediate supervisor and the alleged primary discriminator), Jeremy Hardisty (Chapin’s supervisor and the individual to whom Zubulake originally complained about Chapin), Rose Tong (a human relations representative who was assigned to handle issues concerning Zubulake), Vinay Datta (a co-worker), and Andrew Clarke (another co-worker). The question presented in this dispute is which party should pay for the costs incurred in restoring and producing these backup tapes.

 

In order to obtain a factual basis to support the costshifting analysis, I ordered UBS to restore and produce e-mails from five of the ninety-four backup tapes that UBS had then identified as containing responsive documents; Zubulake was permitted to select the five tapes to be restored.3 UBS now reports, however, that there are only seventy-seven backup tapes that contain responsive data, including the five already restored.4 I further ordered UBS to “prepare an affidavit detailing the results of its search, as well as the time and money spent.”5 UBS has complied by submitting counsel’s declaration.

 

According to the declaration, Zubulake selected the backup tapes corresponding to Matthew Chapin’s e-mails from May, June, July, August, and September 2001.7 That period includes the time from Zubulake’s initial EEOC charge of discrimination (August 2001) until just before her termination (in the first week of October 2001).8 UBS hired an outside vendor, Pinkerton Consulting & Investigations, to perform the restoration.

 

Pinkerton was able to restore each of the backup tapes, yielding a total of 8,344 e-mails.10 That number is somewhat inflated, however, because it does not account for duplicates. Because each month’s backup tape was a snapshot of Chapin’s server for that month — and not an incremental backup reflecting only new material — an e-mail that was on the server for more than one month would appear on more than one backup tape. For example, an e-mail received in January 2001 and deleted in November 2001 would have been restored from all five backup tapes. With duplicates eliminated, the total number of unique emails restored was 6,203.

 

Pinkerton then performed a search for e-mails containing (in either the e-mail’s text or its header information, such as the “subject” line) the terms “Laura”, “Zubulake”, or “LZ”. The searches yielded 1,541 e-mails, or 1,075 if duplicates are eliminated. Of these 1,541 e-mails, UBS deemed approximately 600 to be responsive to Zubulake’s document request and they were produced. UBS also produced, under the terms of the May  Order, fewer than twenty e-mails extracted from UBS’s optical disk storage system.

 

Pinkerton billed UBS 31.5 hours for its restoration services at an hourly rate of $245, six hours for the development, refinement and execution of a search script at $245 an hour, and 101.5 hours of “CPU Bench Utilization” time for use of Pinkerton’s computer systems at a rate of $18.50 per hour.18 Pinkerton also included a five percent “administrative overhead fee” of $459.38.19 Thus, the total cost of restoration and search was $11,524.63.20 In addition, UBS incurred the following costs: $4,633 in attorney time for the document review (11.3 hours at $410 per hour)21 and $2,845.80 in paralegal time for tasks related to document production (16.74 hours at $170 per hour).22 UBS also paid $432.60 in photocopying costs, which, of course, will be paid by Zubulake and is not part of this cost shifting analysis.24 The total cost of restoration and production from the five backup tapes was $19,003.43.25

 

UBS now asks that the cost of any further production estimated to be $273,649.39, based on the cost incurred in restoring five tapes and producing responsive documents from those tapes — be shifted to Zubulake. The total figure includes $165,954.67 to restore and search the tapes and $107,694.72 in attorney and paralegal review costs. These costs will be addressed separately below.

 

II. LEGAL STANDARD

The Federal Rules of Civil Procedure specify that “any matter, not privileged, that is relevant to the claim or defense of any party” is discoverable, except where:

 

(i) the discovery sought is unreasonably cumulative or duplicative, or is obtainable from some other source that is more convenient, less burdensome, or less expensive; (ii) the party seeking discovery has had ample opportunity by discovery in the action to obtain the information sought; or (iii) the burden or expense of the proposed discovery outweighs its likely benefit, taking into account the needs of the case, the amount in controversy, the parties’ resources, the importance of the issues at stake in the litigation, and the importance of the proposed discovery in resolving the issues.

 

Although “the presumption is that the responding party must bear the expense of complying with discovery requests,” requests that run afoul of the Rule 26(b)(2) proportionality test may subject the requesting party to protective orders under Rule 26(c), “including orders conditioning discovery on the requesting party’s payment of the costs of discovery.”28 A court will order such a cost-shifting protective order only upon motion of the responding party to a discovery request, and “for good cause shown.”29 Thus, the responding party has the burden of proof on a motion for cost-shifting.

 

III. DISCUSSION

 

A. Cost-Shifting Generally

In Zubulake I, I considered plaintiff’s request for information contained only on backup tapes and determined that cost-shifting might be appropriate.31 It is worth emphasizing again that cost-shifting is potentially appropriate only when inaccessible data is sought. When a discovery request seeks accessible data — for example, active on-line or near-line data — it is typically inappropriate to consider cost-shifting.

 

In order to determine whether cost-shifting is appropriate for the discovery of inaccessible data, “the requested or produce it in the form requested, the responding party must state an objection complying with these rules. If the court orders the responding party to comply with the request, the court must also order that the requesting party pay the reasonable expenses of any extraordinary steps required to retrieve and produce the information.”); see also American Bar Association Civil Discovery Standards (1998) (Standard 29: “Unless the requesting party can demonstrate a substantial need for it, a party does not ordinarily have a duty to take steps to try to restore electronic information that has been deleted or discarded in the regular course of business but may not have been completely erased from computer memory. . . . The discovering party generally should bear any special expenses incurred by the responding party in producing requested electronic information.

 

The responding party should generally not have to incur undue burden or expense in producing electronic information, including the cost of acquiring or creating software needed to retrieve responsive electronic information for production to the other side.”).

 

 

following factors should be considered, weighted more-or-less in the following order”:

1.      The extent to which the request is specifically tailored to discover relevant

information;

2.      The availability of such information from other sources;

3. The total cost of production, compared to the amount in controversy;

4. The total cost of production, compared to the resources available to each party;

5. The relative ability of each party to control costs and its incentive to do so;

6. The importance of the issues at stake in the litigation; and

7. The relative benefits to the parties of obtaining the information.

 

In establishing this test, I modified the list of factors articulated in Rowe Entertainment, Inc. v. William Morris Agency, Inc.,33 to meet the legitimate concern of those commentators who have argued that “the factors articulated in Rowe [] tend to favor the responding party, and frequently result in shifting the costs of electronic discovery to the requesting party.” Thus, the seven-factor testarticulated in Zubulake I was designed to simplify application of the Rule 26(b)(2) proportionality test in the context of electronic data and to reinforce the traditional presumptive allocation of costs.

 

B. Application of the Seven Factor Test

 

1. Factors One and Two

As I explained in Zubulake I, the first two factors together comprise the “marginal utility test” announced in McPeek v.

 

Ashcroft:

 

The more likely it is that the backup tape contains information that is relevant to a claim or defense, the fairer it is that the [responding party] search

at its own expense. The less likely it is, the more unjust it would be to make the [responding party] search at its own expense. The difference is “at the margin.”

 

These two factors should be weighted the most heavily in the cost-shifting analysis. total costs and ability to control costs will favor the responding party. If courts simply conduct an absolute comparison of the eight Rowe factors, the responding party will need to attain just one more factor to shift the costs to the requesting party. This is a dramatic shift from earlier cases, which were more inclined to follow the presumption in traditional document production, requiring the responding party to pay.”).

 

a. The Extent to Which the Request Is

Specifically Tailored to Discover Relevant

Information

 

The document request at issue asks for “[a]ll documents concerning any communication by or between UBS employees concerning Plaintiff,”37 and was subsequently narrowed to pertain to only five employees (Chapin, Hardisty, Tong, Datta, and Clarke) and to the period from August 1999 to December 2001.38 This is a relatively limited and targeted request, a fact borne out by the e-mails UBS actually produced, both initially and as a result of the sample restoration.

 

At oral argument, Zubulake presented the court with sixty-eight e-mails (of the 600 she received) that she claims are “highly relevant to the issues in this case” and thus require, in her view, that UBS bear the cost of production.39 And indeed, a review of these e-mails reveals that they are relevant. Taken together, they tell a compelling story of the dysfunctional atmosphere surrounding UBS’s U.S. Asian Equities Sales Desk (the “Desk”). Presumably, these sixty-eight e-mails are reasonably representative of the seventy-seven backup tapes.

 

A number of the e-mails complain of Zubulake’s behavior. Zubulake was described by Clarke as engaging in “bitch sessions about the horrible men on the [Desk],” and as a “conduit for a steady stream of distortions, accusations and good ole fashioned back stabbing,”40 and Hardisty noted that Zubulake was disrespectful to Chapin and other members of the Desk.41 And Chapin takes frequent snipes at Zubulake.42 There are also complaints about Chapin’s behavior.43 In addition, Zubulake argues that several of the e-mails contradict testimony given by UBS employees in sworn depositions.

 

In particular, six e-mails singled out by Zubulake as particularly “striking”45 include:

 

·         An e-mail from Hardisty, Chapin’s supervisor, chastising Chapin for saying one thing and doing another with respect to Zubulake. Hardisty said, “As I see it, you do not appear to be upholding your end of the bargain to work with her.” This e-mail stands in contrast to UBS’s response to Zubulake’s EEOC charges, which says that “Mr. Chapin was receptive to Mr. Hardisty’s suggestions [for improving his relationship with Zubulake].”

 

·         An e-mail from Chapin to one of his employees on the  Desk, Joy Kim, suggesting to her how to phrase a complaint against Zubulake. A few hours later, Joy Kim did in fact send an e-mail to Chapin complaining about Zubulake, using precisely the same words that Chapin had suggested. But at his deposition (taken before these e-mails were restored), Chapin claimed that he did not solicit the complaint.

 

·         An e-mail from Chapin to the human resources employee handling Zubulake’s case listing the employees on the Desk and categorizing them as senior, mid-level, or junior salespeople. In its EEOC filing, however, UBS claimed in response to Zubulake’s argument that she was the only senior salesperson on the desk, that it “does not categorize salespeople as ‘junior’ or ‘senior.’” In addition, UBS claimed in its EEOC papers that there were four female salespeople on the Desk, but this email shows only two.

 

·         An e-mail from Chapin to Hardisty acknowledging that Zubulake’s “ability to do a good job . . . is clear,” and that she is “quite capable.”

 

·         An e-mail from Derek Hillan, presumably a UBS employee, to Chapin and Zubulake using vulgar language, although UBS claims that it does not tolerate such language.

 

·         An e-mail from Michael Oertli, presumably a UBS employee, to Chapin explaining that UBS’s poor performance in Singapore was attributable to the fact that it only “covered” eight or nine of twenty-two accounts, and not to Zubulake’s poor performance, as UBS has argued

 

Not surprisingly, UBS argued that these e-mails have very little, if any, relevance to the issues in the case.

 

While all of these e-mails are likely to have some “tendency to make the existence of any fact that is of consequence to the determination of the action more probable or less probable than it would be without the evidence,” none of them provide any direct evidence of discrimination. To be sure, the e-mails reveal a hostile relationship between Chapin and Zubulake — UBS does not contest this. But nowhere (in the sixty-eight e-mails produced to the Court) is there evidence that Chapin’s dislike of Zubulake related to her gender.

 

b. The Availability of Such Information from Other Sources

 

The other half of the marginal utility test is the availability of the relevant data from other sources. Neither party seemed to know how many of the 600 e-mails produced in response to the May 13 Order had been previously produced. UBS argues that “nearly all of the restored e-mails that relate to plaintiff’s allegations in this matter or to the merits of her case were already produced.”54 This statement is perhaps too careful, because UBS goes on to observe that “the vast majority of the restored e-mails that were produced do not relate at all to plaintiff’s allegations in this matter or to the merits of her case.”55 But this determination is not for UBS to make; as the saying goes, “one man’s trash is another man’s treasure.”

 

It is axiomatic that a requesting party may obtain “any matter, not privileged, that is relevant to the claim or defense of any party.” The simple fact is that UBS previously produced only 100 pages of e-mails, but has now produced 853 pages (comprising the 600 responsive e-mails) from the five selected backup tapes alone.58 UBS itself decided that it was obliged to provide these 853 pages of e-mail pursuant to the requirements of Rule 26. Having done so, these numbers lead to the unavoidable conclusion that there are a significant number of responsive emails

that now exist only on backup tapes.

 

If this were not enough, there is some evidence that Chapin was concealing and deleting especially relevant e-mails. When Zubulake first filed her EEOC charge in August 2001, all UBS employees were instructed to save documents relevant to her case.59 In furtherance of this policy, Chapin maintained separate files on Zubulake.60 However, certain e-mails sent after the initial EEOC charge — and particularly relevant to Zubulake’s retaliation claim — were apparently not saved at all. For example, the e-mail from Chapin to Joy Kim instructing her on how to file a complaint against Zubulake61 was not saved, and it bears the subject line “UBS client attorney privilege [sic] only,” although no attorney is copied on the e-mail.62 This potentially useful e-mail was deleted and resided only on UBS’s backup tapes.

 

 

In sum, hundreds of the e-mails produced from the five backup tapes were not previously produced, and so were only available from the tapes. The contents of these e-mails are also new. Although some of the substance is available from other sources (e.g., evidence of the sour relationship between Chapin and Zubulake), a good deal of it is only found on the backup tapes (e.g., inconsistencies with UBS’s EEOC filing and Chapin’s deposition testimony). Moreover, an e-mail contains the precise words used by the author. Because of that, it is a particularly powerful form of proof at trial when offered as an admission of a party opponent.

 

c. Weighing Factors One and Two

 

The sample restoration, which resulted in the production of relevant e-mail, has demonstrated that Zubulake’s discovery request was narrowly tailored to discover relevant information. And while the subject matter of some of those emails was addressed in other documents, these particular e-mails are only available from the backup tapes. Thus, direct evidence of discrimination may only be available through restoration. As a result, the marginal utility of this additional discovery may be quite high.

 

While restoration may be the only means for obtaining direct evidence of discrimination, the existence of that evidence is still speculative. The best that can be said is that Zubulake has demonstrated that the marginal utility is potentially high. All-in-all, because UBS bears the burden of proving that cost shifting is warranted, the marginal utility test tips slightly against cost-shifting.

 

2. Factors Three, Four and Five

“The second group of factors addresses cost issues:

‘How expensive will this production be?’ and, ‘Who can handle

that expense?’”64

 

a. The Total Cost of Production Compared to the

Amount in Controversy

 

UBS spent $11,524.63, or $2,304.93 per tape, to restore the five back-up tapes. Thus, the total cost of restoring the remaining seventy-two tapes extrapolates to $165,954.67.65 In order to assess the amount in controversy, I posed the following question to the parties: Assuming that a jury returns a verdict in favor of plaintiff, what economic damages can the plaintiff reasonably expect to recover? Plaintiff

answered that reasonable damages are between $15,271,361 and $19,227,361, depending upon how front pay is calculated.66 UBS answered that damages could be as high as $1,265,000.67

 

Obviously, this is a significant disparity. At this early stage, I cannot assess the accuracy of either estimate. Plaintiff had every incentive to high-ball the figure and UBS had every incentive to low-ball it. It is clear, however, that this case has the potential for a multi-million dollar recovery. Whatever else might be said, this is not a nuisance value case, a small case or a frivolous case. Most people do not earn $650,000 a year. If Zubulake prevails, her damages award undoubtedly will be higher than that of the vast majority of Title VII plaintiffs.

 

In an ordinary case, a responding party should not be required to pay for the restoration of inaccessible data if the cost of that restoration is significantly disproportionate to the value of the case. Assuming this to be a multi-million dollar case, the cost of restoration is surely not “significantly disproportionate” to the projected value of this case. This factor weighs against cost-shifting.

 

b. The Total Cost of Production Compared to the Resources Available to Each Party

 

There is no question that UBS has exponentially more resources available to it than Zubulake.68 While Zubulake is an accomplished equities trader,69 she has now been unemployed for close to two years. Given the difficulties in the equities market and the fact that she is suing her former employer, she  may not be particularly marketable. On the other hand, she  asserts that she has a $19 million claim against UBS. So while  UBS’s resources clearly dwarf Zubulake’s, she may have the  financial wherewithal to cover at least some of the cost of  restoration. In addition, it is not unheard of for plaintiff’s  firms to front huge expenses when multi-million dollar recoveries are in sight.70 Thus, while this factor weighs against cost shifting, it does not rule it out.

 

c. The Relative Ability of Each Party to Control

Costs and Its Incentive to Do So

 

Restoration of backup tapes must generally be done by an outside vendor.71 Here, UBS had complete control over the selection of the vendor. It is entirely possible that a less expensive vendor could have been found.72 However, once that vendor is selected, costs are not within the control of either party. In addition, because these backup tapes are relatively well-organized — meaning that UBS knows what e-mails can be found on each tape — there is nothing more that Zubulake can do to focus her discovery request or reduce its cost. Zubulake has already made a targeted discovery request and the restoration of the sample tapes has not enabled her to cut back on that request. Thus, this factor is neutral.

 

3. Factor Six: The Importance of the Issues at Stake in the Litigation

 

As noted in Zubulake I, this factor “will only rarely come into play.”74 Although this case revolves around a weighty issue — discrimination in the workplace — it is hardly unique. Claims of discrimination are common, and while discrimination is an important problem, this litigation does not present a particularly novel issue. If I were to consider the issues in this discrimination case sufficiently important to weigh in the cost-shifting analysis, then this factor would be virtually meaningless. Accordingly, this factor is neutral.

 

4. Factor Seven: The Relative Benefits to the Parties of Obtaining the Information

 

Although Zubulake argues that there are potential  benefits to UBS in undertaking the restoration of these backup  tapes in particular, the opportunity to obtain evidence that  may be useful at summary judgment or trial there can be no  question that Zubulake stands to gain far more than does UBS, as  will typically be the case.75 Certainly, absent an order, UBS would not restore any of this data of its own volition. Accordingly, this factor weighs in favor of cost-shifting.

 

5. Summary and Conclusion

 

Factors one through four tip against cost-shifting (although factor two only slightly so). Factors five and six are neutral, and factor seven favors cost-shifting. As noted in my earlier opinion in this case, however, a list of factors is not merely a matter of counting and adding; it is only a guide. Because some of the factors cut against cost shifting, but only slightly so — in particular, the possibility that the continued production will produce valuable new information — some cost shifting is appropriate in this case, although UBS should pay the majority of the costs. There is plainly relevant evidence that is only available on UBS’s backup tapes. At the same time, Zubulake has not been able to show that there is indispensable evidence on those backup tapes (although the fact that Chapin apparently deleted certain e-mails indicates that such evidence may exist).

 

The next question is how much of the cost should be shifted. It is beyond cavil that the precise allocation is a matter of judgment and fairness rather than a mathematical consequence of the seven factors discussed above. Nonetheless, the analysis of  those factors does inform the exercise of discretion. Because the seven factor test requires that UBS pay the lion’s share, the percentage assigned to Zubulake must be less than fifty percent. A share that is too costly may chill the rights of litigants to pursue meritorious claims.77 However, because the success of this search is somewhat speculative, any cost that fairly can be assigned to Zubulake is appropriate and ensures that UBS’s expenses will not be unduly burdensome. A twenty-five percent assignment to Zubulake meets these goals.

 

C. Other Costs

 

The final question is whether this result should apply to the entire cost of the production, or only to the cost of restoring the backup tapes. The difference is not academic — the estimated cost of restoring and searching the remaining backup tapes is $165,954.67, while the estimated cost of producing them (restoration and searching costs plus attorney and paralegal costs) is $273,649.39 ($19,003.43 for the five sample tapes, or $3,800.69 per tape, times seventy-two unrestored tapes), a difference of $107,694.72. As a general rule, where cost-shifting is  appropriate, only the costs of restoration and searching should be shifted. Restoration, of course, is the act of making inaccessible material accessible.

 

That “special purpose” or “extraordinary step” should be the subject of cost-shifting. Search costs should also be shifted because they are so intertwined with the restoration process; a vendor like Pinkerton will not only develop and refine the search script, but also necessarily execute the search as it conducts the restoration. However, the responding party should always bear the cost of reviewing and producing electronic data once it has been converted to an accessible form. This is so for two reasons.

 

First, the producing party has the exclusive ability to control the cost of reviewing the documents. In this case, UBS decided — as is its right — to have a senior associate at a top New York City law firm conduct the privilege review at a cost of $410 per hour. But the job could just as easily have been done (while perhaps not as well) by a first-year associate or contract attorney at a far lower rate. UBS could similarly have obtained paralegal assistance for far less than $170 per hour.

 

Moreover, the producing party unilaterally decides on the review protocol. When reviewing electronic data, that review may range from reading every word of every document to conducting a series of targeted key word searches. Indeed, many parties to document-intensive litigation enter into so-called “claw-back” agreements that allow the parties to forego privilege review  altogether in favor of an agreement to return inadvertently produced privileged documents. The parties here can still

reach such an agreement with respect to the remaining seventy-two tapes and thereby avoid any cost of reviewing these tapes for privilege.

 

 

 

Second, the argument that all costs related to the production of restored data should be shifted misapprehends the nature of the cost-shifting inquiry. Recalling that costshifting is only appropriate for inaccessible but otherwise discoverable — data, it necessarily follows that once the data has been restored to an accessible format and responsive documents located, cost-shifting is no longer appropriate. Had it always been accessible, there is no question that UBS would have had to produce the data at its own cost. Indeed, this is precisely what I ordered in Zubulake I with respect to certain emails kept on UBS’s optical disk system.

 

Documents stored on backup tapes can be likened to paper records locked inside a sophisticated safe to which no one has the key or combination. The cost of accessing those documents may be onerous, and in some cases the parties should split the cost of breaking into the safe. But once the safe is opened, the production of the documents found inside is the sole responsibility of the responding party. The point is simple: technology may increasingly permit litigants to reconstruct lost or inaccessible information,84 but once restored to an accessible form, the usual rules of discovery apply.

IV. CONCLUSION

 

For the reasons set forth above, the costs of restoring any backup tapes are allocated between UBS and Zubulake seventy five percent and twenty-five percent, respectively. All other costs are to be borne exclusively by UBS. Notwithstanding this ruling, UBS can potentially impose a shift of all of its costs, attorney’s fees included, by making an offer to the plaintiff under Rule.

 

– Appearances –

For Plaintiff:

James A. Batson, Esq.

Christina J. Kang, Esq.

Liddle & Robinson, LLP

685 Third Avenue

New York, New York 10017

(212) 687-8500

 

For Defendants:

Kevin B. Leblang, Esq.

Norman C. Simon, Esq.

Kramer Levin Naftalis & Frankel LLP

919 Third Avenue

New York, New York 10022

(212) 715-9100