These appeals require us to apply the principles and case law pertaining to Fed.R.Civ.P. 26(c) to a claim that certain materials obtained in civil discovery but alleged by the producing party to be confidential may be disclosed by the discovering party to the public. We must also consider whether we have appellate jurisdiction over the district court's interlocutory order permitting disclosure of the materials.

The appeal arises from two of the several cases nationwide in which cigarette smokers or their personal representatives have instituted product liability suits against tobacco companies. In both cases, the parties had already engaged in extensive discovery, including production of a very large number of documents by defendants, when the defendants sought protective orders that would prevent the dissemination, either to the public or to counsel in other similar cases, of any documents they had produced or would produce during discovery.1 A federal magistrate entered identical protective orders in both cases along the lines requested by the defendants.

On appeal from the magistrate's orders, the district court substantially revised them. The court altered the procedure that the magistrate's orders had established for deciding disputed claims of confidentiality, and restricted the orders' scope so that release of the documents to the press and public would have followed almost as of course but for this appeal. The revised orders also permitted the documents to be used in other cases in which plaintiffs' counsel was the counsel of record.

The defendants thereupon appealed to this Court and petitioned for mandamus, asserting that the revised orders violated Fed.R.Civ.P. 26(c) and reflected a skewed reading of Seattle Times Co. v. Rhinehart, 467 U.S. 20, 104 S. Ct. 2199, 81 L. Ed. 2d 17 (1984). The defendants also moved for an expedited appeal and a stay of the district court's orders, as well as reinstatement of the magistrate's orders pending appeal. We granted those motions. The plaintiffs moved to dismiss the appeals for want of appellate jurisdiction, and also moved to dismiss the petition for mandamus.

We hold that: (1) we do not have jurisdiction to review the order pursuant to the collateral order doctrine as enunciated in Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 69 S. Ct. 1221, 93 L. Ed. 1528 (1949); (2) we do have mandamus jurisdiction to review the order pursuant to 28 U.S.C. Sec. 1651 (1982); (3) because the district court's reading of Seattle Times constituted a clear error of law, the ruling on the defendants' motion for protective orders was incorrect; and (4) the district court also clearly erred in relying on Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 104 S. Ct. 1949, 80 L. Ed. 2d 502 (1984) to exercise plenary review of the magistrate's protective order, for the court was bound to apply a "clearly erroneous" standard. We therefore grant the writ of mandamus. To assist the district court in future proceedings, we discuss two additional points relevant to this case: the definition of "good cause," and the administration of protective order proceedings.

Rose Cipollone and her husband Antonio filed a complaint against Liggett Group, Inc., Phillip Morris, Inc., and Loew's Theaters, Inc., all manufacturers of cigarettes,2 in the district court for the District of New Jersey on August 1, 1983. Jurisdiction was based on diversity of citizenship. 28 U.S.C. Sec. 1332 (1982). The complaint alleged that defendants manufactured or sold cigarettes and that Rose Cipollone had smoked defendants' cigarettes for almost forty years. As a result of her smoking, the complaint alleged, she acquired bronchogenic carcinoma and other personal injuries; it further alleged that she had experienced severe pain and suffering and that her illness had caused her--and would continue to cause her--great expense. Plaintiffs sought compensation for Rose Cipollone's injuries, suing under theories of negligence and strict liability. Central to plaintiffs' case was their allegation that defendants had withheld scientific evidence from the public and had misrepresented the effects upon health of smoking cigarettes. They also sought compensation for Antonio Cipollone's loss of consortium.

Shortly thereafter, Susan Haines as administratrix ad prosequendum and executrix of the Estate of Peter F. Rossi brought suit in the same court against the same three defendants as well as R.J. Reynolds Tobacco Co. and the Tobacco Institute, Inc. Haines was represented by the same attorney who represented the Cipollones. Jurisdiction was based on diversity, and once again the complaint alleged tortious conduct sounding in strict liability and negligence. The complaint also included an allegation of misrepresentation. The plaintiff sought compensation for the decedent's pain and suffering and for his death, which she alleged was the result of his smoking defendants' cigarettes.

The district court ordered discovery in both cases under the supervision of a federal magistrate. 28 U.S.C. Sec. 636(b)(1)(A) (1982). Discovery proceeded until March 1985, and a large number of documents were produced by the defendants for inspection pursuant to Fed.R.Civ.P. 34. On that date, the defendants moved for an "umbrella" protective order. The defendants argued that such an order would facilitate the discovery process by reducing the number of occasions for lawyers' conferences and discussions about the confidentiality of particular documents. Defendants also argued that they had good cause for the protective order under Fed.R.Civ.P. 26(c)3 and that the closely analogous Seattle Times Co. v. Rhinehart, 467 U.S. 20, 104 S. Ct. 2199, 81 L. Ed. 2d 17 (1984), permitted a protective order in this case. Plaintiffs objected to the defendants' proposal, countering that the defendants' real purpose was to make it impossible for plaintiffs in other suits against the cigarette companies to share information gathered from the defendants. The defendants' strategy, said plaintiffs, was to raise the expense of litigation for future plaintiffs, thus making the cost of suits prohibitive.

After hearing the matter, the magistrate found for defendants. On March 25, 1985, he entered identical protective orders in both cases. The crucial aspects of the protective orders may be summarized as follows: (a) "all information" produced in discovery, presumably confidential and nonconfidential alike, could be used only for the instant cases and not for other cases or other purposes; (b) the defendants had the responsibility in the first instance of deciding in good faith which of their documents were confidential and marking them accordingly; (c) information marked confidential could be examined as a matter of course by plaintiffs' lawyer, his associates, and experts retained by plaintiffs or their lawyer for the cases; (d) if plaintiffs wished to disclose the information to anyone else, they had to inform defendants' counsel, who then had opportunity to apply to the court to prevent that disclosure; and (e) all documents and copies thereof had to be destroyed or returned at the conclusion of the litigation.4

Plaintiffs appealed the protective order to the district court, arguing that the order violated plaintiffs' first amendment rights to disseminate the information that they had received through discovery. Plaintiffs relied on Seattle Times, supra, arguing that the defendants and the magistrate had misconstrued the Supreme Court's holding in that case. They also argued that the defendants had failed to demonstrate good cause as required for a protective order by Fed.R.Civ.P. 26(c).

The district court filed a lengthy opinion, covering its scope of review of the magistrate's decision, the meaning and relevance of Seattle Times, the notion of "good cause" in Fed.R.Civ.P. 26(c), and the proper scope of the protective order. Disposition of the appeal requires that we describe each part of the district court's opinion in some detail.

1. The District Court's Scope of Review of the Magistrate's Protective Order

Although 28 U.S.C. Sec. 636(b)(1)(A) states that a magistrate's order is not to be reconsidered unless it is "clearly erroneous or contrary to law,"5 the district court ruled that its standard of review was plenary, relying on Bose Corp. v. Consumers Union of the United States, Inc., 466 U.S. 485, 104 S. Ct. 1949, 80 L. Ed. 2d 502 (1984), which held that an appellate court has plenary review over the finding of actual malice in libel cases. See Dist.Ct.Op. at A17-A18.6

2. The District Court's Analysis of Seattle Times

The district court next engaged in a lengthy first amendment analysis of protective orders in discovery. It reviewed the conflicting approaches of the circuit courts prior to Seattle Times7 and then observed that Seattle Times had resolved the issue. The court quoted what it believed to be the relevant analysis from that case:

The critical question that this case presents is whether a litigant's freedom comprehends the right to disseminate information that he has obtained pursuant to a court order that both granted him access to that information and placed restraints on the way in which the information might be used. In addressing that question it is necessary to consider whether the "practice in question [furthers] an important or substantial governmental interest" and whether "the limitation of First Amendment freedoms [is] no greater than is necessary to the protection of the particular governmental interest involved." Procunier v. Martinez, 416 U.S. 396, 413, 94 S. Ct. 1800, 1811, 40 L. Ed. 2d 224 (1974).

104 S.Ct. at 2207 (quoted in Dist.Ct.Op. at A21-A22).

The district court believed the passage established that, when a case involves matters of substantial public interest, a protective order implicates first amendment concerns and some constitutional analysis is required. Dist.Ct.Op. at A24-A25. The district court went further, explicitly analogizing the case before it to Seattle Times and holding that the same constitutional inquiry was appropriate in both cases: "It therefore remained there, and remains here, to decide only whether the protective orders at issue limited first amendment freedoms more than necessary or essential to protect the governmental interests furthered by Rule 26(c)." Dist.Ct.Op. at A22.

The court did note one point of confusion about Seattle Times that is relevant to our discussion below. Despite the Supreme Court's apparent endorsement in the above passage of a least restrictive means analysis, its holding subsumes a different analysis entirely. The district court quoted that holding in full:

We therefore hold that where, as in this case, a protective order is entered on a showing of good cause as required by Rule 26(c), is limited to the context of pretrial civil discovery, and does not restrict the dissemination of the information if gained from other sources, it does not offend the First Amendment.

104 S.Ct. at 2209-10 (footnote omitted) (quoted in Dist.Ct.Op. at A22-A23). As the district court noted, this explicit holding appears to exclude any first amendment analysis from the decision about whether a court should issue a protective order; that is, it implies that "if a protective order passes muster under Rule 26(c), it must, of necessity, be constitutional." Dist.Ct.Op. at A24. However, the analytical passage quoted earlier, see supra pp. 1113-14, implies that a court must apply a least restrictive alternative test to all proposed protective orders. Although it noted this apparent contradiction, the district court did not resolve it explicitly, apparently assuming that Seattle Times imposed a least restrictive alternative test and that the test had to be read into the holding. See Dist.Ct.Op. at 24-25; see generally infra part IV (discussing the district court's first amendment analysis).8

3. The District Court's Findings on Good Cause

The district court noted that the party seeking the protective order bore the burden of proving that there was good cause for such an order. It also observed that a protective order could issue only upon a showing that disclosure would result in "clearly defined and serious injury." Dist.Ct.Op. at A26. Although early in its opinion the court suggested that there could be good cause only for revelation of technical information that might hurt one of the defendants' competitive positions,9 the court later made it clear, as the caselaw has established, that lesser concerns, including "embarrassment," might constitute good cause for a protective order. Dist.Ct.Op. at A32 n. 8.10

After discussing these broad legal issues, the court turned to the particular facts before it, and found that neither the magistrate's opinion nor the submissions of the defendants sustained the burden of justifying the protective order. It found that "the reasons asserted are quite conclusory," id. at A28, and that defendants' suggestion that the magistrate's protective order would 'streamline the litigation' was not sufficient to carry the evidentiary burden. Id.

As noted above, the magistrate's order applied to all information produced during discovery. See supra p. 11; Magistrate's Order p 2, supra note 4. The district court criticized this approach, stating that nonconfidential material was, by definition, information for which no Rule 26(c) good cause had been shown and that therefore no protective order should protect such material. Dist.Ct.Op. at A29.

The district court also criticized the portion of the magistrate's order that had prohibited the use in any other case of the materials produced in this case's discovery. The district court said that the prohibition "undermine[d] the purpose of the Federal Rules of Civil Procedure 'to secure the just, speedy, and inexpensive determination of every action.' " Dist.Ct.Op. at A34 (quoting Fed.R.Civ.P. 1) (footnote omitted). Additionally, the court noted that prohibiting the use of materials from one case in other cases would burden both the plaintiffs and the defendant:

There may be some claimants who do not have the resources ... to pursue the thorough investigation which these cases require. To require that each and every plaintiff go through the identical, lone and expensive process would be ludicrous. Even from the point of view of the defendants (though they resist), it would seem that they would benefit by avoiding repetition of the same discovery in each and every case."

Id. at A11.

5. The District Court's Amendment to the Protective Order

The district court amended the magistrate's protective order in light of its conclusions as outlined above. The court's amendments were as follows: (a) whereas the magistrate's protective order had limited the use of all materials produced in discovery, the amended protective order would apply only to confidential materials and would not restrict the use of nonconfidential materials; (b) rather than making defendants' good faith the only limitation on their freedom to designate documents confidential, and forcing the plaintiffs to challenge the designation subject thereafter to rulings by the Court, the amended order required the defendants to demonstrate in a document-by-document showing to the court that each document they believed to be confidential was so in fact; the advantage of this system, the court explained, was that it "does not allow misuse of the confidentiality designation and places the burden of proving such confidentiality squarely upon defendants, as required by Rule 26(c) and the first amendment," id. at A29-A30; (c) although the court agreed that confidential information could not be released to the public, its order differed from the magistrate's in that the court's order allowed plaintiffs' counsel to use any and all confidential materials in cases in which he was a participant, id. at A32; and (d) the amended order eliminated entirely the provision requiring counsel to return or destroy all documents produced in discovery; this was done virtually without discussion, because defendants had not opposed plaintiffs' motion to eliminate the provision.11

The defendants immediately moved the district court for a stay of its own protective order. The district court granted a stay conditioned on defendants' instituting proceedings in the court of appeals, which they did promptly. We granted a further stay pending disposition of the appeal, having been informed that appellees had scheduled a press conference for the morning following expiration of the stay and that they would, at that time, release to the public all the documents obtained in discovery.

Discovery orders, being interlocutory, are not normally appealable. See Borden Co. v. Sylk, 410 F.2d 843, 845 (3d Cir.1969); 8 C. Wright & A. Miller, Federal Practice and Procedure: Civil Sec. 2006 at 29 (1970 & Supp.1985). The first issue before us, therefore, is whether we have appellate jurisdiction. The defendants make two arguments in favor of appellate jurisdiction. First, they assert that the district court's protective order is a collateral order appealable under the rule of Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541 (1949). In the alternative, they argue that this court should exercise its statutory power of mandamus, 28 U.S.C. Sec. 1651 (1982), to review the order. We consider collateral appealability here and the mandamus argument in part III infra.12

Title 28 U.S.C. Sec. 1291 (1982) provides that courts of appeals may review only "final" decisions of the district courts. In Cohen v. Beneficial Industrial Loan Corp., 337 U.S. 541, 69 S. Ct. 1221, 93 L. Ed. 1528 (1949), however, the Supreme Court established a narrow exception to the rule of finality. Cohen held that a prejudgment order of a district court can be reviewed if it falls within

that small class [of prejudgment orders] which finally determine claims of right separable from and collateral to, rights asserted in the action, too important to be denied review, and too independent of the cause itself to require that appellate consideration be deferred until the whole case is adjudicated.

Id. at 546, 69 S.Ct. at 1225. See also Mitchell v. Forsyth, --- U.S. ----, ----, 105 S. Ct. 2806, 2815, 86 L. Ed. 2d 411 (1985); Richardson-Merrell, Inc. v. Koller, --- U.S. ----, ----, 105 S. Ct. 2757, 2761, 86 L. Ed. 2d 340 (1985).

Cohen's progeny have established three requirements for the review of non-final orders: to be reviewed the order must " conclusively determine the disputed question, resolve an important issue completely separate from the merits of the action, and be effectively unreviewable on appeal from final judgment." Coopers & Lybrand v. Livesay, 437 U.S. 463, 468, 98 S. Ct. 2454, 2457, 57 L. Ed. 2d 351 (1978). We have made clear that each of the three requirements must be met before appellate review is permitted. Eavenson, Auchmuty & Greenwald v. Holtzman, 775 F.2d 535, 537 (3d Cir.1985); Metex Corp. v. ACS Industries, Inc., 748 F.2d 150, 153 (3d Cir.1984); Lusardi v. Xerox Corp., 747 F.2d 174, 177 (3d Cir.1984); Gross v. G.D. Searle & Co., 738 F.2d 600, 602 (3d Cir.1984). This approach furthers the important goal of avoiding piecemeal litigation.

The second prong is not met here because defendants' claim touches on the merits of the underlying action. The underlying action raises issues concerning whether and when the defendants knew of the health hazards associated with smoking cigarettes and what steps the defendants allegedly took to mislead the public about those hazards. Defendants contend that the materials should not be disseminated because they would present a distorted and unfair picture about what the defendants knew about the effects of cigarettes on health. Our evaluation of defendants' argument would take us into the merits of the underlying action because we would have to make a judgment about what defendants knew and what steps they may have taken to mislead the public--precisely the issues at the heart of the underlying action. See supra pp. 1111-12.13 Because the second prong is not satisfied, we do not have jurisdiction under the Cohen doctrine.

The All Writs Act, 28 U.S.C. Sec. 1651(a) (1982), provides that "[t]he Supreme Court and all courts established by Act of Congress may issue all writs necessary or appropriate in aid of their respective jurisdictions." Although writs of mandamus are extraordinary devices and we have read Sec. 1651(a) narrowly, mandamus has been held to be appropriate when a failure to issue the writ would lead to the disclosure of confidential materials. See, e.g., Bogosian v. Gulf Oil Corp., 738 F.2d 587 (3d Cir.1984); Iowa Beef Processors, Inc. v. Bagley, 601 F.2d 949 (8th Cir.1979); In re Halkin, 598 F.2d 176 (D.C.Cir.1979). In Sporck v. Peil, 759 F.2d 312, 314 (3d Cir.1985), we held that a writ of mandamus should issue when (A) the party seeking the writ has " 'no other adequate means to attain the relief he desires,' " (quoting Allied Chemical Corp. v. Daiflon, Inc., 449 U.S. 33, 35, 101 S. Ct. 188, 190, 66 L. Ed. 2d 193 (1980)), and (B) the court below has committed a clear error of law. We consider these requisites in turn.

No other paths to appellate review are available to defendants. First, if defendants are required to wait until the final order of the litigation, their appeal on this issue would be valueless. The harms defendants seek to avoid are embarrassment and prejudice in the community at large. Defendants thus require injunctive relief, for compensatory damages would be virtually impossible to assign. Unless the district court's order is vacated, the materials will be released; thereafter, it will be impossible, practically speaking, to rectify the harm. See C & C Products, Inc. v. Messick, 700 F.2d 635, 637-38 (11th Cir.1983) (appeal from district court's modification of a protective order dismissed as moot because the materials had already been released and "no order from this court can undo that situation."). Second, as we have already seen, supra II., the district court's order is not appealable under the collateral order doctrine of Cohen v. Beneficial Industrial Loan Corp.14

Mandamus is not available for abuse of discretion. Rather, we exercise mandamus jurisdiction only if we find that the district court committed a clear error of law. Sporck, 759 F.2d at 314. This requirement is satisfied because the district court made two clear errors of law. First, it misread Seattle Times v. Rhinehart and imposed under Fed.R.Civ.P. 26(c) a more stringent good cause standard than was necessary or appropriate. Second, on account of its misreading of Seattle Times, it exercised plenary review over the magistrate's order when the "clearly erroneous" standard was required. As both of these errors were germane to its decision, see discussion infra, they are independent grounds for reversal. We take these matters up in turn.

IV. THE DISTRICT COURT'S MISREADING OF SEATTLE TIMES

As we have noted, the district court identified an ambiguity in the Seattle Times opinion: it was unclear whether Seattle Times mandated a Rule 26(c) analysis without regard to the first amendment, or whether it required an analysis that included a strict least restrictive means test. See discussion supra pp. 1114-15. See also Post, The Management of Speech: Discretion and Right, 1984 Sup.Ct.Rev. 169, 181-82 (noting the same point). This ambiguity may be significant because the good cause analysis, although by no means toothless, see infra part VI.A. is significantly less stringent than the least restrictive means test. The district court chose the latter alternative without explanation and analyzed the case in first amendment terms, applying the least restrictive means test. See Dist.Ct.Op. at A22. While we recognize the ambiguity in Seattle Times, we believe for several reasons that the district court misinterpreted Seattle Times and that Seattle Times prohibits a court considering a protective order from concerning itself with first amendment considerations.

We recently had opportunity in a case very similar to this one, State of New York v. United States Metal Refining Co., 771 F.2d 796 (3d Cir.1985) to interpret Seattle Times. We found there that Seattle Times confirmed our previous suspicion that protective orders in civil discovery did not require first amendment analysis:

This court has noted that an order prohibiting the disclosure of information obtained under the rules of discovery probably does not run afoul of the first amendment. Rodgers v. United States Steel Corp., 536 F.2d 1001, 1006 (3d Cir.1976).... The Supreme Court confirmed our point of view in the Seattle Times case. 104 S. Ct. at 2009-10.

New York v. United States Metal Refining Co., 771 F.2d at 802. Thus, United States Metal Refining Co. is clear precedent for the interpretation eschewed by the district court.15

This holding would appear to end our inquiry. However, because the district court's interpretation of Seattle Times raises questions not considered in United States Metal Refining Co., it is appropriate and useful to review the Seattle Times opinion in the light of these questions. That review confirms the soundness of United States Metal Refining Co.'s reading of Seattle Times. In the first place, the Supreme Court's holding in Seattle Times was peremptory: "a protective order ... entered on a showing of good cause as required by Rule 26(c) ... does not offend the First Amendment." 104 S. Ct. at 2209-10. This statement leaves no room for lower courts to consider first amendment factors in fashioning or reviewing Rule 26(c) orders. The unequivocal nature of the Court's holding supersedes any ambiguity in its earlier discussion.

Second, the rest of the Supreme Court's opinion, which emphasized that the discovery process was not a forum traditionally open to the public, 104 S. Ct. at 2208, and that the process was "a matter of legislative grace," id. at 2207, to which no first amendment rights attached, is consistent with the position that the first amendment is simply irrelevant to protective orders in civil discovery; it does not comport with the district court's insistence on a less restrictive means test in protective order determinations. Although the Supreme Court's dictum about less restrictive means analysis is to the contrary, see 104 S. Ct. at 2207, this dictum is insufficient to overcome the weight of the Court's holding and the evident direction of the Court's reasoning.

Finally, we note that the overwhelming number of courts that have considered this issue have reached the same conclusion. See Worrell Newspapers of Indiana, Inc. v. Westhafer, 739 F.2d 1219, 1223-24 n. 4 (7th Cir.1984) (in light of Seattle Times, court need only undertake a Rule 26(c) good cause analysis without consideration of First Amendment); Tavoulareas v. Washington Post Co., 737 F.2d 1170, 1172-73 (D.C.Cir.1984) (en banc ) (same); In re Agent Orange Product Liability Litigation, 104 F.R.D. 559, 566 (E.D.N.Y.1985) (same). But see Michelson v. Daly, 590 F. Supp. 261, 266 (N.D.N.Y.1984) (Seattle Times demands a least restrictive alternative test for protective orders in civil discovery). This precedent gives us further confidence in our analysis.

We may summarize thus. Seattle Times required the district court merely to inquire whether the defendants had demonstrated good cause for the protective order; the district court instead applied a least restrictive means test. The good cause standard is significantly less demanding than the least restrictive means test; the court's error, therefore, may have worked a serious detriment to the defendants. The court's error thus constitutes a clear error of law sufficient for our exercise of mandamus jurisdiction.

The district court also erred because it reviewed the magistrate's order under an incorrect standard. Title 28 U.S.C. Sec. 636(b)(1)(A) (1982) explicitly states that the district court may modify the magistrate's order only if the district court finds that the magistrate's ruling was clearly erroneous or contrary to law. The district court in the instant case, however, held that Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 104 S. Ct. 1949, 80 L. Ed. 2d 502 (1984), mandated plenary review regardless of the statutory standard of review. Bose held that when questions of constitutional fact arise in the first amendment context--questions like whether a speaker had "actual malice"; whether speech was libelous or an incitement to riot; whether pictures appeal to "prurient interests" or are "patently offensive"--an appellate court is bound to exercise plenary review on account of the crucial values at stake. Id. 104 S.Ct. at 1961-65. The district court reasoned that because it was acting as an appellate court in reviewing the magistrate's order, Bose should control. Dist.Ct.Op. at A17-A18.

The flaw in this logic stems from the same error discussed above, the district court's misreading of Seattle Times. Believing that Seattle Times made first amendment analysis an important part of its Rule 26(c) inquiry, the district court found that Bose applied. As we have seen, however, Seattle Times says exactly the opposite: that first amendment considerations are irrelevant to Rule 26(c) protective orders. Because the first amendment is irrelevant to the analysis, there are no grounds for extending Bose to this situation. The "clearly erroneous" standard obviously would have been less onerous for the defendants than was the district court's plenary review standard. Thus, the court's error may have harmed the defendants, and this error also constitutes a clear error of law sufficient for our exercise of mandamus jurisdiction.

These errors require that we reverse the district court's judgment and remand for reconsideration of good cause. Although it might be possible for us to review the magistrate's protective order ourselves, we feel it would be unwise to do so. Review of the order will require detailed consideration of the defendants' assertion of good cause. Such consideration would be exceedingly difficult without the district court's prior analysis of the matter under appropriate constitutional standards. Cf. Tavoulareas v. Washington Post, 737 F.2d 1170, 1172 (D.C.Cir.1984) (en banc ) ("It would seem strange for the appellate court ... to decide the 'good cause' question initially--especially when, as here, the District Court, has had no opportunity to decide it free from erroneously imposed constitutional restraints."). It would be equally unwise for us to "tailor" or adjust the order, for the good cause hearing will likely reveal the appropriate shape that the protective order should take and it is thus better that any delineation of specifics await that hearing. We accordingly shall grant the writ, and allow the district court to reconsider the magistrate's protective order in a manner consistent with this opinion.

In view of our holding, the district court will perforce be obliged to take second looks at the good cause issue (no specific good cause findings have been made), and at the magistrate's protective order. With respect to the later issue, we note that our holding has not resolved a critical aspect of the protective order litigation that the record reveals to be still festering: whether the district court was justified in its use of the document-by-document approach as opposed to a broader approach in its reformulation of the magistrate's protective order. These two issues were contested in the district court, and colloquy at oral argument revealed that they are still at issue and will likely arise again. Therefore, we address them for the guidance of the district court.16

Whether defendants have shown good cause for a protective order has been the issue at the heart of this case, and will likely remain so. The defendants assert that although the material they have turned over does not contain trade secrets, it does include materials the dissemination of which would cause them annoyance and embarrassment sufficient to justify a broad protective order. The plaintiffs contend that the defendants have not made a sufficiently convincing showing of the harm they would suffer from dissemination and that their allegations of harm are merely conclusory.

As the district court explained, Rule 26(c) places the burden of persuasion on the party seeking the protective order. To overcome the presumption, the party seeking the protective order must show good cause by demonstrating a particular need for protection. Broad allegations of harm, unsubstantiated by specific examples or articulated reasoning, do not satisfy the Rule 26(c) test. See United States v. Garrett, 571 F.2d 1323, 1326, n. 3 (5th Cir.1978) (requiring "a particular and specific demonstration of fact as distinguished from stereotyped and conclusory statements"); General Dynamics Corp. v. Selb Mfg. Corp., 481 F.2d 1204, 1212 (8th Cir.1973), cert. denied, 414 U.S. 1162, 94 S. Ct. 926, 39 L. Ed. 2d 116 (1974); 8 C. Wright & A. Miller, Federal Practice and Procedure Sec. 2035 (1970 & Supp.1985). Moreover, the harm must be significant, not a mere trifle. See, e.g., Joy v. North, 692 F.2d 880, 894 (2d Cir.1982) (refusing protective order where proponent's only argument in its favor was the broad allegations that the disclosure of certain information would "injure the bank in the industry and local community"), cert. denied sub nom. Citytrust v. Joy, 460 U.S. 1051, 103 S. Ct. 1498, 75 L. Ed. 2d 930 (1983).

Although there appears to be a lurking dispute as to what may constitute good cause for a protective order, see discussion supra at pp. 1114-15, we are satisfied that the district court understood and will apply on remand the principle that Rule 26(c) protects parties from embarrassment as well as from disclosure of trade secrets. We add to the district court's comments only our own understanding that, because release of information not intended by the writer to be for public consumption will almost always have some tendency to embarrass, an applicant for a protective order whose chief concern is embarrassment must demonstrate that the embarrassment will be particularly serious. As embarrassment is usually thought of as a nonmonetizable harm to individuals, it may be especially difficult for a business enterprise, whose primary measure of well-being is presumably monetizable, to argue for a protective order on this ground. Cf. Joy v. North, supra (a protective order will not issue upon the broad allegation that disclosure will result in injury to reputation); to succeed, a business will have to show with some specificity that the embarrassment resulting from dissemination of the information would cause a significant harm to its competitive and financial position.

Under the district court's order, the defendants would be forced to demonstrate to the Court on a document-by-document basis which documents should be protected and not disseminated before they could even be marked "confidential." The district court felt compelled to adopt this solution because it recognized that the burden of persuasion fell on the party seeking the protective order, and it believed that allowing defendants to mark documents confidential in the first instance--bound only by their good faith--and requiring plaintiffs to oppose the confidentiality designation would impermissibly shift the burden of proof to the plaintiffs. Dist.Ct.Op. at A29-30. The defendants object that the district court's order is unduly restrictive and burdensome.

It is correct that the burden of justifying the confidentiality of each and every document sought to be covered by a protective order remains on the party seeking the protective order; any other conclusion would turn Rule 26(c) on its head. That does not mean, however, that the party seeking the protective order must necessarily demonstrate to the court in the first instance on a document-by-document basis that each item should be protected. It is equally consistent with the proper allocation of evidentiary burdens for the court to construct a broad "umbrella" protective order upon a threshold showing by one party (the movant) of good cause. Under this approach, the umbrella order would initially protect all documents that the producing party designated in good faith as confidential.17 After the documents delivered under this umbrella order, the opposing party could indicate precisely which documents it believed to be not confidential, and the movant would have the burden of proof in justifying the protective order with respect to those documents. The burden of proof would be at all times on the movant; only the burden of raising the issue with respect to certain documents would shift to the other party.

As the commentary in the Manual for Complex Litigation Second (MCL 2d) (1985) makes clear, the umbrella order approach has several advantages over the document-by-document method adopted by the district court in a complex case,18 and MCL 2d recommends the use of umbrella orders in complex cases.19 The caselaw also supports the view that the use of umbrella orders in the district court is a useful method of dealing with large-scale discovery.20

There may be cases in which the document-by-document approach adopted by the district court, which deters over-designation of confidentiality and imposes heavier costs on parties making the confidentiality designation, will be preferable. A case in which the district court has reason to believe that virtually all confidentiality designations will be spurious may be such a case. Our purpose in extending the discussion is to explain that the district court erred to the extent that it felt obliged to utilize the document-by-document approach to avoid shifting the burden of proof of confidentiality, and to commend the umbrella approach for consideration of the district courts in this circuit in complex cases.

Because of the district court's misinterpretation of Seattle-Times v. Rhinehart and its consequent errors in defining the appropriate good cause standard and its own scope of review of the magistrate's findings, we will grant the writ.