Opinion for the Court filed by Chief Judge GINSBURG.
GINSBURG, Chief Judge:Mary Ann Workman, the mother of a woman murdered in Somalia while on a mission for the United Methodist Committee on Relief (UMCOR), sued UMCOR, alleging that the defendant negligently failed to prepare her daughter for and protect her from the dangers of her assignment. We hold that the district court properly entered summary judgment for the defendant on the ground that Workman had failed to meet the “heightened foreseeability standard” applicable under the law of the District of Columbia to a case in which the harm arises from the criminal act of a third party.
I. Background
Deena Marie Umbarger, a 1994 graduate of the Yale Law School, left her law practice in 1997 to become a relief worker for UMCOR. Her first international assignment was to Kenya, followed by a brief stint in the Republic of Georgia. In November 1998 she returned to Kenya to live with her fiancé. There UMCOR retained her under a series of short-term consulting agreements until her death in March 1999.
The agreements defined Umbarger’s role as that of an independent contractor charged with evaluating UMCOR’s past relief efforts in Kenya and Somalia and with developing a plan for the future, possibly to include the opening of an UMCOR office in Nairobi. Umbarger was UM-COR’s sole representative in the region; she reported directly to Kemba Eneas, UMCOR’s Program Director for Africa and the Caribbean, who was based in Virginia.
On March 20, 1999 Umbarger traveled to Somalia. An UMCOR memorandum contains the following description, based upon a report from the United States Embassy in Kenya, of the events of that day:
At approximately 8:00 p.m., Deena and Ahmed Hussein [of Somali Community Services/Nairobi] arrived at Nkokani Airport, NW of Lamu, Kenya. Eight people in a vehicle with a Somalian license tag picked her up. Deena and her counterparts registered with the local police at the border to explain their business with the community. Six of the eight pei’sons were there on the invitation of SCS, the other two we[r]e from the Aqwin Militia who were warlords there to monitor their activities. At 5:00 p.m. they arrived at a restaurant 40 km outside of Kiangu, on the border of Kenya and Somalia (in the SW corner of Somalia) where they were to meet five town elders. At 5:30 p.m. before eating, a gunman murdered Deena in the restaurant. The elders caught the gunman; however, other members of the militia shot in the air which startled the elders and caused the gunman’s release. The persons that were with Deena put her back in the vehicle and drove her to the nearest clinic or hospital in Kenya whe[r]e she was dead on arrival.
In February 2000 Workman filed a complaint in the Superior Court for the District of Columbia alleging that UMCOR had “failed to discharge its duty and obligation to protect [Umbarger] as she assisted UMCOR in carrying out its activities in Kenya and Somalia.” Workman sought damages for Umbarger’s death under the District of Columbia Wrongful Death Act, D.C.Code § 16-2701, and under the Survival Statute, D.C.Code § 12-101, which allows a legal representative to bring a claim on behalf of a person who has died. UMCOR, invoking federal diversity jurisdiction, removed the case to the district court. In a ruling not challenged here, the district court dismissed the wrongful death claim because the statute applies only to deaths occurring within the District of Columbia.
*133At the close of discovery, UMCOR moved for summary judgment on the negligence claim under the Survival Statute arguing, as relevant here, that because UMCOR could not have foreseen Umbarger’s murder, it had no duty to protect her from that crime. In support of its motion, UMCOR submitted declarations from Eneas and from Sam W. Dixon, the head of UMCOR’s Nongovernmental Organization Unit. Dixon stated that Umbarger’s duties were “to determine whether there were sources of funding for relief projects in the region and whether the region was safe and secure enough to justify a permanent UMCOR presence,” and that Umbarger was “solely ... responsible]” for the manner in which she carried out these duties, including her decision to travel to Somalia. Dixon characterized Umbarger asUMCOR’s “eyes and ears” in the region; indeed, “[t]he information that UMCOR had about potential security risks in Somalia came primarily from [her].” Dixon also stated:
I did not have, nor was I aware of anyone at UMCOR who had, any advance knowledge or warning that ... her travel to Somalia or work in Kenya would put her at some unique risk beyond the typical security risks known to anyone — including Deena — who provides relief to people living in the most dangerous areas of the world.
Eneas supported Dixon’s description of the relationship between UMCOR and Umbarger. She stated that although Umbarger reported to her, Eneas did not “supervise” Umbarger and did not require Umbarger to travel to Somalia. Eneas stated that Umbarger had gone to Somalia once before, “without incident.” In addition, Eneas reported that during her visit to Nairobi about a week prior to the fateful trip, “[w]e spent a large amount of time together and Deena never told me that she feared traveling to Somalia or that she was concerned that she might be harmed in any way during that trip. To the contrary, Deena was eager to travel to Somalia.”
In opposition to UMCOR’s motion for summary judgment, Workman submitted the declaration of Michael D. O’Neill, the Coordinator for Volunteer Safety and Overseas Security for the Peace Corps. According to O’Neill, even in cases in which decision-making is left to the aid worker, “the institution for whom she works also is expected to provide reasonable safety and security support,” including “proper security assessment of the area in question, appropriate risk reduction strategies, a security management system, a program for sharing information, and training.” By way of example, O’Neill noted that the Peace Corps provides 12 weeks of training to volunteers before they are posted abroad. O’Neill stated that upon his review of the evidence, he had seen nothing to indicate that Umbarger received “proper training and guidance” or that UMCOR maintained a “reasonable safety and security support system” for its overseas workers.
Workman also produced a February 1999 email from Umbarger to Eneas, in which Umbarger had noted possible irregularities in the accounting of relief funds earmarked for Somali Community Services and expressed concern that her investigation of the issue could render a planned trip to Somalia particularly dangerous. She wrote:
I need to spend a fair amount of time with Fawzi [the SCS liaison to UM-COR] getting the particulars of just what was provided to beneficiaries. The local SCS guys here are quite concerned that some of what Fawzi put as outputs in the one final report I do have was actually not correct. This could be a very sensitive (not to *134mention dangerous) issue as we try to confirm some of these things on the ground, particularly if the local people begin to suspect that they never received all of what SCS was supposed to provide. They have told me that a hostage/kidnapping situation would not be farfetched, and I am sure you would be as unhappy receiving a midnight phone call for help as I would be if I had to make one.
In addition, Workman produced an email to Shelly Sutherland, whom Workman described as “a long time UMCOR employee,” in which Umbarger wrote: “I feel like I am at the breaking point with no support system or release valve. The way things are going it wouldn’t surprise me if something happened in Somalia.”
The district court held that Workman’s evidence was insufficient as a matter of law to satisfy the “stringent showing of heightened foreseeability required under District of Columbia law” in a case in which the criminal act of a third party causes injury. On appeal, Workman argues that, in the light of all the evidence, the district court’s conclusion was erroneous.
II. Analysis
In the District of Columbia, a defendant may be liable for harm caused by the criminal act of another only if the crime was particularly foreseeable. See, e.g., Potts v. District of Columbia, 697 A.2d 1249, 1252 (D.C.1997) (“Where an injury is caused by the intervening criminal act of a third party, this court has repeatedly held that liability depends upon a more heightened showing of foreseeability than would be required if the act were merely negligent”). Courts in the District of Columbia regularly grant a defendant’s motion for summary judgment on the ground that the plaintiffs evidence is “insufficient ... to establish that [a] criminal act ... was reasonably foreseeable.” Bailey v. District of Columbia, 668 A.2d 817, 822 (D.C.1995); Potts, 697 A.2d at 1252-53. We review de novo the district court’s determination that Workman failed to present sufficient evidence to preclude summary judgment. DeGraff v. District of Columbia, 120 F.3d 298, 301 (D.C.Cir.1997).
How foreseeable must the crime have been for the plaintiff to get her claim before a jury? The District of Columbia Court of Appeals has said a “heightened showing” is required, the requirement is a “demanding” one, and the proof must be “precise.” Potts, 697 A.2d at 1252. Foreseeability cannot be predicated upon “generic information” such as crime rates, Bailey, 668 A.2d at 820, or evidence that the defendant’s employees worked in a “criminally active environment,” Clement v. Peoples Drug Store, Inc., 634 A.2d 425, 429 (D.C.1993). The plaintiff is not, however, required to show “previous occurrences of the particular type of harm”; the requirement “can be met instead by a combination of factors which give [the] defendant ] an increased awareness of the danger of a particular criminal act.” District of Columbia v. Doe, 524 A.2d 30, 33 (D.C.1987).
Without more specific guidance, we are left to reason by analogy from D.C. cases, as the district court did here. We note first that the Court of Appeals has been reluctant to see a defendant held liable for harm caused by the criminal act of a third party. See, e.g., McKethean v. Washington Metropolitan Area Transit Authority, 588 A.2d 708, 717 (D.C.1991) (evidence regarding rate of collisions at bus stops and high frequency of traffic accidents involving alcohol and drug abuse insufficient); Bailey, 668 A.2d at 820 (finding insufficient “affidavits of witnesses who *135asserted that the area around the school was a ‘high drug area’ and that shootings occurred in that neighborhood”); Clement, 634 A.2d at 427 (similar); Potts, 697 A.2d at 1252 (summary judgment warranted in light of plaintiffs failure to provide evidence of any prior gun-related violence at similar events, “nor any other specific evidence bearing directly on the foreseeability of the shooting incident at issue”). The Court’s reluctance is not surprising, because the heightened foreseeability test provides a limited exception to the “general rule of nonliability” for such claims. Romero v. Nat’l Rifle Ass’n of America, Inc., 749 F.2d 77, 81 (D.C.Cir.1984).
Workman, naturally, relies upon the few cases in which a court has found sufficient evidence of foreseeability either to allow a claim to go to the jury or to uphold a jury verdict for the plaintiff. In District of Columbia v. Doe, the Court of Appeals upheld a jury verdict against the District after a fourth grader was abducted from a classroom and raped. The plaintiffs evidence indicated that “school officials were on notice of the danger to students from assaultive criminal conduct by intruders.” 524 A.2d at 33-34. That evidence included: “crimes against persons in and around the school — an arson in the school and a robbery on the school’s playground; sexual assaults and other violent activity in the surrounding area; and deficient school security — the open rear gate, broken doors, malfunctioning intercom, and presence of adult males who freely roamed throughout the school.” Id. at 34. In those circumstances it would seem only a matter of time until one of the District’s charges was abducted or molested.
In Doe v. Dominion Bank of Washington, 963 F.2d 1552 (1992), a case in which the plaintiff was raped on a vacant floor of an office building during business hours, this court reversed the district court’s grant of a directed verdict in favor of the defendant landlord. In the month preceding the rape, tenants had lodged a number of complaints about threatening intruders in the building. Id. at 1555-56. There had also been reported thefts of personal property from offices in the building and drug use and sexual activity in a building rest room. Id. Again, another crime was just waiting to befall a tenant.
In these cases a good deal of evidence suggested the defendant was on notice there was a substantial risk of harm to the plaintiff; in addition, the relationship between the plaintiff and the defendant suggested the defendant should be held liable as a matter of policy. For example, in Dominion Bank the court imposed a duty of protection because the landlord was in the better position both to know about security threats and to protect against them. Id. at 1559 (relying upon “inability of an individual tenant to control the security of common hallways, elevators, stairwells, and lobbies”). Similarly, District of Columbia v. Doe involved the District’s duty of custodial care for schoolchildren — a point of distinction noted in other cases, including the decision here under review. See, e.g., Clement, 634 A.2d at 429 (“Implicit in Doe’s holding was the notion that particular care is required by school officials when the safety of young children is involved”); Bailey, 668 A.2d at 821 (noting that “victim was young and she was taken from a place that we would expect to be a safe haven, i.e., her classroom”).
From our review of the D.C. cases, we see the requirement that the defendant have been able to foresee that a third party would likely commit a criminal act ordinarily has, and perhaps must have, a relational component. See Romero, 749 F.2d at 81 (“The only District cases departing from [the ‘general rule of nonliability’] involved either a special relationship *136between the parties to the suit ... or a relationship of control between the defendant and the intervening criminal actor”). Indeed, the cases suggest a sliding scale: If the relationship between the parties strongly suggests a duty of protection, then specific evidence of foreseeability is less important, whereas if the relationship is not of a type that entails a duty of protection, then the evidentiary hurdle is higher.
In the case at hand this principle supports the district court’s grant of summary judgment in favor of UMCOR. First, the evidence regarding the relationship between the parties does not suggest that UMCOR should be held responsible for Umbarger’s safety. UMCOR did not send Umbarger to the region; she went there for personal reasons, after which UMCOR contracted with her to assess factors, including security risks, important to UM-COR’s decision whether to open an office there. Umbarger decided how best to early out her mission, including whether and when to travel to Somalia. Because UM-COR was in no better position to provide for Umbarger’s safety than was Umbarger herself, the relationship between the parties is unlike those in which the courts have found a duty of protection. See Kline v. 1500 Massachusetts Ave. Apartment Corp., 489 F.2d 477, 483 (D.C.Cir.1970) (in general, no duty to take precautions against foreseeable criminal attack unless “the ability of one of the parties to provide for his own protection has been limited in some way by his submission to the control of the other”).
Insofar as the relationship between the parties in this case tilts against holding the defendant liable for the harm caused by someone else’s criminal act, the plaintiffs evidence of foreseeability must be particularly strong to survive summary judgment — which it is not. Although there can be no doubt that Umbarger faced far greater risks in Somalia than she would have in a less volatile part of the world, the local courts, as the learned district judge noted, give little or no weight to “generic information” regarding the dangerousness of an area. Bailey, 668 A.2d at 820; Clement, 634 A.2d at 429. The O’Neill declaration, which Workman complains the district court ignored, is at best evidence of this generic type, insofar as it reflects at all upon foreseeability.
Workman’s only specific evidence of foreseeability consisted of the email messages Umbarger sent to Eneas and Sutherland. As for the latter, UMCOR notes, and Workman does not dispute, that there was no evidence that Sutherland was an UMCOR employee at the time she received Umbarger’s email, or that its contents were ever conveyed to anyone at UMCOR. Although Umbarger did in the message to Eneas describe the possibility of her being kidnapped as “not ... farfetched,” there was no evidence that the contingency Umbarger feared might put her in danger — namely, the “local people begin[ning] to suspect that they never received all of [the aid] SCS was supposed to provide” — ever materialized. Moreover, again as noted by the district court, Umbarger in the same message sought to extend her contract and remarked that “[t]hings seem really positive here, and I would really like to follow them up.” Umbarger later made an initial visit to Somalia without incident, and did not reiterate any concern to Eneas when Eneas visited her in Kenya shortly before Umbarger’s second trip to Somalia. In these circumstances, all of which were known to UM-COR, Workman simply asks too much when she says that “UMCOR should have read these emails as a plea from a proud woman asking them to keep her from going to Somalia.” The district court correctly concluded that Umbarger’s message did not put UMCOR on notice of the danger that materialized.
*137We recognize that our approach in resolving this appeal — that is, considering the relationship between the parties in establishing the degree of foreseeability necessary to support liability — might offend a doctrinaire. Ordinarily, the relationship between the parties is the key to determining whether the defendant had a legally enforceable duty to the plaintiff (or her decedent), whereas foreseeability is important to issues of proximate causation and conformity to the standard of care, issues that arise only after a duty has been found. See, e.g., William L. Prosser, et al., The Law of Torts § 53 (5th ed.1984) (“[i]t is better to reserve ‘duty’ for the problem of the relation between individuals which imposes upon one a legal obligation for the benefit of the other ... ”); Dominion Bank, 963 F.2d at 1560 (“Once it is established that duties, generally, are owed to one party ... by another ... the foreseeability of the risk at issue in a particular case might seem more applicable to determining whether the standard of care has been breached”); cf. id. at 1564 (Williams, J., concurring) (foreseeability, “[t]hough framed as an issue relating to ‘duty,’ [at least in the context of] landlord liability for criminal acts [of another] ... merges with the proximate cause analysis”).
Sitting in diversity, however, our task is to apply the law of the District of Columbia as its own courts would apply it, not to second-guess the analytical framework those courts have erected. See Dominion Bank, 963 F.2d at 1560. And the D.C. courts have repeatedly spoken of the heightened foreseeability requirement in terms of duty. E.g., Graham v. M. & J. Corp., 424 A.2d 103, 105 (D.C.1980) (“[floreseeability is the key element in establishing the landlord’s duty” in case involving intervening criminal act); Potts, 697 A.2d at 1252 (“the plaintiff has the burden of establishing that the criminal act was so foreseeable that a duty arises to guard against it”). Although the courts did, in early decisions considering liability for harm caused by the criminal act of a third party, inquire at the threshold into the propriety of imposing a duty of protection based upon the relationship between the plaintiff and the defendant, e.g., Kline, 439 F.2d at 483, the courts have in more recent cases tended to leapfrog directly to the foreseeability issue, with the parties’ relationship, as noted above, a factor relevant to determining whether the requirement of foreseeability has been satisfied. Although perhaps not analytically the most obvious approach, neither does it strike us as imprudent. But even if it did, we would be bound to apply it.
III. Conclusion
Because Workman failed to adduce sufficient evidence to create a genuine issue of heightened foreseeability, the judgment of the district court is
Affirmed.