ORDER

Wade Hemphill was convicted of cocaine conspiracy and distribution charges. He moved for a new trial claiming ineffective assistance of counsel because his appointed counsel allegedly failed to develop his proposed coercion defense. The district court denied the motion, citing a lack of prejudice. Hemphill appeals, challenging that ruling and the district court’s drug-quantity calculation at sentencing, and we affirm.

In November 1999 Hemphill was charged with conspiring with other Gangster Disciples from 1995 to July 1999 to distribute, and possess with intent to distribute, cocaine powder and crack, 21 U.S.C. §§ 846, 841(a)(1), and with distributing 373 grams of crack in November 1998, id. § 841(a)(1). Michael Mann was appointed to represent Hemphill, and after unsuccessful attempts at plea bargaining, the case went to trial. At trial the government’s evidence included Hemphill’s post-arrest confession to involvement in crack dealing and audio tapes of wiretapped conversations between Hemphill and another alleged co-conspirator. A surveillance officer testified that he had observed Hemp-*987Ml receive a package containing 373 grams of crack, and that package was introduced into evidence. Hemphill testified in his own defense and admitted having bought and sold cocaine, but claimed that he had been coerced into doing so by co-conspirators Alex Kinsey and Quincy Jackson. He described a number of incidents in which these men allegedly attacked him and threatened to kill him and his family if he left the drug business. This testimony was Hemphill’s only support for his coercion defense. In June 2000 a jury found him gMty on both counts.

Following the verdicts, attorney Mann withdrew, and new counsel was appointed. Hemphill then moved for a new trial, arguing that Mann had been ineffective. Hemphill claimed that Mann had failed to interview witnesses who could have corroborated his coercion defense: his nieces Okema and Deisha Hemphill, Pierre Stitts, and Terrell Webster. In similar affidavits attached to the motion, Okema and Deisha stated that in September and October 1998, respectively, Hemphill told them that he was being forced to sell drugs by two of his co-defendants, Quincy Jackson and Alex “Rico” Kinsey, and could not go to the police; Okema added that Hemphill attributed his reluctance to go to the police to a belief that one of the men threatening him had contacts within the department who were being paid off. In his affidavit Stitts asserted that he witnessed Jackson and Kinsey-a witness for the government at trial-beat HempMl on one occasion in October 1998. The court held an evidentiary hearing on the motion.

At the hearing attorney Mann testified that Hemphill’s first reference to facts indicating a coercion defense had occurred “a couple of months” before trial. He stated that, at some point during the next month, Hemphill gave him some names of witnesses, including his nieces, Stitts, and Webster. However, he did not contact Okema and Deisha because he believed their evidence would be inadmissable hearsay and, in any event, a jury would view their testimony as biased because they were Hemphill’s nieces. Mann did call and leave several messages for Stitts but never received a response; eventually he received word that Stitts had left the state for the summer. When Mann contacted Webster, Webster told him that he had no information to offer and that he did not wish to be subpoenaed. Therefore, Mann decided that any coercion defense would have to be “proved out of [the defendant’s] own mouth.”

Mann also testified that Hemphill had suggested to him, on the eve of trial, that he call alleged co-conspirator Vernon Everett as a witness. Everett had stated in prior testimony that he had been paying off a Chicago police officer to tip him off to potential raids and snitches. Hemphill believed this information supported his coercion defense by explaining why he had not reported threats against him to the police. Mann, however, did not see a strong connection between Everett’s testimony and the defense, and did not want to risk the admission of testimony regarding additional narcotics trafficking by Hemphill. Therefore, he did not investigate Everett farther.

After hearing the testimony, the district court concluded that only the failure to investigate the possibility of calling Everett warranted further inquiry. The court requested further briefing on this subject, and after receiving the submissions initially granted the motion for a new trial. After the government sought reconsideration, however, the court reversed itself and concluded that Mann’s failure to investigate Everett had not prejudiced Hemphill. The court reasoned that Everett’s past testimony suggested only that police corruption existed, not that it contributed to *988what Hemphill asserted was a “reasonable fear of death or bodily harm.” Therefore, the court concluded, Mann had not rendered constitutionally ineffective assistance in failing to investigate whether to call Everett as a witness.

Hemphill was sentenced in December 2001. The court rejected Hemphill’s request for a downward departure based on coercion, explaining that it did not find the defendant’s witnesses, including Okema and Deisha Hemphill, credible. Relying on testimony from Kinsey and Hemphill himself as to the amount of crack Hemphill had purchased, the court found that Hemphill had transacted in over 1.5 kilograms of crack. Hemphill was sentenced to concurrent 360-month terms, the low end of the guideline range.

On appeal, Hemphill argues that the district court erred in determining that attorney Mann was not ineffective. To prevail on an ineffective assistance claim, a defendant must show that 1) his attorney’s performance fell below an objective standard of reasonableness, and 2) there is a reasonable probability that, but for counsel’s errors, the result of the proceeding would have been different. Strickland v. Washington, 466 U.S. 668, 688, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984). We review the questions of deficient performance and prejudice de novo. Id. at 698, 104 S.Ct. 2052. The district court’s findings of fact are reviewed for clear error. White v. Godinez, 301 F.3d 796, 800 (7th Cir. 2002).

First, Hemphill claims that Mann performed deficiently because he failed to investigate the testimony of potential witnesses Stitts, Everett, and Okema and Deisha Hemphill. He argues that these witnesses could have corroborated his testimony that Kinsey and Jackson attacked him and threatened his life, and that he could not go to the police because they had been bribed. Although he recognizes that Stitts was difficult to contact and that Okema and Deisha were biased, he argues that Mann should have made more of an effort to obtain and introduce this testimony at trial.

We judge an attorney’s performance according to the prevailing professional norms, Washington v. Smith, 219 F.3d 620, 629 (7th Cir. 2000), giving great deference to counsel’s strategic choices viewed as of the time at which they were made, White, 301 F.3d at 802. Failure to properly investigate a defendant’s case may constitute ineffective assistance, but defense counsel need not “track down every lead or ... personally investigate every evidentiary possibility before choosing a defense and developing it.” United States v. Balzano, 916 F.2d 1273, 1294 (7th Cir. 1990) (quotations and citation omitted). Where a witness is known to be exceedingly important, repeated efforts at contact may be necessary. Washington, 219 F.3d at 629-30.

In this case, we believe that Mann’s performance was deficient. Although we agree with Mann that Okema and Deisha’s testimony would have been of questionable admissibility and persuasive power, Man had no compelling strategic reason not to investigate Stitts or Everett. It is true that Mann made an effort to contact Stitts, calling him several times and leaving a message, but given Stitts’ importance as the only eyewitness to attacks on Hemp-hill, Mann should have been more persistent. See Washington, 219 F.3d at 629-30. Further, simply because one of Hemphill’s other witnesses refused to testify, Mann could not assume that Stitts would do the same. Avila v. Galaza, 297 F.3d 911, 920 (9th Cir. 2002). Similarly, although the value of Everett’s testimony was uncertain, it should have been developed through further investigation. White, 301 F.3d at 802. Mann had no substantive reason to reject *989these witnesses; his decision was based on convenience and not on content, and cannot be called strategic. See United States v. Olson, 846 F.2d 1103, 1109 (7th Cir. 1988); United States ex rel Hampton v. Leibach, 347 F.3d 219, 251 (7th Cir. 2003). Viewed as a whole, Mann’s performance was deficient.

We cannot, however, say that Hemphill was prejudiced by Mann’s performance. Hemphill argues that, had Mann performed adequately, the testimony of Stitts, Everett, and Hemphill’s nieces would have been admitted and convinced the jury that he had no reasonable opportunity to avoid death or serious bodily injury if he did not commit the crime. We cannot agree.

“Prejudice” means a “reasonable probability” of a different result without the errors; i.e., a “probability sufficient to undermine confidence in the outcome.” Strickland, 466 U.S. at 694, 104 S.Ct. 2052. The defendant must provide a “comprehensive showing as to what the investigation would have produced.” Montgomery v. Petersen, 846 F.2d 407, 415 (7th Cir. 1988). We examine all the errors made by counsel in the aggregate to determine their effect on the trial, rather than looking at each error individually. Washington, 219 F.3d at 633. Of course, the strength of the case against the defendant is of paramount importance in determining the effect of errors. Id.

In this case, Hemphill’s coercion defense faced an uphill battle. The government’s tapes of Hemphill talking cheerfully and casually with Kinsey and Jackson seriously undermined his claims of fear and coercion. See United States v. Crowder, 36 F.3d 691, 697 (7th Cir. 1994). The government also pointed out that Hemphill did not inform authorities of any threats against him prior to trial. We do not believe that additional witnesses would have changed the result here, especially given the trial judge’s determination at sentencing that Hemphill’s new witnesses were not themselves particularly credible. See Fortenberry v. Haley, 297 F.3d 1213, 1227 (11th Cir. 2002).

Additionally, even had Mann uncovered all of the evidence of coercion Hemphill claims was overlooked, Hemphill would not have been able to show that the threat he feared was “immediate” and that he had no reasonable opportunity to avoid it, as is required for a coercion defense. United States v. Jocic, 207 F.3d 889, 892 (7th Cir. 2000). Hemphill and his proposed witnesses have suggested only that he feared something might happen to him in the future if he did not sell drugs. They have not proffered evidence of a specific and immediate threat. United States v. Fiore, 178 F.3d 917, 922-923 (7th Cir. 1999). Further, even assuming that Hemphill did not have the alternative of going to the police, none of his proposed evidence would have explained why he did not simply flee at some point during the conspiracy. See Jocic, 207 F.3d at 892. So although we agree with Hemphill that counsel could have more zealously pursued his defense, we do not believe that counsel’s omissions prejudiced the outcome of Hemphill’s trial.

We now turn to Hemphill’s challenge to his sentence, specifically to the district court’s determination that he had transacted in over 1.5 kilograms of crack. Hemphill argues that this calculation rests on the testimony of co-conspirator Kinsey, and that in crediting Kinsey the court erroneously concluded that everything Hemphill bought from Kinsey was crack, when in fact Kinsey indicated that some of the purchases were of cocaine powder.

We review a sentencing determination for clear error, which exists only if there is a “definite and firm conviction that a mistake has been committed.” United States v. Chatmon, 324 F.3d 889, 892 (7th Cir. *9902003). The government bears the burden of establishing the quantity of the drugs and the fact that they were in “crack” form by a preponderance of the evidence. United States v. Adams, 125 F.3d 586, 591 (7th Cir. 1997). The court may determine quantity by multiplying the smallest amount of drug obtained by the minimum number of purchases made. United States v. Acosta, 85 F.3d 275, 282 (7th Cir. 1996). The evidence supporting the determination of quantity need not be entirely without inconsistency, but it must be basically reliable. See Chatmon, 324 F.3d at 892; United States v. Noble, 299 F.3d 907, 911 (7th Cir. 2002).

In this case, the district court’s calculation was correct. Hemphill himself admitted to having transacted in from one-eighth to three-eighths of a kilogram of crack on eight to ten occasions. Kinsey’s testimony clarified that at least two of those occasions involved the three-eighths amount. Even using the low figures, this alone accounts for 1.5 kilograms of crack. This calculation was further corroborated by the lab analysis of the 373.1 grams intercepted by the police. Although it is clear that Hemphill also purchased powder cocaine, the specific evidence of crack transactions is sufficient to support the court’s calculation. The district court calculated both the drug quantity and Hemp-hill’s sentence correctly.

For the reasons stated above, we AFFIRM the district court’s judgment.