FOR PUBLICATION
  UNITED STATES COURT OF APPEALS
       FOR THE NINTH CIRCUIT

UNITED STATES OF AMERICA,                        No. 07-10289
                  Plaintiff-Appellee,
                 v.                                D.C. No.
                                                 CR-98-00198-HG
JANICE L. PEREZ,
                                                    OPINION
              Defendant-Appellant.
                                           
         Appeal from the United States District Court
                  for the District of Hawaii
           Helen Gillmor, District Judge, Presiding

                  Argued and Submitted
       September 25, 2007—San Francisco, California

                        Filed May 16, 2008

       Before: John R. Gibson,* Marsha S. Berzon, and
                Carlos T. Bea, Circuit Judges.

                      Opinion by Judge Bea




   *The Honorable John R. Gibson, Senior United States Circuit Judge for
the Eighth Circuit, sitting by designation.

                                 5719
5722                 UNITED STATES v. PEREZ


                           COUNSEL

Peter C. Wolf, Federal Public Defender, Matthew C. Winter,
Assistant Federal Public Defender, Honolulu, Hawaii, for
defendant-appellant Janice Perez.

Edward H. Kubo, Jr., United States Attorney, Tracy A. Hino,
Assistant United States Attorney, Honolulu, Hawaii, for
plaintiff-appellee United States of America.


                            OPINION

BEA, Circuit Judge:

   We are called upon in this case to decide whether a person
on supervised release has a right to cross-examine the labora-
tory technician who tested a urine sample containing an ille-
gal drug, where: (1) the test report itself stated the sample was
“dilute”1—meaning the urine sample had been combined with
another liquid at some point before or during the testing; (2)
the evidence presented showed the person on supervised
release did not have an opportunity herself to dilute nor add




  1
    “Dilute” means to “to make thinner or more liquid by admixture.”
Merriam-Webster’s Online Dictionary http://www.merriam-webster.com
(last visited April 23, 2008).
                    UNITED STATES v. PEREZ                 5723
a substance to the sample; and (3) the result of the urinalysis
was critical to support a finding that the person on supervised
release had possessed or used illegal drugs. Because the ille-
gal drug found in the tested sample could only have come
from Perez’s urine, an added substance, or another liquid, and
because the validity of the urinalysis was the critical issue in
determining whether Perez had used cocaine, we reverse the
district court and hold the person on supervised release had a
right to cross-examine the technician who handled and tested
the sample. Absent such cross examination, the urinalysis
should not have been admitted, and the consequent revocation
of supervised release cannot stand.

   We caution that this is an unusual case with unusual facts
and should not be taken out of context. We do not hold that
a releasee always has a right to cross-examine the technician
who tested a urine sample. This is not a case where other evi-
dence was offered in support of revocation, such as illegal
drugs discovered in the possession of the releasee. Nor is this
a case where multiple urine samples each tested positive.
Here, the urinalysis report was the critical piece of evidence
presented in support of the charge that Perez tested positive
for cocaine. Although urinalysis results may often be suffi-
ciently reliable evidence that the opportunity for cross-
examination is unnecessary for due process purposes, see
United States v. Martin, 

, 313-14 (9th Cir. 1993),
here the report itself showed the sample had been adulterated.
Given that the sample was uncontestably adulterated, the test
results were in fact ineluctably unreliable.

                               I

   In 1998, Janice Perez pleaded guilty to bank robbery in vio-
lation of 18 U.S.C. § 2113(a). She was sentenced to a term of
seventy-seven months imprisonment and three years super-
vised release.

  Perez’s term of supervised release commenced on March 8,
2007. Her release conditions required Perez to “follow
5724                UNITED STATES v. PEREZ
instructions of the probation officer” and to abstain from
using alcohol or drugs not prescribed for her.

   On March 15th, Perez wished to travel to Maui to see her
gravely ill brother, who was on life support. After verifying
her brother’s condition, Perez’s probation officer, Derek Kim,
permitted her to go to Maui, but directed her to come to his
office upon her return, provide her travel itinerary and submit
to a drug test.

   When Perez returned, she went to the probation office and
submitted her travel itinerary, but left without taking a drug
test. Kim telephoned Perez to ask why she did not stay to be
tested; Perez told him she forgot because she was upset after
learning her brother had been taken off life support that day
and had passed away.

   Kim already knew about her brother’s death, and testified,
“so when she told me she forgot because she was upset, you
know, I didn’t think anything of it other than the fact that,
well, I could see that that’s conceivable.” Kim did not ask
Perez to come back into the office that day for a drug test.
Instead, Kim instructed Perez to come by the office for a drug
test the following day (March 20), which she did. That test
established there were no illegal drugs in Perez’s system.

  Kim then directed Perez to report to Freedom Recovery
Services (“FRS”), the outside contractor the probation office
uses to monitor persons on supervised release, to undergo
FRS’s orientation.

  On March 21st, the day after she passed a urine test at the
probation office, Perez duly reported to FRS. Even though
Kim had not directed her to get tested, Perez voluntarily sub-
mitted to another urinalysis.

  Mrs. Billee Schnaible, who worked at FRS, testified she
observed Perez give the urine sample; neither Perez nor the
                       UNITED STATES v. PEREZ                       5725
sample were ever out of Schnaible’s sight. The bathroom used
was specially equipped with mirrors allowing Mrs. Schnaible
to observe Perez and ensure she did not have an opportunity
to adulterate the sample. Mrs. Schnaible testified Perez did
not adulterate or “dilute” the sample.

   According to a dip-stick test, the sample tested positive for
cocaine metabolites. Perez protested vigorously saying, “I just
tested at my PO’s [Probation Officer’s] yesterday and I’m
clean, I’m clean, it can’t be.” Mrs. Schnaible testified that
their usual practice would have been to conduct another test,
but she did not remember collecting a second urine sample or
doing a second test.2

   Later, Dennis Schnaible, another FRS staff member, testi-
fied that after Perez protested, Mrs. Schnaible “got another
testing device, a new one, and dipped it into the same urine.
And the results were the same.” Mr. Schnaible then assured
the lid was on tight so the sample would not leak during ship-
ping, watched Mrs. Schnaible fill out a chain of custody form,
and sent Perez’s urine sample to Scientific Testing Laborato-
ries, Inc. (“STLI”) in Virginia.3 STLI is the designated testing
laboratory for U.S. Probation and Pretrial Services. According
to Mr. Schnaible, no second urine sample was taken from
Perez.

   That same day, on March 21st, after Perez had been to
FRS, she returned to the probation office. Kim was out of
town, but Perez asked another probation officer named Robin
DeMello to take a urine sample and test it for her; DeMello
refused.
  2
    There was no evidence one way or the other as to whether either of the
Schnaibles adulterated the sample, although both testified and were cross-
examined.
  3
    No one suggests either Schnaible adulterated the urine sample.
5726                UNITED STATES v. PEREZ
   Mr. Schnaible testified the testing method used by FRS is
not always reliable, which is why FRS uses the laboratories
at STLI. He recalled one time when the method used by FRS
resulted in four samples testing positive for drugs, but when
those samples were sent to STLI, three of them tested nega-
tive.

   The STLI report showed Perez’s sample tested positive for
cocaine. It also recorded abnormally low readings for the
presence of creatinine (a breakdown product of creatine, an
important part of muscle, produced by one’s kidneys) and the
urine’s specific gravity. The report stated the following:
“Specimen Validity Status: DILUTE.” Perez objected that the
STLI urinalysis report was hearsay, but it was nonetheless
introduced into evidence. No STLI employee testified regard-
ing the urinalysis report.

   Also over Perez’s hearsay objection, Kim testified the low
creatine and specific gravity values meant that “rather than
drinking tons of water, or other types of liquid, that [the peo-
ple at STLI] actually believe a dilute, meaning that it was a
liquid which included some urine.” That the sample was
diluted means that it had been adulterated. It not only con-
tained Perez’s urine, it also contained another unknown liquid
that had been combined with the urine. Kim then speculated
that someone could produce a dilute sample by using a hidden
container to pour adulterated urine into the sample cup.

   Perez argued to the district court the urinalysis result must
be rejected because the sample had been adulterated. Perez
contended she could not have adulterated the sample because
she had been under the vigilant eye of Billee Schnaible when
she gave the sample. She also contended her adamance she
had not taken drugs and her multiple attempts to provide an
additional sample that same day—at both FRS and then at the
probation office—cast doubt on the accuracy of the urinalysis.
The government responded that the sample was still available
for retesting.
                    UNITED STATES v. PEREZ                5727
   The district court rejected Perez’s argument, concluding
Perez had no credibility because of a long history of dishon-
esty. The district court made it clear its adverse credibility
determination was not based on any evidence Perez diluted
her own urine; the only grounds cited by the district court
were some vague references to prior bad acts of the defen-
dant. It is noteworthy there was no evidence introduced at the
revocation hearing of any such prior bad acts. Likewise, there
was no statement by the district court that anything about
Perez’s demeanor or manner of testifying caused her to lose
credibility in the district court’s eyes.

   The district court found Perez guilty of both charges in the
Request for Course of Action, specifically finding that Perez:
(1) “refused to submit to drug testing at the Probation Office
on 3/19/2007 as instructed on 3/15/2007”; and (2) Perez’s
“urine specimen submitted on 3/21/2007 tested positive for
cocaine.”

                              II

   We have jurisdiction under 28 U.S.C. § 1291 and 18 U.S.C.
§ 3742. We review the district court’s decision to revoke a
term of supervised release for an abuse of discretion. United
States v. Verduzco, 

, 1184 (9th Cir. 2003).
Whether a defendant has received due process at a revocation
proceeding is a mixed question of law and fact we review de
novo. See United States v. Havier, 

, 1092 (9th
Cir. 1998). “A due process violation at a revocation proceed-
ing is subject to harmless error analysis.” 

(internal quotation marks omitted). We reverse.

                              III

  A district court may revoke a term of supervised release
only if it “finds by a preponderance of the evidence that the
defendant violated a condition of supervised release.” 18
U.S.C. § 3583(e)(3); see also Morrissey v. Brewer, 

               UNITED STATES v. PEREZ
471, 484 (1972) (holding a revocation cannot stand on “erro-
neous information”); United States v. Martin, 

,
310 (9th Cir. 1993) (holding a revocation of supervised
release must be “based on verified facts”). Although this is a
lower standard than the “beyond a reasonable doubt” standard
required for a criminal conviction, there must still be credible
evidence the releasee actually violated the terms of supervised
release.

                              A.

   [1] With regard to the first count—that Perez “refused” to
submit to drug testing on March 19th—there is no evidence
Perez intentionally refused to be tested. True enough, Perez
was ordered to submit to drug testing after her return from
Maui, and she did not do so. The charge, however, was not
“failure” to submit to testing, the charge was “refusing” to
submit to testing. The only evidence regarding Perez’s refusal
was in her favor. She told Kim she forgot to stay for drug test-
ing on March 19th because her brother died that day. Kim
testified he verified the brother’s death and consequently
“thought nothing” of her failure to stay for drug testing when
she dropped off her travel itinerary. He accepted her explana-
tion as “conceivable.” He did not ask her to come back into
the office that day, and she did not refuse to do so. Further,
she submitted to a drug test the very next day, and tested neg-
ative for the presence of drugs. Accordingly, there was no evi-
dence from which the trial court could find it was more
probable than not Perez “refused” to submit to drug testing.

                              B.

   With regard to the second count—that Perez’s urine speci-
men submitted on 3/21/2007 tested positive for cocaine—
although the admission of a urinalysis without the opportunity
for cross-examination is not necessarily error, given the par-
ticular facts of this case, the admission of this evidence with-
                     UNITED STATES v. PEREZ                  5729
out allowing Perez to cross-examine the laboratory technician
was error.

   [2] Admission of hearsay evidence in revocation of super-
vised release proceedings is governed by the Fifth Amend-
ment right to due process. See United States v. Hall, 

, 985 & n.4 (9th Cir. 2005). A “releasee is guaranteed the
right to confront and cross-examine adverse witnesses at a
revocation hearing, unless the government shows good cause
for not producing the witnesses.” 

(inter-
nal quotation marks omitted); see also Fed. R. Crim. P.
32.1(b)(2)(C) (a releasee is entitled to “an opportunity to
appear, present evidence, and question any adverse witness
unless the court determines that the interest of justice does not
require the witness to appear”).

   [3] In Martin, we construed this right “as requiring that a
supervised releasee receive a fair and meaningful opportunity
to refute or impeach the evidence against him in order ‘to
assure that the finding of a [supervised release] violation will
be based on verified facts.’ 

(citation omit-
ted) (alterations in original). When evaluating the weight of
the releasee’s right, we examine, inter alia: (1) the importance
of the evidence to the district court’s ultimate findings; (2) the
nature of the facts to be proven by the hearsay; and (3)
whether the releasee had any opportunity to refute the evi-
dence. 

   Perez was consistently denied any opportunity to refute the
urinalysis results. Not only was her request to give another
sample on the same day rejected by FRS and the probation
department, but the district court denied her request to cross-
examine the technician. This meant that here, as in Martin,
the releasee had a “virtually complete denial of any opportu-
nity to refute the 

.

   [4] The evidence established the following undisputed
facts: Perez gave a urine sample. Mrs. Schnaible testified that
5730                UNITED STATES v. PEREZ
it was Perez’s urine that was in the sample sent to STLI for
testing, and also that Perez did not have any opportunity to
adulterate the sample. Given the undisputed evidence in this
record that Mrs. Schnaible was watching Perez the entire
time, it cannot be said it is more likely than not that Perez was
the one who diluted the sample. We also know the FRS test-
ing method was unreliable. Mr. Schnaible so testified when he
noted that previously they had sent four positive samples to
the lab, only to have three of them test negative at STLI. The
fact that FRS’s analyses showed cocaine in the sample does
not preclude the possibility that a foreign substance added at
STLI was the source of the cocaine, given that we know
FRS’s tests are unreliable.

   [5] By the time Perez’s sample was tested at STLI in Vir-
ginia, the sample had been diluted with another liquid. We
know this because the urine specific gravity and creatine
levels were low. Under these circumstances, the government’s
offer to re-test the sample did not solve the problem with the
adulteration of the sample. A re-test of the sample might tell
us the foreign substance in the sample and whether it is a sub-
stance commonly found in a laboratory, or it might not. What
a re-test will not tell us is: (1) how the foreign substance got
into the sample; (2) who introduced the foreign substance into
the sample; or (3) whether the foreign substance was the
source of the cocaine metabolites.

   [6] Perez had an obvious interest in cross-examining some-
one from the lab on, inter alia, the results of the report, the
chain of custody of the sample once it arrived at STLI, their
testing methodology, their findings, whether anyone there
could have introduced the foreign substance into the sample,
whether cocaine is stored at the facility, the ratio of the urine
to the other liquid in the sample, the concentration of cocaine
found in the sample, the time cocaine remains in a person’s
system, and the ways in which a sample could become “di-
lute.” Although Perez had an opportunity to cross-examine
the probation officer, Kim, who spoke to someone at STLI,
                     UNITED STATES v. PEREZ                  5731
Kim did not have personal knowledge regarding anything that
happened at the STLI lab in Virginia. The urinalysis and
Kim’s testimony that someone had attempted to dilute the
sample were the only evidence offered to prove Perez’s guilt.
Thus, Perez had a strong interest in being able to confront the
person who actually handled and tested the sample at STLI
and who wrote the report.

   [7] Balanced against the releasee’s interest in cross-
examining a declarant is the government’s interest in not put-
ting the declarant on the stand, including: (1) the expense or
difficulty of getting the witness to the stand; and (2) the reli-
ability of the evidence. 

. Because we
know the evidence was ineluctably unreliable in this case, the
government has a heavy burden to prove it could not have
made the witness available for questioning. The court in Mar-
tin held even if there is a significant expense in bringing the
declarant to testify, the government does not get credit for this
factor if it failed to use an available substitute for live testi-
mony, such as “affidavits, depositions, and documentary evi-
dence.” 

.

   The government argues that because the laboratory techni-
cian was in Virginia and the hearing was in Hawaii, it would
have been quite time-consuming and expensive to transport
the witness to Hawaii for the hearing. Further, the government
maintains the evidence was reliable because STLI was the
national testing laboratory for the U.S. Probation Offices and
U.S. pretrial offices, which shows it “has extensive experi-
ence in this area [and] its reports carry greater indications of
reliability” than other hearsay evidence. 

. All this is true, but the fact remains the sample was
labeled by STLI itself as adulterated. The government could
have offered to allow Perez’s counsel to depose the technician
(including deposing the technician in person, by video-
conferencing, or by telephone), or at least offered a detailed
affidavit explaining the areas of possible cross-examination
5732                    UNITED STATES v. PEREZ
noted above, which would have cost the government very lit-
tle. It did neither.4

   [8] The district court compounded this error by relying on
Perez’s previous violations to disbelieve Perez’s denials of
culpability, rather than on any evidence that Perez committed
the acts alleged here. There was no evidence Perez herself
diluted the sample. The only evidence in this case that relates
to Perez actually makes it less likely she was the one who
diluted the sample. Mrs. Schnaible testified she observed
Perez give the sample and the sample was never out of Mrs.
Schnaible’s sight from the time Perez first gave the sample to
the time it was sealed and mailed off to STLI.

   [9] Further undercutting the district court’s speculation
about what might have happened, Perez immediately
requested that FRS run another test. When FRS did not take
another urine sample, Perez went straight from FRS back to
the probation office, where she requested the probation office
take another sample. We think it unlikely a guilty party would
seek out scientific confirmation of her guilt.

   A term of supervised release cannot be revoked unless
there is actual evidence the person on supervised release com-
mitted the acts alleged in the charging instrument (in this
case, the Request For Course Of Action). See 

(holding a revocation cannot stand on “erroneous
information”); 

(holding a revocation
of supervised release must be “based on verified facts”). Dis-
trict courts must be cautious not to use the releasee’s prior
crimes to find the releasee violated the terms of supervised
release where there is no reliable evidence the releasee did so.
  4
    We do not decide that this substitute for cross-examination would have
satisfied due process. That determination would depend on what the affi-
davit said, who submitted it, why the possibility of a deposition or televi-
sion appearance was rejected, and what opportunities were available to
refute the affidavit. See 

.
                    UNITED STATES v. PEREZ                  5733
   [10] In criminal trials, evidence of other crimes is generally
not admissible to prove action in conformity therewith. See
Fed. R. Evid. 404(b). Evidence of prior convictions for crimes
of dishonesty is admissible to attack the truthfulness of that
witness, Fed. R. Evid. 609(a), but Perez did not testify. Even
though the rules of evidence do not apply to revocation pro-
ceedings, the same logical fallacy exists—you cannot convict
someone of a new offense based solely on their prior bad acts.
It is clear the district court relied solely on prior bad acts by
Perez instead of evidence in this case. For instance, even
though Perez did not testify in this proceeding, at one point
the district court exclaimed, “Her credibility is in the minus
figures [and] If she told me it was daylight, I would go open
the window and check.”

   [11] Finally, the error here could not have been harmless
because the urinalysis was the only actual evidence used to
revoke Perez’s supervised release. See 

n.5. Accordingly, Perez’s interest in cross-examining the lab-
oratory technician so outweighed the government’s interest in
not producing the witness that the admission of the hearsay
evidence violated her Fifth Amendment constitutional right to
due process. See 

.

   The revocation of Perez’s supervised release is reversed
and judgment is rendered in favor of Perez. If Perez has not
already been released, the district court is instructed to vacate
its finding she violated the terms of supervised release, order
her immediate release from any prison term imposed for the
claimed violation of supervised release, and enter an order in
accordance with this opinion.

  REVERSED.