*563OPINION
{1} Defendant Paul Wayne Lovett was charged with murdering two women in two separate, unrelated incidents as well as criminal sexual penetration with respect to one of the victims. Pursuant to Rule 5-203(A) NMRA the two murder charges were joined in one complaint, indictment or information with the intent to try the two murder charges together in one trial. Pursuant to Rule 5-203(C) NMRA Defendant moved to sever the two murder charges into two separate trials. After a hearing, the trial court denied the motion to sever, and Defendant was subsequently convicted of both counts of first-degree murder in one joint trial. We conclude that the trial court committed error when it failed to sever the murder charges into separate trials. Because the error constituted reversible, non-harmless error in relation to one of the murder convictions, we vacate that conviction while upholding the other first-degree murder conviction as well as the conviction for criminal sexual penetration.
BACKGROUND
{2} On January 16, 2002, a young mother named Elizabeth Garcia worked a late shift at a gas station in Hobbs, New Mexico. Under circumstances that are still unknown, Garcia left her workplace sometime before 2:30 in the morning. That afternoon, a local resident found Garcia dead in a vacant field near a dirt road.
{3} Approximately sixteen months later, on May 14, 2003, workers found another young woman, Patty Simon, dead in a caliche pit on the outskirts of Hobbs. On the same day, a delivery man found Defendant, lying on the side of a road outside of Hobbs, shirtless.
{4} The delivery man’s coworker took Defendant into town. Defendant explained to the coworker that he had been talking to a girl named Patty at the bowling alley the night before. According to Defendant, a few guys tried to rough her up and Defendant jumped in. The next thing he remembered was waking up on the side of the road. Defendant said he had a knot on his head, he was scratched, and he seemed shaken, so the coworker asked Defendant if he wanted to seek medical attention, but Defendant refused. After the coworker dropped Defendant off at his apartment, he reported Defendant’s story to the police.
{5} The police had already started investigating Simon’s death. After hearing about Defendant’s story, police officers went to his home, arriving before he had changed clothes or bathed. Defendant was still shirtless and had scratches on his arms and shoulders. He agreed to an interview with the police, and police obtained a search warrant to take certain samples from Defendant, including blood, hair, and a penile swab.
{6} Initially, the police only investigated Defendant in relation to the Simon murder; they did not suspect that he might also be Garcia’s murderer. In fact, when Simon was killed, the police had no suspects in the Garcia murder. Among others, they had investigated Stephen DeMoss, who was Defendant’s brother-in-law at the time Garcia was killed. Police began to investigate Defendant for the Garcia murder only after he began talking about it during an interview concerning the Simon murder.
{7} Eventually, the State charged Defendant with two counts of first-degree murder, one count of kidnaping in relation to the Garcia *564murder, and one count of criminal sexual penetration in relation to the Simon murder. The indictments were joined under Supreme Court Rule 5-203 NMRA .
{8} Defendant moved to sever, and the trial court held a hearing on the matter. After the hearing, the trial court denied Defendant’s motion to sever in a written order, but without explaining the basis for its ruling. The case proceeded to a joint trial, and a death-eligible jury convicted Defendant of two counts of first-degree murder and one count of first-degree criminal sexual penetration.1
{9} Because Defendant received life sentences for each murder, he appeals directly to this Court. See N.M. Const, art. VI, § 2 (“Appeals from a judgment of the [trial] court imposing a sentence of death or life imprisonment shall be taken directly to the supreme court.”); accord Rule 12-102(A)(1) NMRA. On appeal, Defendant argues that his motion for severance was improperly rejected and that the trial court’s failure to sever actually prejudiced his case. He also argues that the fact that a death-eligible jury reviewed a non-death eligible murder prejudiced his case. Holding that the trial court’s failure to sever constituted reversible error, we do not reach Defendant’s second argument.
DISCUSSION
{10} We have previously stated that we review a trial court’s denial of a severance motion for an abuse of discretion. State v. Domzngwez, 2007-NMSC-060, ¶ 10, 142 N.M. 811, 171 P.3d 750. Under certain circumstances like this case, however, the decision to sever may become mandatory, not discretionary, as we explain in this opinion. Although the trial court in this case held a hearing on the matter, and apparently reviewed briefing and case law on the issue, it never explained its decision to deny Defendant’s motion, which makes our task all the more challenging.
{11} “[0]ne test for abuse of discretion [in a failure to sever case] is whether prejudicial testimony, inadmissible in a separate trial, is admitted in a joint trial.” Id. (internal citations and quotation marks omitted). In this case, the State was allowed to introduce in one trial all the evidence of each separate murder plus the criminal sexual penetration. That evidence included background testimony by the victims’ family members and friends, numerous photographs of the respective victims’ crime scenes and mutilated bodies, autopsy reports, bloody clothing samples, testimony by forensic scientists, and testimony by other experts. For better understanding, we first review this evidence in greater detail.2 Then, we consider if the evidence would have been cross-admissible if the State had tried Defendant separately for the two murders. In other words, in a separate trial of the Garcia murder, would the State have been able to present evidence going to the Simon murder, and vice versa? If not, the court committed error in refusing to sever, and we then determine whether the trial court’s error was *565harmless. See generally State v. Tollardo, 2012-NMSC-008, 275 P.3d 110.
Evidence Presented by the State
The Garcia Murder
{12} In addition to the basic information about where and when Garcia’s body was found mentioned above, the State introduced testimony and photographic evidence that Garcia was brutally murdered. T estimony and photographs documented that Garcia’s killer stabbed her fifty-six times and slit her throat. Her killer stabbed her chest, abdomen, back, and hip, and she had other injuries on her hands and arms. Her killer stabbed her with such force that it broke her bones.
{13} Garcia was clothed when she was stabbed and when she was found. Police found tire tracks and footprints at the scene. Blood and scrapes on the ground indicated that Garcia had been dragged from the passenger side of a car.
{14} Several experts testified about the significance of physical evidence found at the Garcia murder scene. One testified that, to a reasonable degree of scientific certainty, Defendant’s semen was found in Garcia’s panties. By contrast, four other suspects, including Defendant’s former brother-in-law, DeMoss, were excluded as semen contributors. The tire tracks were consistent with tires on the car that Defendant drove at the time of Garcia’s murder, and the shoe prints were consistent with the soles of shoes that Defendant owned at that time.
The Simon Murder
{15} In addition to testimony about where Simon was found and Defendant’s predicament that morning — claiming to have been with Simon the night before, the State introduced testimony and photographic evidence documenting the fact that S imon was also brutally murdered. For example, the State introduced testimony and photographs documenting that Simon suffered severe, blunt-force trauma to her head and neck. She had a broken nose and radiating skull fractures. One of her eyes was ruptured, and there were numerous lacerations, bruises, and other injuries to her arms, hands, and the rest of her body, consistent with “defensive wounds.” There was a “large, gaping, incised wound or slash across the upper part of her throat” and obvious injuries to her legs and genitalia. Either the cut to her throat or the blunt-force injuries caused Simon’s death.
{16} Unlike Garcia, police found Simon nude from her bra line down. Her shirt was pulled up over her head. Her legs were spread, and her underwear was around her ankles. Simon’s injuries were consistent with, but not necessarily conclusive of, a sexual penetration. No semen was found at the scene.
{17} Simon’s car was parked at the scene, with a lot of blood found in and around it, including in the trunk. A cigarette butt was found sitting in the.opening to the gas tank, and police collected cigarette butts found on the ground in the area. Only one type of footprint was found near Simon’s body.
{18} Much of the physical evidence tied Defendant to the Simon murder. Investigators found Defendant’s shirt, which was bloody, under a pile of rocks near Simon’s body. DNA on cigarette butts found all around the Simon scene matched Defendant’s DNA. Simon’s blood was on Defendant’s shoes and underwear. The numerous shoe prints around Simon’s body were consistent with the shoes Defendant wore that day. A fiber found on Simon’s hand was consistent with jeans, and Defendant was wearing jeans the day Simon *566was killed. Defendant’s underwear had DNA on it that could not exclude Simon, but could not conclusively be identified as her DNA. DNA found on Defendant’s penis matched Simon with a high likelihood; the likelihood that someone else would match the DNA collected from his penile swab was 1 in 41 Oto 1 in 670.
Defendant’s Prior Statements
{19} The State played a series of videos for the jury that documented interviews between law enforcement officers and Defendant. The interviews initially focused on Defendant’s involvement in the Simon murder. During his second interview with police, however, Defendant began talking about events that had happened around the time Garcia was murdered.
{20} The first interview took place after the delivery man found Defendant on the side of the road, the day police began to investigate the Simon murder. In that interview, D efendant changed his story from what he had told the delivery man’s coworker. Defendant told police that he and Simon had met at her workplace the day before so that he could buy methamphetamine. While they met in her car, a man who Defendant did not recognize, but who Simon knew, approached. The man threatened Simon with a knife and demanded her drugs. Defendant said he tried to hit the man, but that the man slashed Defendant’s shirt, causing the scratch marks on Defendant’s torso.
{21} Defendant said the man made Simon drive her car out of town onto a gravel road. There, the man hit Simon with something, causing her to bleed from the back of her head. Then, the man forced Defendant to drive with Simon in the passenger seat. At some point, the man hit Defendant on the back of the head, and the next thing Defendant remembered, he was in the trunk of the car, without his shoes. Defendant passed out again and woke up on the side of the road with his shoes on again.
{22} Over a month later, in a second interview, police asked Defendant to “tell us what happened . . . from the get go.” Although police were apparently following up on their first interview about the Simon murder, Defendant eventually answered that “[t]he beginning goes back,... uh, to January 15th of . . . what year was it?” January 15th was the day before Elizabeth Garcia disappeared from work and then was found murdered. Defendant said that, on January 15th, his former brother-in-law, DeMoss, had been terrified, afraid that he was going to be framed for a murder. DeMoss needed DNA to “put there.” Defendant said he gave DeMoss a condom with his sperm and several pubic hairs — “there’s always a few loose ones” — in a bag. Defendant told police he thought that DeMoss killed Garcia and that DeMoss shaved his head the next day because people were after him.
{23} After discussing the Garcia murder, police again questioned Defendant about the Simon murder. Defendant began the story as he had during his first interview with the police, but with some additional details: the man who demanded Simon’s drugs had wavy dirty blond hair and “several people, .... in the very beginning told me it was some guy named James.” In this version of the story, however, the man did not knock Defendant out and put him into the trunk. Rather, the man made Defendant get into the trunk. Defendant remembered spending hours in the trunk after the car came to a stop. At some point, after Simon was killed, the man opened the trunk and ordered Defendant to pull Simon from near the car to the spot she was later found. Rather than receiving all his scratches from the man “slashing” him, in this version of *567the story Defendant told police that he inflicted the scratches on his own arms in an attempt to kill himself.
{24} Later during the same, second interview, the police suggested to Defendant thatDeMoss was the assailant who tried to rob Simon and then killed her. Soon after, Defendant changed his story yet again. He maintained much of the initial portion of his story, except that now DeMoss was the assailant. DeMoss hit Simon’s head with something when the car made its first stop.
{25} In this version of the story, Defendant was never trapped in the car’s trunk. Instead, he remained in the car with DeMoss and Simon. When they all arrived at the caliche pit, Defendant said he just walked away, because he did not want to see what was going to happen there. Defendant walked back to the site “late, late, late” and saw Simon, but DeMoss was not there and Defendant did not know how DeMoss had left.
{26} At this point in his interview, the police told Defendant that they thought his DNA would be identified at the scene of the Simon murder and that he should probably be able to explain that. In response, Defendant said he had walked back to the site earlier (as opposed to “late, late, late”) with DeMoss, and gave DeMoss some of his DNA by spitting into a cup. DeMoss then walked over to Simon with the cup. Because DeMoss did this, Defendant said he was sure his spit was somewhere around Simon. Defendant maintained that he did not have sex with Simon and that he did not see anyone else have sex with Simon.
{27} Returning to the subject of the Garcia murder, Defendant said that he was home the night Garcia was killed and that he would not have had the physical strength to commit that crime. He also said DeMoss was angry because DeMoss’s father was interested in Garcia.
{28} Two days later, police interviewed Defendant a third time. During that interview, Defendant’s story changed yet again. He said that after DeMoss initially injured Simon, she walked back to the car herself and got into the back seat. She was bleeding a lot, but she was still conscious and complaining about her head. At the caliche pit, Defendant said that DeMoss and Simon walked away from the car together; and he walked away from them, not wanting anything to do with it. He did not see DeMoss again until DeMoss came over and asked him for spit. Defendant admitted that he was wearing his shoes the whole time and that he had changed his story about where he walked at the scene.
{29} In addition to Defendant’s videotaped statements to the police, a nurse testified that she overheard Defendant say that “they were trying to charge him with two capital murders and they could charge him for ten for all he cared. He only murdered one.” At trial, Defendant did not testify, and thus, he did not explain the nurse’s testimony or the inconsistencies in his own recorded statements.
Cross-Admissibility of Evidence if the Two Murders Had Been Tried Separately
{30} We now address whether the evidence admitted against Defendant for each crime would have been cross-admissible if the State had prosecuted Defendant for the murders separately. Only relevant evidence is admissible. See NMRA 11-402 (2012). Relevant evidence has “a tendency to make a fact more or less probable than it would be without the evidence,” and “is of consequence in determining the action.” NMRA 11-401 (2012).
*568{31} There are important limitations on admitting relevant evidence. As the Court of Appeals explained in State v. Jones, “district courts must be careful in admitting other-bad-acts evidence because of its large potential for prejudice as recognized in the first sentence of [Rule] 11-404(B), which states a general rule of exclusion of such evidence.” State v. Jones, 120 N.M. 185, 187, 899 P.2d 1139, 1141 (Ct. App. 1995). Evidence of other crimes is not admissible to show conformity with those crimes, but may be admitted “for other purposes, such as proving motive, opportunity, intent, preparation, plan, knowledge, identity [or] absence of mistake or lack of accident.” Rule 11-404(B) NMRA. This rule is a product of the “long tradition of courts and commentators expressing fear that jurors are too likely to give undue weight to evidence of a defendant’s prior misconduct and perhaps even to convict the defendant solely because of a belief that the defendant is a bad person.” State v. Ruiz, 2001-NMCA-097, ¶ 13, 131 N.M. 241, 34 P.3d 630 (internal quotations and citations omitted).
{32} “If the evidence is probative of something other than propensity, then we balance the prejudicial effect of the evidence against its probative value.” Id. ¶ 15. (applying Rule 11-403). “[Cjourts must not permit a defendant to be embarrassed in his defense by a multiplicity of charges to be tried before one jury.” State v. Paschall, 74 N.M. 750, 752-53, 398 P.2d 439, 440 (1965).
I den tifying Poten tially Cross-A dm issib le Evidence
{33} In order to apply this framework to the case at hand, we asked the parties to clarify exactly what evidence would or would not have been cross-admissible. Defendant argues that none of the evidence of one crime was admissible in the trial of the other. Specifically, he argues that neither the background testimony, such as the circumstances under which the victims disappeared, the forensic evidence, nor the photographic evidence from one murder would have been admissible in a trial for the other murder.
{34} We agree with Defendant that certain evidence introduced during the joint trial would certainly not have been cross-admissible in separate trials. For example, if the State had tried the Garcia murder separately from the Simon murder, it could have used photographs of Garcia’s body “to illustrate, clarify and corroborate the testimony of witnesses concerning the scene of the crime, wounds of the victim and identity of the deceased.” State v. Noble, 90 N.M. 360, 363, 563 P.2d 1153, 1156 (1977). Those same photographs, however, would certainly not have been admissible in a separate trial of the Simon murder. The same is true of photographs of Simon; they would not have been admissible in a separate trial for the Garcia murder. Those photographs do not tend to show motive, opportunity, or any other relevant fact in relation to the other murder, other than, perhaps, a propensity for violence, which is an impermissible purpose under our rules.
{35} To its credit, the State seems to concede that certain evidence of each crime would not have been cross-admissible. Instead, it cites State v. Gallegos, 2007-NMSC-007, ¶ 45, 141 N.M. 185, 152 P.3d 828, to argue that “[cjomplete cross-admissibility is not required.” The State misunderstands Gallegos, which is a case we authored after the trial court’s hearing on Defendant’s motion to sever, but before Defendant’s trial began.
{36} In Gallegos we held that a trial court abused its discretion when it failed to sever *569charges involving two different victims because evidence would not have been cross-admissible at separate trials. 2007-NMSC-007, ¶ 3. Despite that, we upheld one of the charges because we concluded that the error was harmless, a point we will discuss in more detail later in this opinion. Id. ¶ 44. Thus, contrary to the State’s representation, Gallegos stands for the proposition that a lack of complete cross-admissibility did constitute error, and the charges should have been severed. Ultimately, however, based on the facts of that case, we held that only one conviction should be reversed and the other constituted harmless error.
{37} Because the State does not identify a consequential fact for which much of the bad-act evidence it proffered during trial might have been permissible under the exceptions to Rule 11-404(B), we assume that the general prohibition on propensity evidence applies and that much of the evidence would have been inadmissible in separate trials under Rule 11-404(B). Thus, even ifhypothetically some evidence of the Garcia murder might have been relevant to the Simon murder or vice versa, it was unquestionably error to admit all of the evidence going to both murders in one joint trial.
Applicable Exceptions to Prohibition on Propensity Evidence?
{38} The State also argues that certain evidence was cross admissible for valid purposes other than conformity or propensity. First, because Defendant disputed his identity as the perpetrator of the murders, the State asserts that “any evidence establishing Defendant — as the true perpetrator — was critical and used to refute Defendant’s defense.” The State cites State v. Peters, 1997-NMCA-084, ¶¶ 11-20, 123 N.M. 667, 944 P.2d 896, to support this assertion.
{39} The State misunderstands Peters. Peters is about cross-admissibility of evidence when it suggests that the perpetrator in one crime is the same as the perpetrator in another crime, because the two crimes were committed with a common modus operandi. See id. ¶¶ 14-16, 19-20. Peters does not stand for the proposition that evidence is admissible if it simply suggests that a particular defendant committed multiple crimes. See id.
{40} Peters explains that evidence is cross-admissible to prove identity when it “‘demonstrate[s] a unique or distinct pattern easily attributable to one person.’ Relevancy depends on the degree of similarity.” Id. ¶ 14 (internal citations omitted). Peters was written about a common modus operandi because the two joined crimes were practically the same. Both were armed robberies and rapes of elderly female victims, committed in the victims’ homes at night, using a knife, in the same neighborhood. Id. ¶¶ 3-4, 15. In each crime the perpetrator entered the women’s homes through a window, hit them in the face or head, used his knife to control them, tied their hands and feet, gagged them, covered their faces, raped them, and ejaculated. Id. After each rape, the perpetrator asked each woman where her purse was, and he took money from each purse. Id. The perpetrator had a common modus operandi because he committed the crimes using a “signature,” one recognized not only by police but also by an emergency room nurse who attended to both victims. Id. ¶¶ 15, 20.
{41} No such signature emerges when we compare the Simon and Garcia murders. The killer in the Simon and Garcia murders used different weapons, left the victims in different locations, left entirely different physical evidence at each scene, and inflicted different wounds on the victims. The crimes were *570committed in different places sixteen months apart. The criminal investigators never recognized any pattern in the two crimes until Defendant spoke up and mentioned Garcia in the context of Simon. Unfortunately for the State, the pattern Defendant suggested was that his brother-in-law routinely committed murders and had asked Defendant for DNA to cover his tracks — a pattern unsupported by the evidence; DeMoss was never charged.
{42} The State also argues that evidence presented during the joint trial generally showed that “Defendant harbored the same criminal intent for both murders” and that recurrence of a similar result tends to establish criminal intent. For example, the State points to an expert’s testimony that the perpetrator of both crimes “intended to kill the person and probably had a great deal of emotion.” The State cites to People v. Ewoldt, 7 Cal. 4th 380 (1994), for the argument that this similarity in the evidence tends to establish criminal intent and thus makes evidence cross-admissible.
{43} We disagree. Evidence showing the recurrence of a similar result is just impermissible propensity evidence described using different words. Further, the Ewoldt case from California is not on point. Ewoldt held that certain evidence was cross-admissible because it showed the likelihood that a defendant lacked mistake in committing certain crimes — that a defendant committed a crime intentionally, not accidentally. Whether the Garcia and Simon murders were committed intentionally is not at issue in this case. Defendant does not claim that he accidentally killed Garcia or Simon; he denies killing either of them.
{44} Not only is evidence of a similar result not admissible to show lack of mistake, the fact that a perpetrator had the intent to kill and killed with “a great deal of emotion” is also not sufficient to establish a modus operandi. As Peters recognized, “[i]t is by adding circumstance to circumstance that we obtain a composite feature or mark which as a whole cannot be supposed to be associated with more than a single object.” Id. ¶ 19 (quoting 2 John Henry Wigmore, Evidence § 411 (Chadbourn rev. 1979)). Thus, a similar degree of intent or emotion used to commit two murders is certainly not sufficient to show identity when so many of the characteristics of the Simon and Garcia murders differed.
{45} Next, the State points to Defendant’s statement that he only committed one murder, regardless how many the State charged him with, as admissible in a trial for either murder to support Defendant’s identity and “consciousness of guilt.” According to the testimony, Defendant made this statement after the Simon murder and while he was in custody. Without more information, it is possible that Defendant was referring to the Garcia murder, the Simon murder, or some other murder. Because it is not clear that this evidence relates to one crime over another, it is not evidence showing conformity with a prior bad act. Therefore, we do not consider it in our analysis about cross-admissibility of evidence of one crime in the trial for another.
{46} The State also argues that evidence from each crime was cross-admissible because “the entire case connecting Defendant to the murder of [Garcia] grew out of and originated from the investigation of the death of [Simon].” Specifically, the State cites Defendant’s statement that the “beginning” of the Simon murder started when Garcia was killed. The State argues that this statement is cross-admissible to disprove Defendant’s claims that DeMoss, not Defendant, committed the crimes and that Defendant lacked the opportunity to commit the crimes. The State does not cite any authority for the proposition that, because the investigation for one crime derived from the investigation of *571another, all evidence admissible in one of the trials is also admissible in the other. “We assume[,] where arguments in briefs are unsupported by cited authority, [that] counsel[,] after diligent search, was unable to find any supporting authority.” Matter of Adoption of Doe, 100 N.M. 764, 765, 676 P.2d 1329, 1330 (1984).
{47} In addition, case law from our Court of Appeals suggests that the State’s logic on this point is flawed. In Ruiz, the state charged a defendant with a variety of crimes against three minor girls. 2001-NMCA-097, ¶ 1. The first girl’s statements led police to question the other two girls. Each victim described incidents that happened at different times and under different circumstances, but all accounts involved only the accused, all were sexual in nature, and all took place allegedly in the house of the accused. Id. ¶¶ 2, 4-5. On appeal, the state argued that “the evidence pertaining to each girl was necessary to rebut [defendant’s position that the events never happened or were due to mistaken perceptions on the part of three young girls ....” Id. ¶ 17. The Court of Appeals’ response applies equally to the case before us: “[w]hat the State really wanted was an opportunity to bolster the testimony of the[] three girls.” Id. ¶ 19. There was “no difference between what the [s]tate arguefd] and the use of that same evidence to show propensity: that Defendant did bad things with one girl and it was therefore more likely that he did bad things with the others.” Id. ¶ 18. The convictions in Ruiz were overturned because of the trial court’s failure to sever the charges.
{48} Here, as well, the State wants to enter Defendant’s statements about the Simon murder in the trial for the Garcia murder, and vice versa, to show his conformity with character rather than another, permissible purpose. Defendant’s statements bolster the State’s theory that Defendant has a propensity for lying and committing gruesome murders against women. Conformity with prior bad acts or with character is not a permissible basis to admit Defendant’s statements. See Rule 11-404(A)-(B).
Prejudicial Effect Versus Probative Value?
{49} Now we compare any probative value of the evidence of the Garcia and Simon murders with the prejudicial effect of this evidence. B ecause we have already concluded that evidence of each murder was not offered for any permissible purpose, the inherent prejudicial effect of admitting evidence of prior bad acts necessarily outweighs the probative value of the evidence.
{50} Even more glaringly, the evidence in this particular case had the potential for significant prejudicial effect. The State argues that the evidence presented in Defendant’s trial is distinct from the inflammatory evidence introduced in cases such as Ruiz, which “involv[ed] minor victims and sexual assaults,” because “the evidence was not used to bolster the credibility of the victims or merely to show Defendant’s propensity or bad character.” This argument fails for several reasons.
{51} First, each murder involved a particularly gruesome killing of a young woman, and the State charged Defendant with criminal sexual penetration in relation to the Simon murder. The trial judge even commented to counsel in relation to some of the photographic evidence, “I’m very concerned about how much we can lay on this jury. . . . [T]his is very sensational.” Second, perhaps even more prejudicial to Defendant were Defendant’s statements about the Simon murder. Defendant’s story changed so many times that, in a trial for the Garcia murder, a juror could not help but believe that *572Defendant also lied about his involvement in that case as well. “Once a liar, always a liar” is exactly the kind of propensity evidence rule 404(B) is designed to exclude.
Harmless Error
{52} Because the trial court committed clear error in failing to sever, we proceed to determine whether the error can be regarded, in either trial, as harmless. Since the time the parties submitted briefs to this Court, this Court authored State v. Tollardo, which clarified our harmless-error standard. See 2012-NMSC-008. For nonconstitutional errors, like the violation of evidentiary rules presented in this case, we ask whether there is a reasonable probability that the error affected the jury’s verdict. Id. ¶¶ 35, 42, 43.
{53} The State argues that any error was harmless and that Defendant failed to conduct a harmless-error analysis in his amended brief-in-chief, thus abandoning the issue. While Defendant did not address “harmless error” by name, he did address “actual prejudice.” In the context of failure to sever, we have sometimes called the harmless-error analysis a question of actual prejudice to the accused. See State v. Garcia, 2011-NMSC-003, ¶ 21, 149 N.M. 185, 246 P.3d 1057 (“The failure to sever did not result in any prejudice to Defendant and the trial court's decision not to sever constitutes harmless error.”); State v. Dominguez, 2007-NMSC-060, ¶ 13, 142 N.M. 811, 171 P.3d 750 (“As we have stated, judicial error by itself is not necessarily grounds for reversal in the absence of actual prejudice."). In fact, in Tollardo we recognized that New Mexico courts have sometimes called harmless error “prejudice.” See 2012-NMSC-008, ¶ 25 (“Although not always employing the term ‘harmless error,’ New Mexico courts historically evaluated claims of error by inquiring into how severely the defendant was affected thereby. In State v. Coyle, for example, this Court considered whether or not a trial error worked a ‘prejudice’ against the defendant . . .”). Accordingly, Defendant did not abandon the issue of harmless error.
{54} In determining whether error is harmless, “courts should evaluate all of the circumstances surrounding the error” and should not mechanically apply a multi-factor test. Id. ¶43. That evaluation should include “an examination of the error itself,” including the source and emphasis on the error. See id. ¶¶ 43-44. Evidence of guilt apart from the error can provide context but “can never be the singular focus of the harmless error analysis.” Id. ¶ 43. Courts may also consider the weight placed on the erroneously admitted evidence by the State. Id. ¶ 44. An error may be prejudicial with respect to one conviction, but harmless with respect to another. Id. ¶ 44 (citing Johnson, 2004-NMSC-029, ¶ 31, 136 N.M. 348, 98 P.3d 998). Thus, courts must separately assess the effect the error may have had on each of the convictions.
{55} Applying this analysis, we first consider the error itself, its source, and the emphasis placed upon it by the State. Courts may, depending upon the circumstances of the cases before them, examine “the importance of the [erroneously admitted evidence] in the prosecution’s case,” as well as “whether the [error] was cumulative” or instead introduced new facts.” Id. ¶ 43. In this case, the error was the trial court’s failure to sever two murder prosecutions, each supported by different evidence. Thus, by definition, the error introduced new facts: all the facts related to the Garcia murder were new facts in the case of the Simon murder and all the facts related to the Simon murder were new facts in the case of the Garcia murder.
*573 The Drew Factors
{56} In Gallegos we adopted the “Drew factors” for appellate courts to determine actual prejudice to an accused in the failure-to-sever context. See 2007-NMSC-007, ¶ 42. As mentioned earlier, we published Gallegos after the trial court’s hearing on Defendant’s motion to sever, but before Defendant’s trial ended. We do not know if Gallegos was brought to the trial judge’s attention during trial.
[Fjactors weighing in favor of prejudice include: (1) the prosecution intertwining the offenses in opening statement, during its case-in-chief, or in closing argument; (2) the defendant being found guilty on all counts; (3) factual similarities linking the offenses; (4) offenses that are inflammatory in nature; (5) unusually long and complex trials; and (6) a conviction on a charge where the evidence is thin. On the other hand, factors tending to show that a defendant was not prejudiced by going to trial on the joined offenses include: (1) dissimilar offenses such that a jury would not confuse them; (2) the defendant being acquitted of some charges; and (3) proper jury instructions that adequately make clear to the jury that it must not consider evidence inadmissible to a particular count when coming to a verdict on that count.
Id. ¶ 41 (internal citations omitted). After we applied these factors in Gallegos, we refined our harmless error analysis in Tollardo; the Drew factors, however, are still relevant to our harmless-error inquiry. For example, the first Drew factor, how the prosecution intertwined the cases throughout the trial, may indicate the relative emphasis the State placed on the error in question in this case. As directed in Tollardo, we do not apply the Drew factors mechanically. The factors may have varying degrees of relevancy in different cases.
{57} We consider the Drew factors that are relevant to this case to analyze whether the erroneously admitted evidence from the Garcia murder “probably” affected the jury’s verdict in the Simon murder. Likewise, we analyze whether the erroneously admitted evidence from the Simon murder probably affected the jury’s verdict in the Garcia murder. We begin our analysis by determining whether the State intertwined the facts of the two murders in its opening statement, its case in chief, or its closing statements.
The State’s Opening Statement
{58} The State immediately intertwined the two cases, beginning with the opening statement, by telling the jury “this is actually two cases brought together in one.” Although the State specified which charges were related to each victim, it went on to say that Defendant “took great steps to conceal his involvement in these crimes,” and that the theme of the whole trial was “there is nothing that is hidden which will not be revealed.” Thus, by portraying the crimes as having a common theme, the State demonstrated its reliance on the jury’s perception that the crimes were related and that Defendant acted similarly in each. The State then introduced each of the victims by photograph and biographical information in succession, first Garcia, then Simon.
{59} Rather than describing the evidence supporting a conviction for each murder separately, the State generally told the story of the two murders through the eyes of the *574investigators. First, the prosecutor described the day Garcia was killed and the initial investigation into her death. Then, after telling the jury that police were unable to determine Garcia’s murderer, or even match the semen found in her panties with a known suspect, the State “fast forward[ed]” sixteen months, to Simon’s murder. The State remarked that Simon, “too, suffered a very, very violent death” and that “significant amounts of blood were found over a broad area of this crime scene, as well.” Simon’s “throat, too, had been slit.”
{60} The State described Defendant’s apprehension by the police the day that workers found him on the side of the road and his initial explanation of his predicament. Then, notably, the prosecutor suggested that police decided to interview Defendantbecause they thought he might have had something to do with the Garcia murder in addition to the Simon murder, a suggestion that is demonstrably false. In the same statement, the prosecutor confused the names of the victims. The prosecutor told the jury that police did not have any leads in the Garcia murder, so “when [the police] found out that Elizabeth [Garcia] had been — I’m sorry — that Patty [Simon] had been killed, that she was left also on the outskirts of town, that her body was semi-nude, that there were some similarities between the two — the two crimes, they decided, ‘Why not? Let’s talk to [Defendant].”’
{61} The evidence, however, was that, over a month after Simon was killed, a police officer from the Hobbs Police Department who had been working on the Garcia murder became interested in interviewing Defendant, for vague reasons: he made contact with the Lea County Sheriffs Department, he had heard of the death of Simon, and he wanted to help out with the Simon murder investigation because “we had two young mothers that had been murdered.”
{62} The State then alternated the substance of its opening argument between discussion of the two murders. First it described additional evidence that the jury would hear, linking Defendant to the Garcia murder. After describing the investigation that linked the footprints and tire tracks found at the Garcia crime scene with Defendant, the State turned back to the footprints associated with the Simon murder and the ways in which investigators linked footprints from the Simon crime scene to Defendant.
{63} After describing more about the Simon murder crime scene, again the State told the jury more about what its theory of the case in relation to the Garcia murder. Finally, the State gave its theory of the case in relation to the Simon case.
The State’s Case in Chief
{64} The State also organized its case in chief from the perspective of an investigator, in general revealing information about the Garcia and Simon murders as the evidence was discovered by police, chronologically. Thus, rather than first admitting all the evidence relating to the Garcia murder, concluding that evidence, and then moving on to evidence pertaining to the Simon murder, the State interspersed the evidence supporting a conviction for each murder much in the same manner it did during its opening statement. First it presented background information about each crime. From evidence relating to Garcia’s disappearance and the scene where she was found, the State moved on to testimony by Defendant’s wife to whom Defendant was married when Simon was killed, testimony by witnesses who found Simon’s body, and by the workers who found *575Defendant on the side of the road after Simon’s murder.
{65} The State even separated the testimony of some witnesses into multiple parts, not necessarily based on the crime about which they were testifying. Defendant obj ected to this practice, arguing that recalling witnesses simply provided dramatic effect. The trial court permitted the State’s format for introducing evidence, despite recognizing its confusing nature:
the case is made very complicated by the fact that there are two separate murder counts being considered and there’s this unusual timing with the — some 16, 18 months between and then how the case has been developed by the State, such that I’m going to grant latitude to the State in presenting this evidence to try to make it in a more presentable but not necessarily a more dramatic fashion.
{66} Defendant also objected to the State’s plan to present all of Defendant’s recorded interviews with police at the same time. Although the police interviews were primarily about the Simon murder, one included a significant discussion of the Garcia murder. The State argued that Defendant’s statements to the police should be presented at the same time because “they’re intricate, they ’re interwoven, they relate to one another ... since these statements are so interrelated, since they contradict each other so much . . . the jury should have all this evidence presented to them at the same time . . . .” The State’s desire to present all the statements at once because they were “interwoven,” relating to both the Simon murder and Garcia murder, illustrates the State’s emphasis on the fact that Defendant faced charges for two murders, tried together.
{67} After playing the video recordings to the jury of Defendant’s statements to the police that tied him to both crimes, the State essentially alternated between presenting witness testimony relating to the Garcia murder and evidence relating to the Simon murder. Testimony about the Simon autopsy was followed by testimony about the footwear worn in the Garcia murder. Then, testimony about the tire tracks at the Garcia murder scene was followed by testimony about the shoes worn at the Simon murder scene. Then, expert testimony about the Garcia autopsy included testimony about the likelihood that larvae may break down the components of semen (a fact only relevant to the Simon murder). Then, testimony about the DNA on cigarette butts found at the Simon murder scene, which was followed by testimony that Defendant’s semen was found in Garcia’s panties, as well as testimony about where investigators found Simon’s blood, and testimony about whether investigators found Simon’s DNA on Defendant’s penile swab.
{68} In addition to the order in which the State chose to offer evidence, the State’s case in chief intertwined the cases at times by confusing them. For example, during one direct examination the State accidently asked a witness about the site where Garcia’s body was found (Continental Road), when it meant to ask about the site where Simon’s body was found (Shell Road).
The State’s Closing Statement
{69} During closing argument, the State began by explaining the jury instructions for the Garcia and Simon murders. Right away there was some confusion on the State’s part as to which instructions applied to which crimes:
I think I may have gotten it backwards here. One of them is *576going to be with respect to Patty Simon, and the other one will be with respect to Elizabeth Garcia. Okay? I may have a different instruction up here; but you’re going to have an instruction with you that instead of Patty Simon ’ here, it says ‘Elizabeth Garcia. ’ Okay? Same elements as the first crime.”
{70} Like it did during its opening statement, the State suggested that police connected the two murders before they even had their first recorded interview with Defendant. After discussing the lack of a suspect in the Garcia case at the time Simon was murdered, the prosecutor said, “[One of the detectives investigating the Garcia murder] thinks to himself, ‘You know what, I need to check [Defendant] out. We have another young woman dead in a remote area of Hobbs 16 months after — I need to go talk to this guy.’” The mentioned detective testified during trial that he decided to interview Defendant over a month after the Simon murder, in order to help with the Simon murder investigation, not because of any relation to the Garcia murder.
{71} In addition to this misrepresentation, two other aspects of the State’s closing argument are noteworthy. The State suggested that Defendant’s claim that his brother-in-law, DeMoss, committed the Garcia murder and then asked Defendant for DNA to cover up the crime was “laughable; laughable, especially when you look at the next case.” Thus, the State relied upon the jury’s knowledge of the Simon case in order to discredit Defendant in the Garcia case. Later, after replaying one of Defendant’s statements to the police, the State suggested thatbecause Defendant was a liar in the Simon murder, he also lied in the Garcia case. For example, the State pointed out Defendant’s explanation as to how his DNA was found in Garcia’spanties was the same as his explanation as to why police might also identify his DNA at the Simon murder scene:
So, once again, we hear the same story. [DeMoss] asked [Defendant] for DNA, and [Defendant] gave it to him. Fool me once, shame on you. Fool me twice, shame on me.
What do you think about that? What do you think about the first time [Defendant] says, “Well, yeah. [DeMoss] came — I let him borrow my DNA. That time I gave him some sperm and some pubic hairs, but he was just — he just told me that someone was going to get killed. This time I know [DeMoss] actually killed somebody, and then [DeMoss] came to me and said T need some DNA.’ All right. I can just give you spit this time.” Think about that, ladies and gentleman. Think about it in the context of all the lies he’s already told, lie after lie after lie after lie.
Each argumentagainstDefendant’s credibility relied on the joinder of the two cases, that the State tried Defendant for both murders at once.
{72} It is fairly obvious that the State’s opening, closing and case-in-chief placed significant emphasis on the fact of joinder, that Defendant faced charges for both the Simon murder and Garcia murder in the same trial. By using evidence of Defendant’s actions and statements in relation to the Simon murder to prove that he also committed the Garcia murder, the State relied heavily on the trial court’s erroneous failure to sever in order to prove its case.
*577 Other Drew Factors
{73} Beyond the State’s significant intertwining of the two cases, several other Drew factors are relevant in this case and suggest actual prejudice to Defendant. For example, the jury found Defendant guilty of all three counts against him. Although the State counters that the trial judge granted a directed verdict on the kidnaping charge, that point is irrelevant. The Drew factors direct us to consider convictions and acquittals (not directed verdicts) because they are intended to evaluate whether the State’s presentation of evidence to a jury caused juror bias. A trial judge’s realization that the State has not provided enough evidence to sustain a conviction on a particular charge does not provide the same insight about a jury’s likely bias.
{74} Another Drew factor suggesting actual prejudice to Defendant is that the State relied on the factual similarities of the offenses during its arguments. For example, in addition to the other instances mentioned during closing argument, the prosecutor argued that
[tjhere is no question in these two cases these women were deliberately murdered. The amount and the type of injuries that they suffered can lead to no other conclusion at all.
Despite exploiting joinder in this manner, the State now argues that “the jury was able to consider and compartmentalize the evidence with respect to each count,” this despite the State’s efforts to blur those distinctions throughout the trial. The support the State provides for this sweeping assertion is that “different witnesses testified [concerning] different offenses, and the issues were relatively simple.” While many witnesses only testified about one of the murders, our analysis of the State’s presentation of the evidence makes clear that the State went out of its way to question certain witnesses about both murders and to intertwine the testimony about each murder.
{75} It is not clear what the State means by the “issues were relatively simple.” The trial was for two death-penalty-eligible murders. It involved two unrelated victims, different defenses, dozens of witnesses, and a significant amount of expert testimony. The fact that the State was confused at times, using the wrong victim’s name, indicating the wrong jury instruction, and confusing the locations where the victims were found, suggests that the jury was also confused at times. The trial lasted a full two weeks, not including jury selection or sentencing. For all these reasons, we disagree that this was a simple case. If the issues in this trial were simple, it would be difficult to identify a complex case.
{76} Another Drew factor suggesting actual prejudice to Defendant is that the offenses were inflammatory in nature. See, e.g., Bracy v. Schomig, 286 F.3d 406, 416 (7th Cir. 2002) (“The evidence of the Arizona murders was admitted and it was inflammatory. It was the story of a nasty home invasion resulting in the brutal murder of two people.” (emphasis added)). Both murders were particularly violent and the evidence included gruesome photographs of mutiiated women. One of the murders involved genital mutilation, as well as rape. This was not a case about bad checks or trespass.
{77} The State argues that the last Drew factor is present, that “the jury instructions were adequate,” weighing against a conclusion that Defendant was actually prejudiced. In support, the State points out that the jury instructions directed the jury to consider each count separately and there were no juror notes *578indicating juror confusion. “The risk of prejudice can be offset by the defendant being acquitted of some charges and ‘proper jury instructions that adequately make clear to the jury that it must not consider evidence inadmissible to a particular count when coming to a verdict on that count.’” Gallegos, 2007-NMSC-007, ¶ 41. In State v. Paiz, we also noted that such an instruction could have guarded against juror confusion. See 2011-NMSC-008, ¶ 25, 149 N.M. 412, 249 P.3d 1235. In this case, however, we fail to understand how adequate jury instructions could cure the likelihood of juror confusion and bias. During the two week trial, the State relied particularly heavily on the intertwining of the evidence supporting the two murder charges.
Evidence of Defendant’s Guilt, Separate from the Error
{78} As directed by Tollardo, we also consider the evidence of Defendant’s guilt, separate from the error, in our harmless error analysis. 2012-NMSC-008, ¶ 43. Thus, we consider the evidence against Defendant in the Garcia murder without evidence of the Simon murder. Then, we also consider the evidence against Defendant in the Simon murder without evidence of the Garcia murder.
{79} As mentioned earlier, police did not suspect Defendant in the Garcia murder for over sixteen months. Only when Defendant suggested to police that his brother-in-law had planted his sperm at the Garcia crime scene, did police compare the sperm found in Garcia’s panties with Defendant’s DNA. Before that the police had no match. The match connected Defendant to Garcia, but not necessarily at the time of her death and not necessarily in a violent encounter.
{80} Defendant had an alibi in the Garcia case. In his recorded statement, Defendant said he was home the night Garcia was killed and that he went to physical therapy the next morning. Defendant’s physical therapist confirmed Defendant’s attendance at 7:45 that morning and said that he remembered nothing unusual about Defendant during the therapy session that lasted forty-five minutes to an hour. Defendant’s ex-wife confirmed Defendant’s presence at home the night Garcia was killed. Defendant’s wife also testified, however, that she did not know about other things that Defendant was doing around that time. Thus, evidence corroborated Defendant’s claim of an alibi to a certain extent, although it was also not categorically confirmed. Significantly, however, without the evidence of the Simon murder, a jury hearing just the evidence about the Garcia murder would not know that Defendant repeatedly lied about his involvement in another murder.
{81} Unlike the Simon murder, no evidence at the Garcia murder scene connected Defendant, specifically, with Garcia’s death. Although Defendant owned shoes and tires at the time of the crime that were consistent with the shoe and tire prints found near Garcia’s body, analysts could not compare unique characteristics between the impressions at the crime scene and Defendant’s shoes and tires because Defendant no longer owned the shoes or tires.
{82} By contrast, Defendant was undoubtably present when Simon was killed. Not only did he consistently admit to being present, the physical evidence supported a conclusion that he was involved in her death. The State’s experts testified that Simon’s blood was on Defendant’s shoes, shirt, and pants. Defendant’s underwear had DNA on it that could not exclude Simon as a contributor, and Defendant’s penis had DNA on it that was likely Simon’s. Physical evidence at that crime scene pointed to Defendant: shoe *579prints, cigarette butts, Defendant’s shirt hidden in a pile of rocks, and tracks leading from the crime scene toward the place Defendant was later found.
{83} The evidence against Defendant in the Simon case was strong in several respects. There was significant evidence that Defendant lied repeatedly about his involvement in the Simon murder. The recorded interviews demonstrated that Defendant varied his stories at the suggestion of the police. A majority of the stories had to be lies. In fact, none of Defendant’s stories about his involvement in the Simon murder were consistent with the physical evidence presented.
{84} Considering our analysis oftheDrew factors and the evidence of Defendant’s guilt in each these crimes, we conclude that the trial court’s failure to sever caused Defendant actual prejudice in relation to his conviction for the Garcia murder. While there was certainly sufficient evidence to convict Defendant in the Garcia murder, any reasonable person would conclude, as a reasonable probability, that trying the two murders together made the Garcia case stronger and Defendant’s conviction more likely. Thus, the trial court committed reversible error — not harmless error — by failing to sever the two prosecutions, at least with respect to the Garcia prosecution.
{85} We cannot conclude, however, that trying the Garcia case alongside the Simon case affected the jury verdict in the Simon case, at least not to a reasonable probability. It seems more likely that Defendant stood the same high probability of conviction for the Simon murder with or without the evidence relating to the Garcia murder. Thus, we only vacate Defendant’s conviction for first-degree murder in relation to Garcia. See Gallegos, 2007-NMSC-007, ¶ 47 (concluding that a trial court’s failure to sever two joined crimes only actually prejudiced a defendant in relation to one of the crimes, and thus only reversing the conviction for which the defendant suffered actual prejudice).
CONCLUSION
{86} For the forgoing reasons, we reverse Defendant’s conviction for first-degree murder in relation to Garcia and uphold his conviction for first-degree murder and criminal sexual penetration in relation to Simon.
{87} IT IS SO ORDERED.
RICHARD C. BOSSON, Justice
WE CONCUR:
PETRA JIMENEZ MAES, Chief Justice
PATRICIO M. SERNA, Justice
EDWARD L. CHÁVEZ, Justice
BARBARA J. VIGIL, Judge