OPINION OF THE COURT
RENDELL, Circuit Judge.
Shawn Langford appeals the sentence imposed after he pled guilty to bank robbery in violation of 18 U.S.C. § 2113(a), armed bank robbery in violation of 18 U.S.C. § 2113(d), and carrying and brandishing a firearm during a crime of violence in violation of 18 U.S.C. § 924(c)(1)(A)(ii). Langford was sentenced to 46 months’ imprisonment for Counts One and Two, to be served concurrently, and a mandatory term of 84 months’ imprisonment for Count Three, to be served consecutively, for a total of 130 months’ imprisonment. Langford argues that the District Court improperly calculated his criminal history score and consequently chose an erroneous Sentencing Guidelines range as the first step in the sentencing process, and that he should therefore be resentenced. The government urges that the error is harmless because the applicable Guidelines range overlaps with the correct range. The application of the harmless error standard to a sentence in this fact setting is an issue of first impression in our Court. We will join our sister courts of appeals who have decided this issue and hold that the error is not harmless. We will accordingly vacate Langford’s sentence and remand to the District Court for resentencing.
I. FACTS AND PROCEDURAL HISTORY
On March 9, 2005, Langford and his uncle, Charles Collier, a career criminal, robbed the Iron and Glass Bank in Scott Township, Pennsylvania. PSR ¶ 5-6. Armed with a pistol, Langford remained in the lobby while Collier vaulted the teller counter, announced the robbery, and emptied the money from the teller drawers. PSR ¶ 7. Both men fled, were captured by [208]*208the police following a chase, and confessed to the bank robbery.
Langford was indicted for bank robbery in violation of 18 U.S.C. § 2113(a), armed bank robbery in violation of 18 U.S.C. § 2113(d), and carrying and brandishing a firearm during a crime of violence in violation of 18 U.S.C. § 924(c)(1)(A)(ii). He pled guilty to the charges on December 20, 2005. App. 85.
Langford was eighteen years old at the time of the bank robbery and this was his first adult offense. App. 34. Langford does, however, have a history of adjudications as a juvenile. In the Presentence Investigation Report (“PSR”), the probation officer concluded that consideration of three of Langford’s prior adjudications of juvenile delinquency resulted in five criminal history points.
The two adjudications not at issue on appeal were as follows:
1. In 2001, at age fourteen, Langford was adjudicated delinquent for criminal conspiracy, possession, and possession with intent to deliver crack cocaine and ordered committed to a Community Intensive Supervision Program (and subsequently a detention center for violating the terms of his program). PSR ¶ 31, App. 126.
2. In 2003, at age sixteen, he was adjudicated delinquent for robbery, criminal conspiracy, and fleeing the police. PSR ¶ 32.
While the charges were pending for his second offense, Langford was released with electronic home monitoring. He failed to appear for arraignment and a warrant was issued. On September 29, 2003, he was apprehended by the police as he attempted to steal a vehicle. PSR ¶ 33. That same day, a petition for his second adjudication was filed in juvenile court charging the defendant with the previous robbery charge. PSR ¶ 32. As to the robbery charge, he was adjudicated delinquent on October 28, 2003 and committed to a youth development center. PSR ¶ 32. On October 28, 2003, he was also adjudicated delinquent as to criminal attempt (auto theft), possession of instruments of crime, resisting arrest, criminal mischief, and disorderly conduct. The court discontinued that third adjudication and, according to the PSR, ordered the defendant to provide a DNA sample. PSR ¶ 33. This third adjudication (the “auto theft adjudication”) is at issue on appeal.
Because the 2005 bank robbery occurred less than two years after Langford’s release from juvenile commitment to a youth development center, the probation officer added an additional criminal history two points, establishing a criminal history category of IV. PSR ¶ 33.
At sentencing, Langford’s counsel argued that the appropriate criminal history category was III, rather than IV, because the last adjudication did not result in a sentence and, accordingly, no point should have been added. App. 116. The Court disagreed and calculated a criminal history category of IV which, when combined with a total offense level of 19, resulted in a Sentencing Guidelines range of 46 to 57 months’ incarceration for Counts One and Two. App. 123. Langford was sentenced to 46 months for Counts One and Two, to be served concurrently. For Count Three, carrying and brandishing a firearm during a crime of violence, Langford was subject to a mandatory minimum term of 84 months’ imprisonment to be served consecutively to the bank robbery charges. Langford’s total sentence, therefore, was 130 months’ imprisonment. Langford does not appeal his sentence for Count Three, but rather contends that the District Court miscalculated the Sentencing Guidelines range for Counts One and Two, thus resulting in a longer overall sentence.
[209]*209Langford timely appealed. We have jurisdiction under 28 U.S.C. § 1291 and 18 U.S.C. § 3742(a). See United States v. Lloyd, 469 F.3d 319, 321 (3d Cir.2006). We review the District Court’s interpretation of the Sentencing Guidelines de novo. United States v. Pojilenko, 416 F.3d 243, 246 (3d Cir.2005).
II. CALCULATION OF LANGFORD’S CRIMINAL HISTORY CATEGORY
Langford argues that the District Court should not have imposed a criminal history point for his juvenile adjudication for attempted auto theft because it resulted in a “discontinuance” of the delinquency petition. He maintains that a discontinuance is not a “sentence” within the meaning of U.S.S.G. § 4A1.2(a). Because no “sentence” was imposed, the adjudication should not have been counted, and the proper criminal history category was III, not IV. Accordingly, the Sentencing Guidelines range should have been 37 to 46 months.
U.S.S.G. § 4A1.2(a)(1) defines a prior sentence as “any sentence previously imposed upon adjudication of guilt, whether by guilty plea, trial, or plea of nolo contendere, for conduct not part of the instant offense.” To be sure, as the government argues, juvenile adjudications are not exempted from the calculation of a defendant’s criminal history score. United States v. Bucaro, 898 F.2d 368, 373 (3d Cir.1990). Not all juvenile adjudications, however, result in criminal history points, largely because “[attempting to count every juvenile adjudication would have the potential for creating large disparities due to the differential availability of records.” U.S.S.G. § 4A1.2, cmt. n. 7. Therefore, for offenses committed prior to age eighteen, the Guidelines comments limit criminal history points to “those that resulted in adult sentences of imprisonment exceeding one year and one month, or resulted in imposition of an adult or juvenile sentence or release from confinement on that sentence within five years of the defendant’s commencement of the instant offense.” Id. (emphasis added). Because the other categories of juvenile adjudications that are to be counted are not relevant here, “the imposition of an adult or juvenile sentence” is a prerequisite to using Langford’s auto theft adjudication in the calculation of his criminal history score.
The key question is whether the discontinuance of Langford’s auto theft adjudication constitutes a “sentence” under the Guidelines. We must review the operations of the Pennsylvania juvenile system to determine whether a “discontinuance” falls within the definition of a prior “sentence” set forth in U.S.S.G. § 4A1.2. See United States v. McKoy, 452 F.3d 234, 238-40 (3d Cir.2006) (noting that while we use federal, rather than state, law definitions and terminology, the state proceeding and statutory scheme is relevant to determining whether the adjudication actually resulted in a sentence as defined by the Guideline and its commentary); accord U.S. v. Morgan, 390 F.3d 1072, 1074 (8th Cir.2004).
In Pennsylvania, a juvenile delinquency adjudication requires a court to find beyond a reasonable doubt that a child committed acts that would constitute crimes if committed by an adult. See 42 Pa.C.S.A. § 6341(b). Typically, the court then orders a disposition. Completed with the aid of a comprehensive social study and investigation, a disposition may operate as the functional equivalent of an adult sentence. See id. § 6339. However, “[i]f the court finds that the child is not in need of treatment, supervision or rehabilitation it shall dismiss the proceeding and discharge the [210]*210child from any detention or other restriction theretofore ordered.” See id. § 6341(b). Here, the juvenile court adjudicated Langford delinquent in connection with the attempted auto theft offense. Thereafter, however, it ordered that the petition be discontinued.
In the PSR for Langford’s sentencing, the probation officer added one criminal history point for this adjudication in reliance on U.S.S.G. § 4A1.2(a)(3), which provides that “[a] conviction for which the imposition or execution of sentence was totally suspended or stayed shall be counted as a prior sentence under § 4A1.1(c).” Langford objected on the ground that no sentence had actually been imposed. In response, the probation officer admitted that the adjudication had resulted in the “imposition of no sentence,” yet reached the improbable conclusion that it could be counted as a prior sentence because “[t]he imposition of no sentence, as occurred here, is akin to a ‘suspended’ sentence.” PSR 2d Addendum. On appeal, the government asks us to use this analogy to find that the discontinuance was a sentence.1
If in fact the juvenile court had imposed probation or a suspended sentence, our review would end here. The juvenile court, however, did not impose a sentence and then suspend its operation; it discontinued the action. Refusing to impose a sentence is not the same as suspending a sentence. Black’s Law Dictionary 1446 (6th ed.1992) (defining a suspended sentence as “[a] conviction of a crime followed by a sentence that is given formally, but not actually served”). Even if we understood the juvenile court to have continued the disposition hearings under 42 Pa.C.S. § 6341(e) (which we do not), we would not count a continuance as a “sentence” under the Guidelines as this is, in essence, a juvenile diversion. See U.S.S.G. § 4A1.2(f) (noting that “[a] diversionary disposition resulting from a finding or admission of guilt, or a plea of nolo contendere ” from a juvenile court is not considered a sentence under § 4A1.1(c)); United States v. McKoy, 452 F.3d 234, 238 (3d Cir.2006).
The government refers us to no authority from which we can find that a discontinuance by a juvenile court constitutes a sentence. Instead, it cites numerous eases where courts reached the unremarkable conclusion that a suspended or probationary sentence constitutes a sentence under U.S.S.G. § 4A1.2(a). See, e.g., United States v. Holland, 195 F.3d 415 (8th Cir.1999) (examining a suspended sentence imposed by juvenile court); United States v. Holland, 26 F.3d 26 (5th Cir.1994) (same).
Lastly, the government urges us to assume that the juvenile court discontinued Langford’s attempted auto theft adjudication only because Langford had already been committed to a juvenile institution on a separate offense, so that “there was simply no reason for the juvenile court to impose such a sentence once again a few weeks later.” Appellee’s Br. at 17. We decline the government’s invitation to engage in conjecture. Even if we were able to ascertain the juvenile court’s motives, they are irrelevant; the essential fact here is that the court discontinued the petition and did not impose a sentence.
In light of the foregoing, we hold that the discontinuance of the juvenile adjudication here is not a sentence for the purposes of U.S.S.G. § 4A1.2(a) and should not have been used in the calculation of the [211]*211defendant’s criminal history under the Guidelines.2
Since the discontinuance of a juvenile adjudication cannot be considered a sentence for the purpose of U.S.S.G. § 4A1.2(a), adding a point to Langford’s criminal history on the basis of this adjudication was error. This error, in turn, affected the calculation of the overall criminal history category — moving it from category III to IV — and the subsequent Guidelines range calculation — changing it from a range of 37 to 46 months for Counts One and Two to a range of 46 to 57 months.
III. THE EFFECT OF AN INCORRECT GUIDELINES CALCULATION
Although the Guidelines are now advisory and a sentencing court has great discretion over the substance of the sentence, the correct calculation of the applicable Guidelines range remains an important procedural requirement. First of all, as before Booker, the sentencing court is required to calculate the Guidelines range in each case, and that calculation is the focus of the parties’ arguments. Second, a district court is required to consider the Guidelines range, pursuant to § 3553(a)(4), and use that range as a starting point for the entirety of the § 3553(a) analysis. Based on its consideration of the § 3553(a) factors, the Court must state the reasons for its sentence and explain whether a within-Guidelines sentence is appropriate in the particular case, a process which generally will require a correct Guidelines calculation. Third and finally, a correctly calculated Guidelines range will often be a necessary precondition of our reasonableness review. Where a district court begins with an erroneous range, it will be difficult for us to determine that it fulfilled its duty to consider the Guidelines and reason through to the ultimate sentence. We will discuss these considerations in turn.
A. Duty to calculate the Guidelines range in each individual case
In rendering the Guidelines advisory, the Supreme Court made clear that sentencing courts are required to “consider” the Guidelines in crafting a sentence. United States v. Booker, 543 U.S. 220, 245-46, 125 S.Ct. 738, 160 L.Ed.2d 621 (2005). Our Court thereafter provided district courts with a three-step process to follow in order to comply with the Supreme Court’s ruling in Booker:
(1) Courts must continue to calculate a defendant’s Guidelines sentence precisely as they would have before Booker.
(2) In doing so, they must formally rule on the motions of both parties and state on the record whether they are granting a departure and how that departure affects the Guidelines calculation, and take into account our Circuit’s pre-Booker case law, which continues to have advisory force.
(3) Finally, they are to exercise their discretion by considering the relevant § 3553(a) factors in setting the sentence they impose regardless of whether it varies from the sentence calculated under the Guidelines.
[212]*212United States v. Gunter, 462 F.3d 237, 247 (3d Cir.2006) (quotation marks, brackets, and citations omitted); see also United States v. Hawk Wing, 433 F.3d 622, 631 (8th Cir.2006) (stating that courts should calculate Guidelines ranges just as they would have before Booker); United States v. Crosby, 397 F.3d 103, 112 (2d Cir.2005) (“The applicable Guidelines range is normally to be determined in the same manner as before Booker/Fanfan.”). When a sentencing court miscalculates the applicable range, it fails to discharge its duties under step one of Gunter. As we made clear in United States v. Jackson, “because the Guidelines still play an integral role in criminal sentencing, we require that the entirety of the Guidelines calculation be done correctly.” 467 F.3d 834, 838 n. 4 (3d Cir.2006) (citations omitted).
B. The “starting point” of a district court’s § 3553(a) analysis
The correct Guidelines calculation is not merely one of three steps, but rather constitutes the “natural starting point” from which the sentencing court exercises its discretion under § 3553(a) at Gunter's third step. United States v. Cooper, 437 F.3d 324, 331 (3d Cir.2006). As the Supreme Court recently confirmed in Gall v. United States, “a district court should begin all sentencing proceedings by correctly calculating the applicable Guidelines range. As a matter of administration and to secure nationwide consistency, the Guidelines should be the starting point and the initial benchmark.” — U.S. -, 128 S.Ct. 586, 596, 169 L.Ed.2d 445 (2007). The Court further observed that “[t]he fact that § 3553(a) explicitly directs sentencing courts to consider the Guidelines supports the premise that district courts must begin their analysis with the Guidelines and remain cognizant of them throughout the sentencing process.” Id. at 597 n. 6. The failure to correctly apply the Guidelines was specifically listed by the Supreme Court in Gall as a “significant procedural error.” Id. at 597. A correct calculation, therefore, is crucial to the sentencing process and result.
An erroneous calculation of the Guidelines will frustrate the sentencing court’s ability to give meaningful consideration to “the kinds of sentence and the sentencing range established for ... the applicable category of offense committed by the applicable category of defendant as set forth in the guidelines .... ” as required by 18 U.S.C. § 3553(a)(4). The Supreme Court recently noted that “[a] district judge must include the Guidelines range in the array of factors warranting consideration.” Kimbrough v. United States, — U.S. -, 128 S.Ct. 558, 564, 169 L.Ed.2d 481 (2007) (emphasis added). As we have observed, “[bjecause the Guidelines reflect the collected wisdom of various institutions, they deserve careful consideration in each case. Because they have been produced at Congress’s direction, they cannot be ignored.” United States v. Goff, 501 F.3d 250, 257 (3d Cir.2007); accord United States v. Lalonde, 509 F.3d 750, 763 & n. 5 (6th Cir.2007) (noting that after Gall, a court of appeals “cannot find that a sentencing court has properly considered the § 3553(a) factors if it miscalculated the advisory Guidelines range”).
The proper Guidelines benchmarks — offense level, criminal history, enhancements, and ultimate range — are necessary prerequisites to a court’s analysis under § 3553(a)(4) in general and, more specifically, § 3553(a)(6) (concerning disparity between defendants) and § 3553(a)(5) (having to do with the Sentencing Commission commentary). For example, where a court miscalculates a defendant’s criminal history, its attempts to avoid disparity between defendants pursuant to § 3553(a)(6) will be misguided as it ineluctably will [213]*213compare the defendant to others who have committed the same offense but are in a different criminal history category. Similarly, if a sentencing court incorrectly decides that a reckless endangerment adjustment under U.S.S.G. § 3C1.2 applies, it may rely on inapplicable Sentencing Commission comments as it evaluates the § 3553(a) factors.
Moreover, a sentencing court’s exercise of its discretion to impose a sentence outside the Guidelines range or to determine that “a within-Guidelines sentence is ‘greater than necessary’ to serve the objectives of sentencing,” Kimbrough, 128 S.Ct. at 564, will necessarily be skewed when it misperceives the applicable range. Without knowing the correct range, a district court may impose an outside-the-Guidelines sentence without providing adequate explanation or, alternately, may impose a sentence believed to be at one end of the range or below the range, but that actually falls within the correct range.
Imposing a sentence outside the correctly calculated Guidelines range without explanation would fly in the face of the Supreme Court’s and our precedent. As the Supreme Court noted in Rita v. United States, § 3553(c) calls for a sentencing judge “at the time of sentencing” to “state in open court the reasons for its imposition of the particular sentence.” — U.S. -, 127 S.Ct. 2456, 2468, 168 L.Ed.2d 203 (2007). Without a correct Guidelines range, a sentencing court will fail to comply with the Supreme Court’s holding that a sentencing court must properly justify a sentence based on the record and Guidelines calculation before it. Id. at 2465-68 (reiterating the importance of the sentencing court’s subjecting the sentence to thorough adversarial testing). As we said in United States v. Fisher,
[i]f, after calculating the appropriate Guidelines, a district judge finds that the imposition of a within-Guidelines sentence would visit an injustice upon the defendant pursuant to 18 U.S.C. § 3553(a), it is incumbent upon the judge to say so, and sentence below the Guidelines range. Conversely, when the Guidelines range is too low to satisfy 18 U.S.C. § 3553(a), the district judge must explain why this is so and vary upward.
502 F.3d 293, 308 (3d Cir.2007) (citing Rita, 127 S.Ct. at 2464, 2466). Due process concerns underlie these requirements. United States v. Ausburn, 502 F.3d 313, 322 (3d Cir.2007) (“[D]ue process in criminal sentencing requires that a defendant receive notice of, and a reasonable opportunity to comment on, (a) the alleged factual predicate for his sentence, and (b) the potential punishments which may be imposed at sentence.”); United States v. Fuller, 426 F.3d 556, 565 (2d Cir.2005) (in discussing the pre- and post-Booker cases, the court noted that a district court’s failure to comply with § 3553(c)(2) denies a defendant “ ‘the right to argue more effectively ... whether ... a sentence is ‘reasonable’ ’ ”).
C. An incorrect calculation of the Guidelines range can thwart reasonableness review
Our reasonableness review relies on a district court’s reasoning from the starting point of the correctly calculated Guidelines through the § 3553(a) factors. Our Court, our sister courts of appeals, and the Supreme Court agree that a district court’s use of the incorrect Guidelines range impedes our ability to conduct review of the ultimate sentence.
We have emphasized that a sentencing court’s failure to execute Gunter’s first step will tend to thwart our reasonableness review. See Jackson, 467 F.3d at 838-39 (“[District courts must still calculate what the proper Guidelines sentencing range is, [214]*214otherwise the Guidelines cannot be considered properly at Gunter’s third step.”). For, the correct computation of the Guidelines range and any departures therefrom “serves to clarify the basis for the sentence imposed” and thus facilitates reasonableness review. United States v. Floyd, 499 F.3d 308, 311 (3d Cir.2007). In United States v. Ali, we explained that, by relying on an incorrectly calculated Guidelines range and an improper departure determination, a sentencing judge “necessarily was unable meaningfully to consider the recommended Guidelines range as required by § 3553(a)(4).” 508 F.3d 136, 154 (3d Cir.2007). Thus, we concluded, “the preliminary errors at steps one and two tainted the step three analysis and resulting sentence.” Id. (remanding for resen-tencing based on the court’s error at Gun-ter's step one).
Our sister courts of appeals agree that “the correct guidelines range is still the critical starting point for the imposition of a sentence” and a prerequisite to reasonableness review. Crawford, 407 F.3d at 1178-79; Hawk Wing, 433 F.3d at 631 (internal quotation marks omitted); United States v. Zeigler, 463 F.3d 814, 819 (8th Cir.2006) (Hansen, J., concurring) (“Generally, if the district court errs in applying the Guidelines at step one or fails to consider a requested departure at step two, we cannot conduct a reasonableness review because the district court’s critical starting point, a correctly determined advisory Guidelines range, may be flawed.”); United States v. Staten, 466 F.3d 708, 713 (9th Cir.2006) (“If ... ‘there was material error in the Guidelines calculation that serves as the starting point for the district court’s sentencing decision, we will remand for resentencing pursuant to 18 U.S.C. § 3742(f), without reaching the question of whether the sentence as a whole is reasonable in light of § 3553(a).’”). As the Court of Appeals for the Eighth Circuit has stated, where “the sentence fails the first step of our analysis [because of an incorrect application of the Guidelines], we need not reach the second step, a determination of whether the imposed sentence is reasonable in light of § 3553(a).” United States v. Mashek, 406 F.3d 1012, 1020 (8th Cir.2005); see also United States v. Williams, 456 F.3d 1353, 1360 (11th Cir.2006) (only if the Guidelines calculation is correct or the error harmless can the court go on to consider whether the sentence is reasonable); United States v. HernandezCastillo, 449 F.3d 1127, 1129-30 (10th Cir.2006) (same).
The importance of a correctly calculated range to our reasonableness review is evident in the Supreme Court’s opinions as well. While Gall reinforced a district court’s discretionary authority to choose the substance of a sentence, it also clarified the role of courts of appeals in reviewing procedural and substantive errors in sentencing. In both Gall and Kimbrough, the Court began by noting that the sentencing court had properly calculated and considered the advisory Guidelines range and only then turned to the sentencing court’s consideration of the § 3553(a) factors. Kimbrough, 128 S.Ct. at 575; Gall, 128 S.Ct. at 598. Accordingly, in Gall, the Court instructed us to “ensure that the district court committed no significant procedural error, such as failing to calculate (or improperly calculating) the Guidelines range, treating the Guidelines as mandatory, failing to consider the § 3553(a) factors, selecting a sentence based on clearly erroneous facts, or failing to adequately explain the chosen sentence-including an explanation for any deviation from the Guidelines range” before considering the substantive reasonableness of a sentence. Gall, 128 S.Ct. at 597. This first step, ensuring that the sentencing court did not make a procedural error, is increasingly [215]*215important in light of the Supreme Court’s decision that “a Guidelines sentence will usually be reasonable.” Rita, 127 S.Ct. at 2468.
In sum, while “the district court is free to make its own reasonable application of the § 3553(a) factors, and to reject (after due consideration) the advice of the Guidelines,” Kimbrough, 128 S.Ct. at 577 (Scalia, J., concurring), it must first duly consider the correct Guidelines. Thus, a district court’s incorrect Guidelines calculation will thwart not only its ability to accomplish the analysis it is to undertake, but our reasonableness review as well.
IV. HARMLESS ERROR IN THE SENTENCING CONTEXT
We suggest that, given the importance of a correct Guidelines calculation both to the sentencing process that district courts are required to conduct and to our ability to carry out reasonableness review, the use of an erroneous Guidelines range will typically require reversal under 18 U.S.C. § 3742(f). Nonetheless, under certain, limited circumstances, miscalculation of the Guidelines may be harmless. The government urges that this is one such time because the correct and incorrect ranges here overlap. We are not so sure.
According to our traditional harmless error standard, a non-constitutional error is harmless when “it is highly probable that the error did not prejudice” the defendant. Government of Virgin Islands v. Toto, 529 F.2d 278, 284 (3d Cir.1976). “ ‘High probability’ requires that the court possess a ‘sure conviction that the error did not prejudice’ the defendant.” United States v. Zehrbach, 47 F.3d 1252, 1265 (3d Cir.1995) (quoting United States v. Jannotti, 729 F.2d 213, 219-20 (3d Cir.1984)); accord United States v. Wallace, 461 F.3d 15, 44 (1st Cir.2006) (vacating and remanding under the plain error standard because it “lack[ed] the ‘definite and firm conviction’ that removal of the inappropriate grounds would not be likely to alter the district court’s view of the sentence rightfully to be imposed”); United States v. Duckro, 466 F.3d 438, 446 (6th Cir.2006) (“[w]here a district court makes a mistake in calculating a Guidelines range for purposes of determining a sentence under section 3553(a), we are required to remand for resentencing ‘unless we are certain that any such error was harmless-i,e. any such error “did not affect the district court’s selection of the sentence imposed.” ’ ”) (citations omitted). As the Supreme Court has instructed, the proponent of the sentence bears the burden of “persuading] the court of appeals that the district court would have imposed the same sentence absent the erroneous factor.” Williams v. United States, 503 U.S. 193, 203, 112 S.Ct. 1112, 117 L.Ed.2d 341 (1992). For the error to be harmless, it must be clear that the error did not affect the district court’s selection of the sentence imposed. Id. at 203, 112 S.Ct. 1112. Accordingly, we will remand for resentencing “unless [we] conclude on the record as a whole ... that the error did not affect the district court’s selection of the sentence imposed.” Id.
We submit that the improper calculation of the Guidelines range can rarely be shown not to affect the sentence imposed. In the typical case in which an error in the calculating of Sentencing Guidelines has been held harmless, the sentence was dictated not by the erroneously calculated Guideline, but by a statutory minimum or maximum or another properly calculated Guideline. United States v. Frazier, 213 F.3d 409, 417-18 (7th Cir.2000) (finding that the sentencing court’s miscalculation of the defendant’s criminal history category was irrelevant to the sentence imposed because his offense level carried a life [216]*216sentence); see also United States v. Wilken, 498 F.3d 1160 (10th Cir.2007) (holding that the erroneous application of an enhancement did not affect the district court’s selection of the sentence imposed because defendant qualified as a career offender and as such was subject to categorically prescribed offense and criminal history levels); United States v. Long Soldier, 431 F.3d 1120, 1121 (8th Cir.2005) (“The district court’s erroneous belief that it could not depart upward did not deprive Long Soldier of any substantial-or even beneficial-right. He was deprived merely of the opportunity to receive an upward departure and, perhaps, a longer sentence. As such, any error is harmless.... ”). In such cases, although an error was made, it could not have impacted the ultimate sentence.
The government contends that a sentencing error is also harmless where, as here, the sentence imposed falls into the “overlap” between the incorrect Guidelines range used by the sentencing court and the correct Guidelines range. Although some courts have adopted an “overlapping range” rationale, we conclude that such an “overlap” does not necessarily render an error in the Guidelines calculation harmless. Such an overlap, alone, proves too little. The record must show that the sentencing judge would have imposed the same sentence under a correct Guidelines range, that is, that the sentencing Guidelines range did not affect the sentence actually imposed. The overlap may be helpful, but it is the sentencing judge’s reasoning, not the overlap alone, that will be determinative.3
In United States v. Knight, we made clear that we do not agree that an overlap between ranges renders an error harmless. 266 F.3d 203 (3d Cir.2001). In Knight, the District Court erroneously calculated the Guidelines range as 151 to 188 months and imposed a 162-month sentence that fell within the correct Guidelines range of 140 to 175 months. Id. at 205. Under the exacting plain error standard, we held that “application of an incorrect Federal Sentencing Guidelines range presumptively affects substantial rights, even if it results in a sentence that is also within the correct range.” Id.; see also United States v. Wood, 486 F.3d 781, 790-91 (3d Cir.2007) (relying on Knight post -Booker and vacating and remanding); United States v. Felton, 55 F.3d 861, 869 n. 3 (3d Cir.1995) (“This circuit and others have found that the miscalculation of a defendant’s offense level ‘certainly is error that seriously affeet[s] the defendant’s rights, and so amounts to plain error.’ ”) (citation omitted); United States v. Pollen, 978 F.2d 78, 90 (3d Cir.1992) (“The district court’s improper calculation ..., resulting in a significantly higher Guideline sentencing range, certainly is an error that seriously affected [defendant’s substantial [217]*217rights and so amounts to plain error”). We reviewed numerous cases wherein our sister courts of appeals similarly concluded that the selection of an incorrect Guidelines range was plain error even though the actual sentence happened to fall within the correct Guidelines range. Id. at 208-10. Recognizing that some cases had been to the contrary, we decided that our case law was more sound in that it better protects the defendant’s right to a sentence “imposed pursuant to correctly applied law” and “better effectuates the Guidelines’ purpose to institute fair and uniform sentencing.” Id. at 210. We reviewed the record and determined that “we would be unable to conclude that it is even reasonably likely that the same sentence would have been imposed if the correct range and history were considered.” Id. at 208.
Even when the sentence is below the Guidelines range, the record must be unambiguous that the miscalculation of the range had no effect. Accordingly, in United States v. Thayer, we reviewed a sentence where the district court had erred in calculating the range, but had also granted a six-level downward departure to arrive at an 18-month sentence. 201 F.3d 214 (3d Cir.1999). The government argued that remand was unnecessary because, even if the court used the correct Guidelines range and again departed downward six levels, the range would be 12 to 18 months. Id. at 228. It also contended that, irrespective of the Guidelines range, “the District Court considered 18 months incarceration the proper sentence in Thayer’s case and departed downward six levels in order to obtain that sentence.” Id. at 230. We said that, although the court had stated that it would not impose a sentence of probation, it was unclear on the record that it would not again depart six levels and impose an even lower sentence at the bottom of the Guidelines range. Id. at 230 (stating that “the record support for the possibility Thayer would have received a shorter sentence but for the § 2F1.1(b)(4)(B) enhancement is sufficient to require remand”).
Moreover, when the starting point for the § 3553(a) analysis is incorrect, the end point, i.e., the resulting sentence, can rarely be shown to be unaffected. As noted above, the record must show that the incorrect calculation of the Guidelines did not “result in the district court selecting from the wrong guideline range” and “did not affect the sentence imposed.” Williams, 503 U.S. at 202, 112 S.Ct. 1112. Although many of our harmless error cases were decided in the mandatory Guidelines regime, they point in the same direction under an advisory regime:4 we cannot presume that a district court would have imposed the same sentence, given the opportunity to consider the correctly calculated Guideline. See Pollen, 978 F.2d at 89-90 (noting that speculation would be inappropriate given the nature of a sentencing court’s decision); United States v. Crawford, 407 F.3d 1174, 1183 (11th Cir.2005) (‘We cannot presume that, in the absence of those errors, the district court would have decided that a downward departure was warranted in calculating an advisory guideline range.”). As the Court [218]*218of Appeals for the Tenth Circuit said, “what the district court will do upon resen-tencing absent the illegal presumption ‘places us in the zone of speculation and conjecture.’ ” United States v. Conlan, 500 F.3d 1167 (10th Cir.2007) (citation omitted). An assumption that the sentencing court would have arrived at the same precise sentence is speculation indeed.
In the rare case, a district court may choose to disregard the Guidelines as too severe in such a way that we can be certain that the miscalculation had no effect on the sentence imposed. United States v. Flores may be that unusual case where the sentence imposed was not tied to the Guidelines range or a specific departure or variance from the Guidelines, but rather represented a discretionary sentence imposed based on 3553(a)’s parsimony provision. 454 F.3d 149 (3d Cir.2006). There, after calculating an advisory Guidelines range of 70 to 87 months’ imprisonment, the district court sentenced Flores to 32 months’ imprisonment without granting a formal departure. Id. at 162. On appeal, we concluded that the Guidelines range made no difference: the “District Court clearly considered all the factors in 18 U.S.C. § 3553(a) in reaching its sentence and used its discretion in light of these factors, rather than in the application of a specific downward departure, to go below his advisory Guidelines range to identify the appropriate sentence for Flores.” Id. Thus, the sentencing court could not be said to have imposed the sentence “as a result” of any alleged errors in the calculation of the Guideline.
In order to conclude that a district court would not have imposed a different sentence, the record must be clear. A “blanket statement” that the sentence imposed is fair is not sufficient; a district court must determine a Guidelines range without the miscalculation error and explain any variance from it based on § 3553(a) factors. See United States v. Icaza, 492 F.3d 967, 971 (8th Cir.2007) (“[T]he record must clearly show not only that the district court intended to provide an alternative sentence, but also that the alternative sentence was based on an identifiable, correctly calculated guidelines range.”); United States v. Funk, 477 F.3d 421, 430 (6th Cir.2007) (“If we were to view Funk’s sentence simply as a sentence where ‘the district court independently [chose] to deviate from the advisory guidelines range,’ we would still need to determine whether the district court adequately justified the extent of this deviation.”) (citation omitted). Sentencing in the post-Booker era is a process, not a mere affixing of a point within a mandated range as was previously the case.
The dissent urges that we resolve the issue before us by looking to see if the correct range seems close enough to the actual sentence imposed. If so, the argument goes, the sentence is “reasonable.” However, this ignores the fact that the failure to start with the correct Guideline range is legal error that thwarts reasonableness review — that is, it cuts off our review process before we even reach the issue of reasonableness. As part of the sentencing process, error can occur at the outset, as the Supreme Court noted in Gall, and we must determine if such error is harmless. If it does not impact the analysis and ultimate sentence such that we can say that it probably made no difference, then the ultimate sentence may be reviewed for reasonableness. However, if this cannot be said with some degree of comfort, the sentence must be vacated and the case remanded.
Similarly flawed is the dissent’s adoption of a new test — is the procedural error “insignificant”? In Gall, the Supreme Court clearly considered error in the [219]*219Guideline range to be significant, noting that such errors thwart our reasonableness review and including all such errors in its listing of “significant” procedural errors. Gall, 128 S.Ct. at 597. Moreover, there is no legal test for “insignificance.” Instead, harmlessness is the appropriate barometer and here it cannot be met.
V. APPLICATION OF THE ANALYSIS TO LANGFORD’S SENTENCE
The present case is not that rare case where we can be sure that an erroneous Guidelines calculation did not affect the sentencing process and the sentence ultimately imposed. Contrary to the government’s view, we cannot conclude that the miscalculation of Langford’s criminal history category was harmless.
At Langford’s sentencing, the District Court said that the Sentencing Guidelines “have been deemed to be advisory in nature. They still, however, remain a factor that Court is required to consider in imposing sentence.” App. 122. The District Court did an admirable job of considering the 3553(a) factors and evaluating the characteristics specific to Langford and his offense. The Court then imposed a sentence at the lowest point in the advisory Guidelines range it had calculated.
The government is correct that the 46-month sentence was within the Guidelines range in either case. However, if the criminal history point had not been added, the Court could have imposed a 37-month sentence without departing from the Guidelines, and the 46 months it did impose would have been at the top, not at the bottom, of the proper range.
There is absolutely nothing in the record to indicate that the District Court would have imposed the same sentence under a lower Guidelines range. We must decline the government’s invitation to affirm on the theory that the District Court might have imposed the same sentence. See Thayer, 201 F.3d 214; United States v. Duckro, 466 F.3d 438 (6th Cir.2006) (holding that, even where the district court departed downward significantly from the originally (incorrectly) calculated range, one could not presume that the court would have departed less under a correct and lower Guidelines range).5 We are not persuaded that the record is clear that the sentence imposed was not a result of the erroneous sentencing Guidelines range.
We will remand for the District Court to determine the sentence that should be imposed in light of the correct Guidelines range, considering the 3553(a) factors. Solem v. Helm, 463 U.S. 277, 290 n. 16, 103 S.Ct. 3001, 77 L.Ed.2d 637 (1983) (“[I]t is not the role of an appellate court to substitute its judgment for that of the sentencing court as to the appropriateness of a particular sentence.”).6
[220]*220The goals of uniformity and sentencing discretion are furthered by a remand. Where we conclude that the District Court might have ended up with a different sentence had it started at the right point, giving the Court the opportunity to reconsider the sentence and start at the right place in resentencing actually affords deference and respect for the District Court judge. Our failure to do so would be presumptuous on our part; it is not our role to say that the sentencing judge would consider the sentence he gave, which was at the low end of the incorrectly calculated range, to be appropriate when the correct Guideline range is lower than was assumed. Moreover, insisting on a uniform point of departure from which all sentencing courts can exercise their discretion promotes uniformity in the sentencing of defendants with similar criminal history and offense levels. Surely, a remand with opportunity for reasoning anew is required in order to further both goals.
VI. LANGFORD’S REMAINING ARGUMENT
Langford also argues that his sentence was unreasonable because the District Court violated the law by giving presumptive weight to the Guidelines and imposing a sentence greater than necessary to meet the purposes of sentencing.7 Because Langford was sentenced as a result of an incorrect application of the Guidelines, the sentence fails the first step of our Gunter analysis and, therefore, we need not reach the third step of determining whether the sentence is reasonable in light of the § 3553(a) factors. See Mashek, 406 F.3d at 1020.
VII. CONCLUSION
For the foregoing reasons, we will vacate Langford’s sentence and remand to the District Court for resentencing.