Princeton Vanguard, LLC (“Princeton Vanguard”) appeals from the final decision of the Trademark Trial and Appeal Board (“the Board”) cancelling its registration of the mark PRETZEL CRISPS for pretzel crackers on the Supplemental Register and denying its application to register PRETZEL CRISPS on the Principal Register. Frito-Lay N. Am., Inc. v. Princeton Vanguard, LLC, 109 U.S.P.Q.2d 1949 (T.T.A.B. Feb. 28, 2014) (“Board Decision”). Because the Board applied the incorrect legal standard in evaluating whether the mark is generic, we vacate and remand for further proceedings.
Background
On April 21, 2004, Princeton Vanguard filed U.S. Trademark Application Serial No. 78/405,596, seeking to register PRETZEL CRISPS in standard character format for “pretzels” on an intent-to-use basis under § 1(b) of the Lanham Act, 15 U.S.C. § 1051. The trademark examining attorney refused registration on the Principal Register on grounds that the proposed mark was merely descriptive under 15 U.S.C. § 1052(e)(1). In response, Princeton Vanguard: (1) amended its identification of goods from “pretzels” to “pretzel crackers;” (2) disclaimed the exclusive right to use the term “pretzel” apart from the mark as a whole; and (3) requested registration on the Supplemental Register. Board Decision, 109 U.S.P.Q.2d at 1950 & n. 1. Princeton Vanguard subsequently obtained Registration No. 2,980,303 for the PRETZEL CRISPS mark on the Supplemental Register.
Several years later, Princeton Vanguard filed U.S. Trademark Application Serial No. 76/700,802, seeking to register PRETZEL CRISPS in standard character format for “pretzel crackers” on the Principal Register. In its application, Princeton Vanguard identified October 6, 2004 as its first use of the mark in commerce, disclaimed the exclusive right to use the term “pretzel” apart from the mark as shown, and claimed acquired distinctiveness in the mark as a whole. Id. at 1950 n. 2.
On July 2, 2010, Frito-Lay North America, Inc. (“Frito-Lay”) filed a notice of opposition to Princeton Vanguard’s Application Serial No. 76/700,802 to register PRETZEL CRISPS on the Principal Register. In its opposition, Frito-Lay argued *963that the term PRETZEL CRISPS is generic for pretzel crackers and thus is not registrable. In the alternative, Frito-Lay asserted that PRETZEL CRISPS is highly descriptive of a type of cracker product and has not acquired distinctiveness. Id. at 1950. Frito-Lay subsequently filed a petition to cancel Supplemental Registration No. 2,980,303 on the same grounds. The petition for cancellation was consolidated with the opposition proceeding and both parties filed motions for summary judgment.
In denying Frito-Lay’s motion for summary judgment that the mark is generic, the Board found that Princeton Vanguard’s submissions, including survey evidence and news articles, “give[ ] rise to a genuine dispute regarding the element of whether the public understands ‘pretzel crisps’ as the generic wording for pretzel crackers, or as designating applicant as the source of pretzel cracker products bearing the mark.” Frito-Lay N. Am., Inc. v. Princeton Vanguard, LLC, Opp’n Nos. 91190246 & 91195552, Canc’n No. 92053001, slip op. at 8 (T.T.A.B. Feb. 9, 2011). The Board likewise denied Princeton Vanguard’s later motion for summary judgment, finding a genuine dispute as to whether the mark is generic for pretzel crackers and whether it has acquired distinctiveness. Frito-Lay N. Am., Inc. v. Princeton Vanguard, LLC, Opp’n No. 91195552, Canc’n No. 92053001, slip op. at 3-4 (T.T.A.B. Sept. 5, 2012). The parties agreed to proceed to trial on the summary judgment record, as well as supplemental expert declarations. That record included, among other things, testimony in the form of declarations with attached exhibits, media references, third party use in the food industry, use by the parties themselves, and expert survey evidence. Board Decision, 109 U.S.P.Q.2d at 1951. Both parties filed trial briefs and, pursuant to Princeton Vanguard’s request, the Board conducted oral argument.
On February 28, 2014, the Board sustained Frito-Lay’s opposition to Princeton Vanguard’s application and granted its petition for cancellation. Id. at I960.1 At the outset, the Board found that “pretzel crisps” is a compound term, not a phrase, and analyzed the terms individually. Id. at 1953 (noting that “compound words that do not add new meaning may be analyzed by their constituent terms”). Specifically, the Board found that the term “pretzel” in PRETZEL CRISPS is generic for pretzels and pretzel snacks, and the term “crisps” is generic for crackers. In reaching this conclusion, the Board cited: (1) media references and third-party use of the term “crisps” to identify crackers; (2) registrations disclaiming the term “crisps”; (3) dictionary definitions of the word “crisp”; (4) Princeton Vanguard’s admission that its packages for its PRETZEL CRISPS products provide nutritional facts for a serving size of a stated number of “crisps”; (5) a few generic references to the combined term “pretzel crisps;” and (6) the surveys of two of the parties’ experts: Dr. Alex Simonson for Frito-Lay and Dr. E. Deborah Jay for Princeton Vanguard.
Dr. Simonson’s survey found that 41% of respondents, thought PRETZEL CRISPS was a brand name, 41% thought it was a category name, and 18% answered “don’t know” or “not sure.” Id. at 1957. Dr. Jay noted several problems with Dr. Simonson’s methodology, including that “the universe of survey participants was underinclusive,” the two options of giving no opinion — “don’t know” and “not sure”— may have caused confusion, and Dr. Si-monson faded to conduct a “mini-test” to *964determine whether participants understood the difference between brand names and category (generic) names. Id. at 1957-58. The Board agreed with Dr. Jay’s criticisms of Dr. Simonson’s survey, and thus gave Frito Lay’s survey little probative weight. As to Dr. Jay’s survey — which found that 55% of respondents thought that PRETZEL CRISPS was a brand name, while 36% thought it was a common or generic name — the Board noted Dr. Simonson’s objection to the screening criteria, but did not adopt it. Id. at 1958.
The Board indicated that it considered the entire record, including the surveys, but gave “controlling weight to the dictionary definitions, evidence of use by the public, including use by the media and by third-parties in the food industry, and evidence of use by defendant itself.” Id. at 1960. On this record, the Board found “PRETZEL CRISPS” is generic for “pretzel crackers.” Id. The Board explained that its conclusion would be the same if it -had analyzed PRETZEL CRISPS as a phrase instead of a compound term, because “the words strung together as a unified phrase also create a meaning that we find to be understood by the relevant public as generic for ‘pretzel crackers.’ ” Id. Given the Board’s finding of genericness, it did not address whether the term PRETZEL CRISPS, when used in connection with pretzel crackers, has acquired distinctiveness. Id. at 1960 n. 13.
Princeton Vanguard timely appealed to this court. We have jurisdiction pursuant to 28 U.S.C. § 1295(a)(4)(B).
Discussion
We review the Board’s legal conclusions de novo and its factual findings for substantial evidence. In re Pacer Tech., 338 F.3d 1348, 1349 (Fed.Cir.2003). Whether the Board applied the correct legal standard in assessing the mark is a question of law we review de novo. See In re Dial-A-Mattress Operating Corp., 240 F.3d 1341, 1345 (Fed.Cir.2001). Whether an asserted mark is generic is a question of fact. In re Hotels.com, LP, 573 F.3d 1300, 1301 (Fed.Cir.2009). Accordingly, “[o]n appellate review of the Board’s factual finding of genericness, we determine whether, on the entirety of the record, there was substantial evidence to support the determination.” Id. at 1302. Substantial evidence requires “more than a mere scintilla” and is “ ‘such relevant evidence as a reasonable mind would accept as adequate’ to support a conclusion.” Pacer Tech., 338 F.3d at 1349 (quoting Consol. Edison v. NLRB, 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938)).
On appeal, Princeton Vanguard argues that the Board erred in its generieness analysis when it assessed the PRETZEL CRISPS mark as a compound term instead of a phrase. Specifically, Princeton Vanguard contends that the Board “took the short-cut of analyzing whether the words ‘pretzel’ and ‘crisps’ are each generic for a pretzel and a cracker, and then it merely assumed the public would understand the combined mark PRETZEL CRISPS to be generic for ‘pretzel crackers’ without due consideration of the actual evidence of record.” Appellant Br. 9. According to Princeton Vanguard, by focusing solely on the mark’s constituent parts, the Board: (1) deviated from this court’s precedent, which requires consideration of the mark it its entirety; and (2) failed to consider the evidence of record, which shows that the purchasing public understands the term PRETZEL CRISPS to be a brand name.
For the reasons explained below, we agree with Princeton Vanguard that the Board applied the incorrect legal standard when it failed to consider the relevant *965public’s understanding of the PRETZEL CRISPS mark in its entirety.
A. The Board Applied the Incorrect Legal Standard
A generic term “is the common descriptive name of a class of goods or services.” H. Marvin Ginn Corp. v. Int’l Ass’n of Fire Chiefs, Inc., 782 F.2d 987, 989 (Fed.Cir.1986). Because generic terms “are by definition incapable of indicating a particular source of the goods or services,” they cannot be registered as trademarks. Dial-A-Mattress, 240 F.3d at 1344. “The critical issue in genericness cases is whether members of the relevant public primarily use or understand the term sought to be protected to refer to the genus of goods or services in question.” Marvin Ginn, 782 F.2d at 989-90.
We have said that determining a mark’s genericness- requires “a two-step inquiry: First, what is the genus of goods or services at issue? Second, is the term sought to be registered or retained on the register understood by the relevant public primarily to refer to that genus of goods or services?” Id. at 990. Evidence of the public’s understanding of the mark may be obtained from “any competent source, such as consumer surveys, dictionaries, newspapers and other publications.” In re Northland Aluminum Prods., Inc., 777 F.2d 1556, 1559 (Fed.Cir.1985). In an opposition or cancellation proceeding, the oppo-ser or petitioner bears the burden of proving generic-ness by a preponderance of the evidence. See Magic Wand, Inc. v. RDB, Inc., 940 F.2d 638, 641-42 (Fed.Cir.1991).2
Applying the first prong of the Marvin Ginn test, the Board defined the genus of goods at issue as “pretzel crackers.” Board Decision, 109 U.S.P.Q.2d at 1952. Turning to the second prong, the Board identified the relevant public as “ordinary consumers who purchase and eat pretzel crackers.” Id. Neither party disputes these findings on appeal.
The Board began its discussion of the public’s perception of the mark by stating that it “first must decide how to analyze the term.” Id. Although the Board acknowledged that the ultimate inquiry is whether the mark as a whole is generic, it then cited In re Gould Paper Corp., 834 F.2d 1017 (Fed.Cir.1987), for the proposition that, “in cases where the proposed mark is a compound term (in other words a combination of two or more terms in ordinary grammatical construction), gener-icness may be established with evidence of the meaning of the constituent words.” Id. (citing Gould, 834 F.2d at 1019). The Board indicated that, “[b]y contrast, “where the proposed mark is a phrase ... the board cannot simply cite definitions *966and generic uses of the constituent terms of a mark; it must conduct an inquiry into the meaning of the disputed phrase as a whole.’ ” Id. (quoting Dial-A-Mattress, 240 F.3d at 1345 (citing In re Am. Fertility Soc’y, 188 F.3d 1341, 1347 (Fed.Cir.1999))).
The Board appears to believe that there is a dichotomy in the standard applicable to a particular mark depending on whether it is a compound term or a phrase. According to the Board, if the mark is a compound term, then Gould applies, and it can focus on the individual words, but if it is a phrase, American Fertility requires that the Board consider the mark in its entirety. Id. at 1953. Because the Board found “no additional meaning added to ‘PRETZEL CRISPS’ in relation to ‘pretzel crackers,’ when the individual terms are combined,” the Board analyzed it as a compound term. Id. The Board then considered the terms individually and concluded that “pretzel” is generic for pretzels and pretzel snacks, and “crisps” is generic for crackers. Id.
The problem with the Board’s analysis is that there is only one legal standard for genericness: the two-part test set forth in Marvin Ginn. Am. Fertility, 188 F.3d at 1348. As noted, to determine whether a mark is generic under that test, the Board must first identify the genus of goods or services at issue, and then assess whether the public understands the mark, as a whole, to refer to that genus. Marvin Ginn, 782 F.2d at 990. On appeal, Frito-Lay cites our decisions in Gould and American Fertility to suggest that the Board can somehow short-cut its analysis of the public’s perception where “the purported mark is a chmpound term consisting merely of two generic words.” Appellee Br. 21. As discussed below, however, there is no such short-cut, and the test for genericness is the same, regardless of whether the mark is a compound term or a phrase. Neither Gould nor American Fertility hold otherwise.
The applicant in Gould sought to register the mark SCREENWIPE for goods identified as “pre-moistened, anti-static cloth for cleaning computer and television screens.” 834 F.2d at 1017. While the Board looked to the individual definitions of “screen” and “wipe,” we found that “Gould’s own submissions provided the most damaging evidence that its alleged mark is generic and would be perceived by the purchasing public as merely a common name for its goods rather than a mark identifying the good’s source.” Id. at 1018-19. Indeed, Gould described its own product as “a ... wipe ... for ... screens.” Id. at 1019. Given this admission, we noted that the “compound immediately and unequivocally describes the purpose, function and nature of the goods as Gould itself tells us.” Id. (“Gould has simply joined the two most pertinent and individually generic terms applicable to its product, and then attempts to appropriate the ordinary compound thus created as its trademark.”). In that context, where the mark in its entirety has exactly the same meaning as the individual words, we stated that “the PTO has satisfied its evidentiary burden if, as it did in this case, it produces evidence ... that the separate words joined to form a compound have a meaning identical to the meaning common usage would ascribe to those words as a compound.” Id. at 1018. Because “the terms remain as generic in the compound as individually,” we concluded that the compound itself was generic. Id. at 1019.
In American Fertility, we explained that: (1) Gould did not alter the legal standard for genericness; and (2) the correct legal test “is set forth in Marvin Ginn and is to be applied to a mark ... as a whole, for the whole may be greater *967than the sum of its parts.” Am. Fertility, 188 F.3d at 1348. In that case, the Board held that the proposed mark — SOCIETY FOR REPRODUCTIVE MEDICINE— was generic based on evidence of record relating solely to the individual terms. Id. at 1344 (“Despite the lack of evidence of the public’s understanding of the phrase as a whole, the Board held that the ‘combination of the terms ‘society’ and ‘reproductive medicine’ results in a designation, SOCIETY FOR REPRODUCTIVE MEDICINE, which is also generic.’ ”) (citation omitted). We vacated the Board’s decision on appeal, finding that it applied the incorrect legal standard when it “assumed the genericness of a phrase as a whole based solely on proof of the genericness of its individual terms.” Id. at 1342.
In reaching this conclusion in American Fertility, we clarified that the Gould decision “did not purport to modify Marvin Ginn ... and seemingly sought to follow” it by focusing on Gould’s generic use in its entirety. Id. at 1347. We reiterated that Marvin Ginn requires evidence of the “genus of goods or services at issue” and the “understanding by the general public that the mark refers primarily to ‘that genus of goods or services.’ ” Id. “Properly interpreted, Gould does not justify a short-cut around this test, but rather found that the evidence presented met this burden.” Id. at 1348. We further indicated that the Board “cannot simply cite definitions and generic uses of the constituent terms of a mark ... in lieu of conducting an inquiry into the meaning of the disputed phrase as a whole to hold a mark ... generic.” Id. at 1347. Accordingly, we remanded the case for the Board to apply the Marvin Ginn test to the mark as a whole. Id. at 1349.
In subsequent decisions, this court has reiterated that, “[a]n inquiry into the public’s understanding of a mark requires consideration of the mark as a whole. Even if each of the constituent words in a combination mark is generic, the combination is not generic unless the entire formulation does not add any meaning to the otherwise generic mark.” In re Steelbuilding.com, 415 F.3d 1293, 1297 (Fed.Cir.2005). In Steelbuilding, the applicant sought to register the mark STEELBUILDING.COM in connection with “computerized on-line retail services in the field of pre-engi-neered metal buildings and roofing systems.” Id. at 1296. The Board found that the mark was either generic or highly descriptive, and that there was insufficient evidence of acquired distinctiveness. Id. On appeal, we found that the record did not contain substantial evidence that “STEELBUILDING,” “in common usage, is a compound word used to mean either ‘steel building’ or ‘steel buildings.’ ” Id. at 1298 (citing Gould, 834 F.2d at 1018). In reaching this conclusion, we emphasized that Gould required consideration of the mark in its entirety, and that the applicant in Gould “admitted that ‘screen wipe’ denoted a ‘screen wipe.’ ” Id. at 1298. In other words, the combined term at issue in Gould — SCREENWIPE—itself described the genus of goods at issue. In contrast, the applicant denied that “STEELBUILD-ING” describes merely “steel buildings.” Id. Although the Board cited evidence showing that “steel building” was generic, that evidence did not “address directly the composite term STEELBUILDING.” Id. at 1299. We concluded, therefore, that the record did not contain sufficient evidence to support the Board’s finding that STEELBUILDING.COM was generic for the applicant’s services. Id.3
*968On appeal, Frito-Lay argues that American Fertility is distinguishable from this case because it involved a phrase, not a compound term. According to Frito-Lay, the applicable legal standard depends on an initial “factual determination of whether the purported mark is a compound term or a phrase.” Appellee Br. 36-37.4 We disagree. Regardless of whether the mark is a compound term or a phrase, the applicable test is the same and the Board must consider the record evidence of the public’s understanding of the mark as a whole. Am. Fertility, 188 F.3d at 1348-49. Our decision in Gould merely provides additional assistance in assessing the genericness of compound terms where it can be shown that “the public understands the individual terms to be generic,” and the joining of those terms into one compound word provides no additional meaning. Id. It is not a short-cut and does not supplant the two-part test set forth in Marvin Ginn.
Even in cases where we have recognized that Gould provides an evidentiary standard applicable to compound terms, we have nonetheless emphasized that the Board must consider the mark in its entirety. See Hotels.com, 573 F.3d at 1304-06 (citing Gould and concluding that “the Board satisfied its evidentiary burden, by demonstrating that the separate terms ‘hotel’ and ‘.com’ in combination have a meaning identical to the common meaning of the separate components”); see also In re 1800Mattress.com IP, LLC, 586 F.3d 1359, 1363 (Fed.Cir.2009).
In 1800Mattress.com, for example, the Board initially considered the terms “mattress” and “.com” individually and determined that they were both generic. 586 F.3d at 1363. We explained. that the Board “then considered the mark as a whole and determined that the combination added no new meaning, relying on the prevalence of the term ‘mattress.com’ in the website addresses of several online mattress retailers that provide the same services as Dial-A-Mattress.” Id. In affirming the Board’s decision, we indicated that the Board “properly determined” that “the correct inquiry is whether the relevant public would understand, when hearing the term ‘mattress.com,’ that it refers to online mattress stores.” Id. at 1364. Accordingly, even in circumstances where the Board finds it useful to consider the public’s understanding of the individual words in a compound term as a first step in its analysis, the Board must then consider available record evidence of the public’s understanding of whether joining those individual words into one lends additional meaning to the mark as a whole. See id. *969at 1363-64; see Hotels.com, 573 F.3d at 1304-06.
As noted, here, the Board determined that PRETZEL CRISPS is a compound term, not a phrase, and evaluated the terms individually under Gould. Board Decision, 109 U.S.P.Q.2d at 1953. The Board concluded that the “commonly understood meaning of the words ‘pretzel’ and ‘crisps,’ demonstrates that purchasers understand that ‘PRETZEL CRISPS’ identifies ‘pretzel crackers.’ ” Id. at 1959. Where, as here, the record is replete with evidence of the public’s perception of the term PRETZEL CRISPS as a whole, it is unclear why the Board would resort to analyzing the terms individually or why it would believe doing so would aid its analysis. In any event, as we have explained, the appropriate legal standard set forth in Marvin Ginn requires consideration of the mark as a whole.
At the end of its decision, the Board stated in passing that, although it analyzed the term PRETZEL CRISPS as a compound term, “were we to analyze it as a phrase, on this record, our conclusion would be the same, as the words strung together as a unified phrase also create a meaning that we find to be understood by the relevant public as generic for ‘pretzel crackers.’ ” Board Decision, 109 U.S.P.Q.2d at 1960. But the Board’s decision lacks any indicia that it actually engaged in such an analysis. Specifically, as explained below, there is no evidence that the Board conducted the necessary step of comparing its findings with respect to the individual words to the record evidence demonstrating the public’s understanding of the combined term: PRETZEL CRISPS. By failing to do so, the Board took the type of short-cut analysis we have said is prohibited and ignored evidence that might compel a contrary conclusion. See Am. Fertility, 188 F.3d at 1348 (stating that “Gould does not justify a short-cut around”. the test articulated in Marvin Ginn). We conclude, therefore, that the Board applied the incorrect legal standard in its genericness determination. On re- mand, the Board must consider evidence of the relevant public’s understanding of the term PRETZEL CRISPS in its entirety.
B. Evidence of Record
As previously discussed, the relevant public’s perception is the primary consideration in determining whether a term is generic. In re Merrill Lynch, Pierce, Fenner & Smith, Inc., 828 F.2d 1567, 1569 (Fed.Cir.1987) (“It is basic to the inquiry to determine whether members of the relevant public primarily use or understand the term to refer to the genus of goods or services.”). And, as noted, evidence of the public’s perception may be obtained from “any competent source, such as consumer surveys, dictionaries, newspapers and other publications.” Northland Aluminum, 777 F.2d at 1559.
One of our sister circuits has indicated that “direct consumer evidence, e.g., consumer surveys and testimony is preferable to indirect forms of evidence.” Berner Int’l Corp. v. Mars Sales, Co., 987 F.2d 975, 982-83 (3d Cir.1993) (“Consumer surveys have become almost de rigueur in litigation over genericness.”) (internal citation and quotation marks omitted). We likewise have recognized that “consumer surveys may be a preferred method of proving genericness.” BellSouth Corp. v. DataNational Corp., 60 F.3d 1565, 1570 (Fed.Cir.1995) (“While consumer surveys may be a preferred method of proving genericness under the proper test of purchaser understanding, we are satisfied that on the facts of this case genericness has been established under that test.”).
On appeal, Princeton Vanguard argues that the Board cherry-picked the media *970references in the record and chose only those references that supported genericness. According to Princeton Vanguard, the Board failed to consider evidence that, “since the launch of PRETZEL CRISPS pretzel crackers, 86% of unsolicited media references clearly use the term PRETZEL CRISPS” as a brand name. Appellant Br. 23. Next, Princeton Vanguard argues that the Board failed to consider the declarations it presented from four independent participants in the snack food industry who testified that they use and understand the term PRETZEL CRISPS exclusively to refer • to Snack Factory’s PRETZEL CRISPS products.5 Instead, the Board cited “two instances in which a snack food company used the term • PRETZEL CRISPS in an arguably generic fashion.” Id. at 27. Finally, Princeton Vanguard argues that the Board erred in disregarding the results of Dr. Jay’s survey without any explanation.
In response, Frito-Lay argues that, “even if the Board somehow ‘cherry picked’ generic uses of ‘pretzel crisps’ by the media (which it did not), there is ample other evidence supporting its genericness finding, including dictionary definitions; generic use by Princeton, competitors, and consumers; and survey evidence.” Appel-lee Br. 43-44. Frito-Lay maintains that the Board’s findings are reasonable and supported by substantial evidence.
Given our conclusion that the Board applied the incorrect standard for genericness and our decision to remand for application of the correct legal test, we need not analyze the parties’ specific arguments with respect to the evidence of record. We reiterate, however, that substantial evidence review “requires an examination of the record as a whole, taking into account both the evidence that justifies and detracts from an agency’s opinion.” Falkner v. Inglis, 448 F.3d 1357, 1363 (Fed.Cir.2006). Our review under that standard “can only take place when the agency explains its decisions with sufficient precision, including the underlying factfindings and the agency’s rationale.” Packard Press, Inc. v. Hewlett-Packard Co., 227 F.3d 1352, 1357 (Fed.Cir.2000).
The record here contains significant evidence in the form of declarations, survey evidence, and evidence of use of PRETZEL CRISPS in the snack food industry and by the media and Frito-Lay does not dispute Princeton Vanguard’s description of much of it. Though the Board is not required to discuss every piece of evidence, it cannot focus primarily on evidence of the word “crisps” in isolation, select a few pieces of evidence involving the combined term “pretzel crisps,” and conclude that the trademark is generic. Nor can it disregard the results of survey evidence without explanation. Just as it may not short-cut its legal analysis, the Board may not short-cut its consideration of the factual record before it.
As previously noted, the Board expressly agreed with Dr. Jay’s criticisms of Dr. Simonson’s survey and gave his findings “little probative weight.” Board Decision, 109 U.S.P.Q.2d at 1958. In particular, the Board concluded that the two “don’t know” and “not sure” answers “potentially were confusing to survey participants, and may have le[d] those who understood the survey question to elect to indicate they did not.” Id. As to Dr. Jay’s survey, which found that 55% of respondents thought that PRETZEL CRISPS was a brand name, the Board merely noted Dr. Simon-son’s criticism, which was that “less than 65% of the initial group ‘of qualified re*971spondents’ was entered into the survey due to the underinclusive nature of the questions.” Id. at 1958. The Board did not agree with this critique, however. Id. Nor did the Board call into question Dr. Jay’s conclusion that “the primary significance of the name ‘PRETZEL CRISPS’ to past and prospective purchasers of salty snacks is as a brand name.” Id. Nevertheless, in finding the mark generic, the Board indicated that it considered the evidence of record “including the surveys (which in any event arrive at different conclusions),” but gave controlling weight to dictionary definitions, evidence of use by the public, and evidence of use by Princeton Vanguard. Id. at 1960. The Board seems to have treated the surveys as though they cancelled each other out, but failed to offer any explanation for doing so. The Board thus overlooked or disregarded a genericness survey as to which it apparently found no flaw. On remand, the Board will have the opportunity to make the relevant factual findings based on all of the evidence of record, and must give appropriate consideration to the proffered survey evidence.
Conclusion
Because we find that the Board applied the incorrect legal standard in assessing whether the term PRETZEL CRISPS is generic, we vacate the Board’s decision cancelling Registration No. 2,980,303 and its decision sustaining Frito-Lay’s opposition to Application Serial No. 76/700,802. We remand for application of the appropriate standard as set forth in Marvin Ginn and discussed herein.
VACATED AND REMANDED