ORDER GRANTING MOTION FOR LEAVE TO SUPPLEMENT BRIEFING; AND

ORDER DENYING MOTION FOR REMAND

After Defendant Pacer Cartage removed this putative class action from state court, Plaintiff Manuela Mendoza moved to remand. Later, Edwin Molina was substituted in as named Plaintiff, and therefore replaces Mendoza as movant. Pacer later asked the Court to consider a supplemental response.

The unopposed motion for leave to file a supplemental response (Docket no. 13) is GRANTED. The Court will consider it along with the other briefing on the issue of remand.

Removal

Under 28 U.S.C. § 1441 (a), a case can be removed from state to federal court, provided it could originally have been brought in federal court. This statute is construed strictly against removal, and “[fjederal jurisdiction must be rejected if there is any doubt as to the right of removal in the first instance.” Gaus v. Miles, Inc., 980 F.2d 564, 566 (9th Cir.1992); see also Boggs v. Lewis, 863 F.2d 662, 663 (9th Cir.1988). The removing party bears the burden of establishing that the court has subject matter jurisdiction. Abrego Abrego v. The Dow Chemical Co., 443 F.3d 676, 685 (9th Cir.2006).

Federal courts are presumed to lack jurisdiction, and the burden always falls on the party invoking it. See Gen. Atomic Co. v. United Nuclear Corp., 655 F.2d 968, 968-69 (9th Cir.1981). Although the Court looks in the first instance to the remand motion, the Court is also under an independent obligation to confirm its own jurisdiction even if jurisdictional defects *1063 are not raised by the parties. See United Investors Life Ins. Co. v. Waddell & Reed Inc., 360 F.3d 960, 966 (9th Cir.2004). Thus, Molina’s failure to raise issues or make arguments does not prevent the Court from considering those issues or arguments sua sponte; rather, the Court is obligated to consider any possible defects in jurisdiction its analysis may discover, even if Molina has not raised them. By contrast, the Court is not allowed to create arguments or engage in speculation in favor of jurisdiction; any doubts must be resolved against jurisdiction and in favor of remand. Gaus, 980 F.2d at 566 .

In removing the case, Pacer relied on the Class Action Fairness Act, 28 U.S.C. § 1332 (d), under which the Court has jurisdiction over matters where, among other things, the amount in controversy exceeds five million dollars. The remand motion argues that the threshold is not met, and the Court therefore lacks jurisdiction.

Amount in Controversy

This is a wage and hour case, arising under California law. The complaint alleges that Pacer misclassified employees as independent contractors, wrongly required class members to work through meal breaks, did not provide required rest breaks, and failed to pay overtime. The notice of removal bases the amount in controversy on these three claims.

As mentioned in the notice of removal, the complaint does not plead a particular amount in controversy or seek particular amounts of damages. Pacer therefore must establish by a preponderance of evidence that the aggregate amount in controversy exceeds $5 million. Rodriguez v. AT & T Mobility Servs., LLC, 728 F.3d 975, 981 (9th Cir.2013). This cannot be based on speculation and conjecture, see Roth v. Comerica Bank, 799 F.Supp.2d 1107, 1118 (C.D.Cal., 2010), but must be based instead on facts. See Korn v. Polo Ralph Lauren Corp., 536 F.Supp.2d 1199, 1206 (E.D.Cal., 2008) (citing Gaus, 980 F.2d at 567 ).

This means that Pacer is required to “set forth underlying facts to support key variables used in [its] calculations.” See Manier v. Medtech Products, Inc., 2014 WL 1609655 , at *2 (S.D.Cal., Apr. 22, 2014). The facts may come from the complaint itself (i.e., facts pleaded on the face of the complaint), or facts set forth in the removal petition. Abrego, 443 F.3d at 690 . The Court may also consider “summary-judgment-type” evidence. Id. Estimates must be reasonable and fact-based, not speculative or inflated. See Romsa v. Ikea U.S. West, Inc., 2014 WL 4273265 , at *2 (C.D.Cal., Aug. 28, 2014) (citing Cohn v. Petsmart, Inc., 281 F.3d 837, 840 (9th Cir.2002)). See also Behrazfar v. Unisys Corp., 687 F.Supp.2d 999, 1004 (C.D.Cal., 2009) (finding by a preponderance of evidence that the amount in controversy was met, where calculations were “relatively conservative, made in good faith, and based on evidence whenever possible”). Because Pacer bears the burden of showing removal was proper, its failure to produce evidence (whether by pointing out allegations in the complaint or by presenting facts in its own briefing) would mean it has not carried its burden, at least as to that claim. See, e.g., Reames v. AB Car Rental Servs., Inc., 899 F.Supp.2d 1012, 1016 (D.Or., 2012) (“[B]e-cause defendants submitted no evidence as to the amount of plaintiffs attorney fees incurred as of the time of removal, defendants have not met their burden” of establishing that removal was proper).

There is no real dispute between the parties about these standards. Rather, the focus is on how they apply to the pleadings and facts. The remand motion *1064 argues that Pacer’s estimates are inflated, not conservative, and not based on facts.

The complaint gives an estimate of over 200 putative class members. (Compl., ¶ 24.) Pacer’s CEO, in a declaration attached to the notice of removal, estimates based on a review of records at least 309 class members, some of whom are no longer working for Pacer. (Decl. of Van Noel (Docket no. 1-5), ¶¶ 3-5.) According to Pacer’s supplemental briefing, there are at least 584 putative class members, who worked a total of at least 49,640 weeks during the period covered by this action. (Docket no. 13.) This estimate is based on new information unavailable at the time the motion was originally briefed. Although the number of workers is considerably higher, the number of weeks increased by only a few thousand. Apparently, the newly-identified workers did not work as much time during the relevant period as did the previously-identified workers.

Analysis of Molina’s Position

Both Molina and Pacer cite Ray v. Wells Fargo Bank, N.A., 2011 WL 1790123 , at *7 (C.D.Cal., May 9, 2011), a wage and hour action in which conservative estimates were used to determine the amount in controversy. Molina argues that although Pacer’s estimates are supported in part by declarations, the amount of meal and rest periods missed and the number of days worked overtime are “pure conjecture,” unsupported by any facts. Molina also points out that Pacer’s estimates are significantly higher than those approved in Ray .

Molina agrees that the workers’ hourly rate of $15 for ordinary work and 41,090 weeks 1 are based, on declarations, but contests the amount of meal periods per week missed. Molina points out that Molina relies on an allegation in the complaint that class members were, “routinely required to work without an uninterrupted meal break,” to conclude that this happened at least twice a week. (Mot., 5:25-28) (citing Notice of Removal 4 n. 3 and Compl., ¶¶ 16, 41.)

Molina also distinguishes Ray . In that case, the plaintiff alleged that all class members were paid on a salary basis with no overtime compensation and were “consistently” required to work overtime without extra pay. The Ray court, based on this, determined that a conservative estimate was one hour of unpaid overtime every two weeks, and one missed meal period every two weeks. Molina argues that “routinely” does not imply that putative class members were denied two meal breaks a week or two rest periods per week, or anything close to it. But here, Molina is merging two different claims. The claim for missed meal breaks does allege that employees “routinely” had to work through meals. (Compl., ¶ 41.) And the Court agrees that “routinely” leaves open a wide range of frequency. But with regard to overtime, the complaint alleges that “during the Class period, Plaintiff and other members of the Class consistently worked three to seven (7) days per week for nine (9) hours or more.” (Compl., ¶ 55 (emphasis added).) This implies at least three hours of overtime per week (for those who worked three days), and “consistently” suggests it was at least commonplace.

Molina argues that Pacer’s “conservative estimate” of five hours of overtime per week is similarly based on conjecture, and does not represent a conservative esti *1065 mate. The notice of removal says Pacer selected five hours because the complaint alleged that class members worked from three to seven hours per week, and five is the midpoint of that range.

While the complaint requests attorney’s fees as provided by law, it did not request a particular amount or percentage. Molina argues that Pacer’s estimate that these will amount to 25% of the overall recovery is merely speculation. For claims arising from a violation of the Labor Code, California provides for an award of attorney’s fees based on the lodestar method, not on a percentage of the recovery, see Ketchum v. Moses, 24 Cal.4th 1122, 1135-36 , 104 Cal.Rptr.2d 377 , 17 P.3d 735 (2001), so a reasonable estimate of attorney’s fees would be based on the number of hours it is expected Plaintiffs counsel will work, multiplied by a reasonable fee. This can be shown by introducing evidence regarding the usual amount of fees recovered in similar actions, but Pacer did not do this.

Analysis of Pacer’s Position

Pacer cites to Mendoza’s work logs, showing how many overtime hours she worked during the relevant period, or how many meal or rest breaks she missed. From this, Pacer infers what it considers a conservative and reasonable estimate of what other workers’ claims are likely to be. At the outset is worth remembering that this is a class of hundreds of workers, and that when, as here, class members may have claims of various sizes, the person selected to be class representative is commonly, though not always, someone with a relatively large claim, or even the largest claim. See Charles A. Wright & Arthur R. Miller, 7A Federal Practice & Procedure § 1767 (3d ed.2007). It is also to be expected that one of the most aggrieved class members would step forward as the named plaintiff. Compare City of Greenville v. Syngenta Crop Protection, Inc., 904 F.Supp.2d 902, 904 (S.D.Ill., 2012) (“As would be expected, systems with ... the largest claims, filed claims at an even higher rate.”)

But the size of Mendoza’s claim, or the frequency with which she worked unpaid overtime or missed breaks says little about what the rest of the class experienced. In class actions with variable claim amounts, the claim size frequently skews towards the lower end of the scale. See, e.g., Williams v. Aramark Sports, LLC, 2011 WL 4018205 , at *6 n. 6 (E.D.Pa„ Sept. 9, 2011) (in wage and hour class action, highest payable claim was over five times what the average claimant would receive). The evidence of whát one selected worker experienced does not speak to the overall distribution of overtime hours or missed breaks.

Pacer’s opposition estimates the hours of overtime based on the complaint, which says that when workers were required to work overtime, they were required to work at least one hour of overtime. Pacer also points to ¶ 55 of the complaint, which alleges that “during the Class period, Plaintiff and other members of the Class consistently worked three to seven (7) days per week for nine (9) hours or more.” The Court construes this to mean that class members consistently worked overtime for at least one hour and for at least three days a week. “Consistently” implies it was widespread and occurred as a matter of course, and that it was not irregular or sporadic. It does not necessarily mean it occurred every week with every worker, though. With regard to the average number of hours per week, Pacer has not demonstrated any basis for selecting five hours of overtime per worker per week. In a situation such as the complaint alleges, there are likely to be far more claims at the estimated low end than at the estimated high end. Selecting the *1066 midpoint (five hours per week) is likely to skew the numbers upward. But, at the very least, it cannot be called a conservative estimate. Rather, a conservative estimate would be that the class members worked three hours overtime per week; this would account for those weeks when they worked more (as alleged), as well as some weeks when they worked fewer overtime hours, or perhaps none at all (“leaving room for “consistently” to mean something less than every week without fail.”)

Pacer cites to Mendoza’s work logs, to show that she worked an average of nearly ten hours of overtime per week. But, as discussed Mendoza’s work habits do not necessarily speak to those of the class, and it is very possible most other workers far less overtime than she did. The Court also notes that the hourly logs for Mendoza show wide variations in the amount of overtime she worked, from a high of 17 hours to a low of 0.5 hours. There is no reason to infer that other workers steadily or consistently worked overtime at the same rate.

For each hour of overtime, the class members were allegedly underpaid by $22.50, 2 which represents, the uncontested estimated hourly rate of $15.00 times the required 1.5 multiplier for overtime. Multiplying this by three hours per week and by the higher total of 49,640 weeks worked by class members during the relevant period, the result is $8,350,700. This represents a conservative estimate of the unpaid overtime wages claimed in the complaint.

With regard to the meal and rest break claims however, Pacer’s opposition doesn’t address the heart of the motion’s argument. Instead, it cites Ray’s interpretation of an allegation, and applies it to different facts. In Ray , the court considered an allegation that putative class members “routinely” worked overtime hours and determined that a conservative estimate was that they worked one overtime hour every two weeks. With regard to the missed meal breaks, the complaint alleges that Pacer

(Compl., ¶ 41.) With regard to rest breaks, the allegations are almost identical. {See Compl., ¶ 48.) From this, Pacer relies on Ray to infer that putative class members are alleged to have missed at least two breaks per week; according to the opposition, “Ray stands for the proposition that ‘routinely’ means at least once per week.... ” (Opp’n at 14:6.)

Ray , however, uses the word “routinely” once, and even then only when characterizing the complaint, not when quoting it. Ray construed an allegation of “consistent” overtime work, concluding (as this order does) that a conservative estimate was one hour of overtime per week. The court, in the course of discussing what “consistent” overtime meant, used the word “routinely” to describe it. Ray, 2011 WL 1790123 at *6. It should also be remembered that the Ray court was talking about weekly overtime, not about something potentially less regular such as missing breaks.

“Routinely” does not refer so much to frequency as to custom, policy or habit. The allegation that Pacer either directed or acquiesced in this suggests that it may have happened when Pacer found it conve *1067 nient and workers were therefore told to interrupt or skip their breaks, or that it’ happened whenever a worker felt pressured to interrupt or miss a break. “Routinely” implies this happened as a matter of course whenever circumstances required it, but it does not speak to how often those circumstances arose. See, e.g., Sana v. Hawaiian Cruises, Ltd., 181 F.3d 1041, 1047 (9th Cir.1999) (construing meaning of “routinely” for purposes of Fed.R.Evid. 803(6) without regard to frequency). Nor does it explain whether breaks were interrupted or skipped altogether. Either of those would entitle workers to the same extra pay, but interruption of breaks suggests the alleged violations may have been more ad hoc and occasional, whereas requiring workers to skip breaks altogether suggests Pacer might have been so busy in general that workers had to skip breaks to keep up.

In other words, there is no basis for estimating that putative class members missed the required breaks twice a week. Clearly, the complaint is alleging that it happened repeatedly, but how often that was is undetermined because there is no evidence showing how often workers found themselves under pressure to skip breaks, or how often supervisors directed to do so. For instance, it could plausibly have been as infrequently, averaged over the whole class, as one missed break every two weeks.

The opposition cites to Mendoza’s record, showing that she missed a break 62% of the time. But, as discussed above, there is no basis for determining how similar other putative class members’ work habits were to Mendoza’s. Supposing, for example, that she was under significantly more pressure than most other workers, she might have missed breaks far more often than most. There is, in other words, no evidence supporting an inference that the rest of the class missed a similar number of breaks. They might have, or they might not have; no evidence nudges a conclusion either way.

Molina argues that Ray’s conservative estimate was one hour every two weeks, suggesting that the Court use this estimate. In fact, Ray’s estimate was based on an allegation of “consistent” overtime work, rather than “routinely” missed breaks. But the Ray court was attempting to engage in conservative estimation, and the Court believes that based on the complaint’s allegations here, one missed meal break and one missed rest break every two weeks (that is, one missed break of either type per week) is a conservative estimate.

As the briefing makes clear, California law entitles employees to an extra hour of pay for each day when an uninterrupted meal or rest break is not provided as required. The class members, conservatively, are claiming $15.00 multiplied by 49,640, or $744,600. This puts the total amount in controversy at $4,095,300.

To reach the jurisdictional threshold, Pacer would need to show Molina was claiming just over $900,000 more. But, as noted, no evidence at all was introduced regarding the amount of attorney’s fees claimed. Pacer’s pereentage-of-the-recovery estimate is based on class action settlements, which courts sometimes approve, where the fee award by agreement is capped at a percentage of the recovery. See Merchant v. OfficeTeam, 2013 WL 1153648 , at *7 (Cal.App. 2 Dist., Mar. 21, 2013) (discussing “clear sailing” provisions in class action settlement agreements). Pacer has introduced evidence showing that in other cases the settlement agreement has provided for a 25% to 30% attorney’s fee payment, and that class counsel have themselves sought the same percentage. If Molina had requested any particu *1068 lar amount in settlement, including attorney’s fees, that request would be relevant to establishing the amount of attorney’s fees sought. See Cohn v. Petsmart, Inc., 281 F.3d 837, 840 (9th Cir.2002) (“A settlement letter is relevant evidence of the amount in controversy if it appears to reflect a reasonable estimate of the plaintiffs claim.”) But settlement offers in other cases, which may or may not turn out to be similar to this one, are not particularly helpful. See Mireles v. Wells Fargo Bank, N.A., 845 F.Supp.2d 1034, 1055 (C.D.Cal., 2012) (“While settlements and jury verdicts in similar cases can provide evidence of the amount in controversy, the cases must be factually identical or, at a minimum, analogous to the case at issue.”)

Pacer cites various district court cases for the proposition that the Ninth Circuit has adopted a presumptively reasonable “benchmark” of 25%, (Opp’n at 17:6-18:2), and that this is acceptable even in the absence of evidence. (Id. at 18:3-8.) But even assuming the Ninth Circuit (rather than various district courts) had adopted this rule, it would be insufficient here, because fees are not recoverable for all claims. California statutes provide for an award of attorney’s fees for parties who prevail on actions for nonpayment of wages, failure to pay overtime, and violations of California’s Unfair Competition Law — but not failure to provide rest or meal breaks. See Kirby v. Immoos Fire Protection, Inc., 53 Cal.4th 1244, 1248 , 140 Cal.Rptr.3d 173 , 274 P.3d 1160 (2012); Franco v. Arakelian Enterprises, Inc., 149 Cal.Rptr.3d 530 , 559 (Cal.App. 2 Dist.2012) (reversed on other grounds as noted in — Cal.4th-, 176 Cal.Rptr.3d 265 , 331 P.3d 1274 (2014)) (extending Kirby’s reasoning regarding rest breaks to meal breaks). The fee award would therefore need to be reduced to account for the two claims for which a fee award is not available. See Staley v. Carlson, 2013 WL 6795404 , at *10-11 (CaLApp. 1 Dist., Dec. 23, 2013) (discussing apportionment of lodestar when fees are recoverable in connection with only some of them). It might be increased to account for other factors, but those have not been briefed. Thus far, the only possible recovery for which attorney’s fees could be awarded is the estimated $3,350,700 award for unpaid overtime. Even if that were increased by 25%, the threshold would not be met.

But the Court-is not convinced there is a presumption in all class actions that a request for attorney’s fees more likely than not establishes that plaintiffs are asking. for 25% of the total recovery (or even of that part of the recovery for which fees are awardable). Pacer cites only cases where the court has discretion to use either the percentage method, or the lodestar approach. See Trang v. Turbine Engine Components Technologies Corp., 2012 WL 6618854 , at *6 (C.D.Cal., Dec. 19, 2012). Here, if Molina prevails at trial, the Court cannot use the 25% “benchmark” which can be appropriate when the percentage method is used, see id., l but must use the lodestar approach. The only decisions that can be said to be on point are those where the estimate of attorney’s fees sought appears to have been undisputed.

In short, there is really no way to know what amount Molina is seeking in attorney’s fees. Clearly, it must be a significant amount, but beyond that it is unclear. It might exceed the $900,000 needed to get Pacer over the amount-in-controversy threshold, or it might not. In what appears to be a larger and more complex wage and hour class action than this, the lodestar figure was just over $725,000. See Amaral v. Cintas Corp. No. 2, 163 Cal.App.4th 1157, 1216 , 78 Cal.Rptr.3d 572 (Cal.App. 1 Dist.2008). In other cases, the figure has been much lower. See Knik v. *1069 Marriott Int'l., Inc., 2010 WL 2765106 , at *4 (Cal.App. 2 Dist., July 14, 2010) (approving trial court’s award of $93,161 for year-long litigation of wage and hour class action).

Thus far, the amount in controversy is not met. Pacer has, however, provided additional evidence not included in its no-, tice of removal, pertaining to the complaint’s seventh claim, for failure to provide accurate wage statements. This is not, as Molina argues, a new basis for jurisdiction; rather, it is new argument and new evidence in support of the same jurisdictional claim. The Court may properly consider such evidence. Cohn, 281 F.3d at 840 n. 1 (citing Willingham v. Morgan, 395 U.S. 402 , 407 n. 3, 89 S.Ct. 1813 , 23 L.Ed.2d 396 (1969)) (court may construe opposition to motion for remand as amendment to notice of removal, and may consider relevant evidence presented in the opposition).

Cal. Labor Code § 226 provides for an award of damages for each week for which an employee is not provided with a timely and accurate wage statement. This award is a minimum of $50 for the first pay period and $100 for each later pay period, not to exceed $4,000. Pacer has submitted evidence showing 232 putative class members are seeking the $4,000 maximum, which adds $928,000 to the amount in controversy, and another 77 class members worked 1,075 weeks, putting over $100,000 more in controversy. This does not account for the additional class members or the weeks they worked, as discussed in the supplemental briefing. Those additional workers would push the amount in controversy even higher. With this addition, the Court finds the amount in controversy is met.

Conclusion and Order

For these reasons, the Court finds the amount in controversy is met, and the Court can exercise diversity jurisdiction over this matter. The motion to remand is DENIED.

In making this ruling, the Court has accepted that putative class members are seeking $22.50 per hour for unpaid overtime. {See note 2, supra.) The assumption seems to have been that workers were paid nothing at all for any overtime hours they worked. If this is incorrect, any party knowing it to be incorrect (or both parties jointly) must promptly file a notice stating the actual hourly amount being sought.

IT IS SO ORDERED.

. Because Molina has not challenged the supplemental declaration, the slightly higher figure of 49,640 weeks is also acceptable.

. The claim is that workers were misclassified, and therefore paid nothing at all for extra hours. If it were the case that workers were paid at the normal rate, the amount in controversy for this claim would be $7.50 per hour.