Dear Adrian, David,

Physics:
are 3.05-1.35-1.27=0.43 supposed to be identical to the 0.40 from paper 2012-050?
If so, mention that the result of the previous paper is consistent with this one.

L.248: This value is from the printed copy of the PDG (which is also what you reference), but the PDG live now lists 0.124 \pm 0.011 http://pdglive.lbl.gov/DataBlock.action?node=S086DRA, which again is not quite the same as the HFAG values (http://www.slac.stanford.edu/xorg/hfag/osc/summer_2015/). Has this been discussed in the review?

General: You use a lot of jargon, especially wrt particle names. Please stick to the PDG. D** is not defined, please do so. Please use the correct name D_{s1}(2460)^\pm.

Also, you never give uncertainties except in the systematic section. We think values should always be given with an uncertainty, which can then be commented on in the systematics section. For example, see Eq. L.72, 113.

Abstract, Table 5 and Summary are not consistent in the notation of the three measured branching fractions.
Explain also why B0->D*\pm_s D\mp with both charge combinations and B0s->D*+_s D*-_s with only one charge combination are written.
It is also not immediately clear if in B0s->D(*)+_s D(*)-_s both D_s have to be in the same state or not and if the superscript * refers to the first excited state or to any excited state.
Please, define the notations mentioned above clearly in the introduction.

The title mentions one inclusive branching fraction, while we measure three exclusive branching fractions and add them up. Do this justify calling this decay inclusive?

Introduction : It is not clear where equation (1) comes from. It is not directly included in Ref. [1], so how is it determined? Also, doesn’t this formula require HQE? 
As the rest of the paragraph indicates that this formula is not expected to hold, maybe it is possible to mention the estimates from theory as determined in Ref. [3], and to mention that measuring this branching fraction with increased precision will improve estimates on the CP-even contributions to B0s decays. 

In addition, especially examining the contribution of D_{s1}(2460)^\pm in the DsDs fit, is it possible that higher excited states of the Ds meson than Ds* contribute to the relative decay width difference? Please mention this.

It is not clear what exactly you measure. Sometimes there's mention of 2 channels plus one inclusive, sometimes three plus one. Please add a paragraph at the end of the introduction on what you measure in this paper. L.60-64 can be used as starting point.

Signal selection: Why is B0->D\pm D\mp not used as normalisation channel?

For Table 1: It is unclear what you mean by initial selection and Rec. X Trig. From the analysis note, it seems that initial selection is the stripping efficiency, while Rec. X Trig. is a combination of the generator level efficiency and trigger efficiency. However, this cut on the trigger is only applied after all other cuts have been included, according to the analysis note. Ideally, this cut would be applied immediately after the stripping selection, to get the table in line with the analysis. At the least, could Rec. be renamed to Acc. and interchanged with Initial Selection?

In addition, the loss in efficiency due to background vetoes seems quite high. Is it possible to apply a PID cut as well for these background vetoes, such that less signal is cuts away?

Signal and background shapes: According to this section, in the final fit only (some of) the shape parameters of the combinatorial background and the Bs->DsDs signal are floated, while MC is used to fix all of the others. Would it be possible to apply Gaussian constraints, using parameter values and uncertainties from MC? Additionally, with more data it might be possible to float more of these parameters. Thus, both trying to increase the efficiency and using the full dataset might lead to a large improvement of the fit model.

Systematics: As most of the parameters have been fixed in the fit used to obtain the yields, a lot of systematic checks have to be included. It seems like only one of these checks (for Ds(2460) background) includes extra floating parameters. Additionally, just comparing the fit result on data with different signal or background shapes does not probe the bias introduced by a model. Could you include toy studies to estimate the fit model uncertainties?


Line by line
Abstract: remove "using the ... B(B->DD)"
Line 5: Replace “(Gamma_s) is the difference (average) of the decay widths" by "and Gamma_s are the difference of the decay widths and their average, respectively".
L.6: Replace "of these quantities" by "of this quantity" (the relative decay width difference).
L.7: A justification on why we want to determine these quantities (NP?) would be welcome.
L.8: Explain why one would expect these decays to be CP even. That seems to be a very unnatural hypothesis at first sight.
L.10: what is "this value"? Above is an equation with many values. What do you mean with direct determination?
L.16: in Ref. [3].
L.18: how is b->ccs an FCNC? Maybe just mention that the ~3% of B0s->D(*)+_s D(*)-_s is a large contribution to the total inclusive b->c\bar{c}s branching fraction.
L.44,49: \gevc
L.59: signal selection
L.60. correspond -> corresponds, why is not 3fb-1 used?
L.66: clarify charge conjugation at earlier point in paper, analysis -> paper
Eq.2: rel should be in roman, but we presume this is channel-dependent.
L.72: give uncertainties and cite PDG.
L.75: ... mass of the Ds+Ds- system.
L.84: charmed ?
L.88: remove "fully". background events -> backgrounds
L.90: ... to the signal. Candidates ..
L.93: B^0_{(S)}
L.112: data with D0->K-pi+
Table 1: cuts -> requirements (twice). Why no uncertainties? Why different number of digits across columns.
L.116: why three channels?
L.118: Say in introduction what you do with Bs->DsDs
Table 2: move to bottom of page. Rel should be in roman.
L.123: extract -> determine (also 163)
L.132-133: We fail to understand this sentence. From fitting what to what? What does "in the full data fit" mean?
L.139-143: Too long sentence. We suggest : "The wrong sign ... looks for Bs->Ds+Ds+ decays, which emulate the pairing of genuine or fake \Ds mesons."
L.146: Ds** is undefined. We often write "where Ds** stands for any Ds resonance of mass higher than that of Ds* and decaying to a Ds meson", or similar.
L.155-159: This sentence is too long an unclear. Just say the model is justified by simulation.
L.165: what does "full" mean here? Same in Fig.1.
Table 3: The yields seem a bit squeezed. Is that the usual LaTeX spacing?
L.169-172: Remove that sentence. It conveys no information beyond what is in the peak.
L.174: This sentence is very unclear (what comes from Ref [23]?). Add uncertainties. from Ref. [23].
Figure 2: where does the green cat-head-shaped distribution come from? It's not clearly described in the text.
L.189-195: Are remeasure and remodel English words?
L.200: the knowledge of what of Bs->Ds**D? BF?
Table 5: use PDG name, where are the fit uncertainties for B0->DsD decays taken into account?
L.241: from Ref. [3].
Fig.3: Please rotate by 90^0 as is usual for such plots. Also, make the theory marker different. from Ref. [3]
L.246: \cal missing
L.256: what do you mean with clean?
[5-6] collaboration
[17] The template now says "and 2015 update", but that depends on what value of DGs/Gs you use.
[23] The template suggests to also cite the paper.

Cheers,

Patrick