The flagged page
Grading report
81 exchanges read · 16 moments flagged · outline checked against 4 topics
8
answers that stopped being answers
Outline coverage: 4 of 4 topics actually covered (100%)
Search engineering headcount and investment gap with Google
Mobile distribution deals (Verizon, RIM, Nokia) and their outcomes
User switching from Bing defaults to Google
Microsoft resources and capital investment in Bing
Moments worth a second look
Ranked: answers that stopped being answers first, then the long narratives, then the evasions.
Q. A habit that was built on having the leading, best product in the market, correct?A. You know, look, I look at it and say, I'm focused on building a product against Google which is all that and more, as you said. And so, therefore, in order to compete with it, you have to compete for distribution, you've got to put up with press like this, and then get to the other side with a better product ultimately.
Non-answer62 words that never touch the question. This is where the follow-up lived.
Q. Okay. The -- and you can put this up. The very first sentence said, "A Google phone handcuffed to Microsoft's Bing -- at Bing's search engine sounds like a bad joke. But it's a real product, one for which Verizon Wireless charges 199.99. You were aware at this time -- you were overseeing Bing search. You were aware that these Verizon devices that came with Bing as the exclusive preloaded search engine were encountering bad tech reviews like this, weren't you?A. Yeah, we saw -- saw a lot of this. In fact, a lot of this even informed a lot of my conversation even subsequently with even Apple. Knowing that, look, I think you need to have a good product, and more importantly, you need to be able to get through this bad-press period and turbulence and to get to the other side, because in some sense, you have to persist to change. And you sort of mentioned it very well, when Google's such a user habit that is well ingrained, that there will be a little bit of, both from the intelligentsia to the users, a lot of change, management that you have to persist through.
Non-answer116 words that never touch the question. This is where the follow-up lived.
Q. Now, various of these Verizon devices that were preloaded with Bing were panned by various tech reviewers, right?A. I don't have all of the details but, yeah, I mean, these were very different times, even. It was sort of -- IPhone was just coming out, I would imagine. And what is mobile search today was -- is what mobile search was in that period as a habit. And even the browsers and everything else, you know, the speeds, were, I think what, 2G, 3G, so it was a different world, but, yes.
Non-answer72 words that never touch the question. This is where the follow-up lived.
Q. Okay. To the extent you've got issues with distribution, it's on mobile, right?A. That's correct. And also on desktop. I'm grounded on the fact that, you know, we have low share of our browser where Google Chrome has. But, yes, to your fundamental point about we have the ability to set defaults on desktop is -- is accurate.
Non-answer44 words that never touch the question. This is where the follow-up lived.
Q. And you don't have any reason to doubt the accuracy of these numbers, do you?A. Yeah, as represented here, absolutely not. And I think the one thing that all these estimates are, this is the hard part about, yeah, how do you separate out what is the YouTube investment, what's the Google Search investment. So our teams, you know, obviously, to me, to make the arguments that I should spend more would make the argument of picking the largest number, and it's okay.
Non-answer68 words that never touch the question. This is where the follow-up lived.
Q. The forecast that Microsoft was making during the time period that you were overseeing search, this '7 to '11 period, Microsoft didn't have enough servers to support even the forecasted growth you were hoping to get, correct?A. You know, each -- I felt, you know, we definitely were pushing for more and more investment, but I felt well supported and well funded to keep continuing to improve the product and compete. But, yes, I'm sure there's lots and lots of emails where I'm asking for more, I'm asking for -- advocating for our team to get the investment required.
Non-answer60 words that never touch the question. This is where the follow-up lived.
Q. Do have an estimate today as to the disparity between search engineering resources between Bing and Google?A. I don't. The one thing, as I look at all these numbers again, is we've gotten much better at getting, again, focused on -- you know, like it is brought up on international markets, for example. We prioritized our U.S. market, so in some sense we sort of said, let's hunker down on things we can really make a difference in and then concentrate our head count forces. So I don't have an estimate as to exactly how we match up; and at this point, in fact, with all the LLM investments and so on, it becomes even harder to estimate.
Non-answer101 words that never touch the question. This is where the follow-up lived.
Q. And that was accurate in November of 2007, correct?A. Yeah, I mean, it -- just to be very transparent, I mean, we have a culture of a lot of self-critical stuff. And so, yes, I do feel like, you know, I would say even that today, we need to do more work, and we are behind Google. So in some sense, that's the spirit of sort of me trying to get into my new job and work it.
Non-answer69 words that never touch the question. This is where the follow-up lived.
Q. Yes, sir.A. I don't recall it, but I'm sure we did.
Evasive answer"I don't recall" — the door to a clean admission was open here.
Q. Sir -- and if you look in the last sentence there, he calls out Milan. Who is Milan? It seems like it's the person who's cc'ed on the document.A. Yeah, I don't recall what Milan did.
Evasive answer"I don't recall" — the door to a clean admission was open here.
Q. Well, they switched from Bing to Google on their BlackBerry devices even though Bing was the default, right?A. If you have Google as the most dominant search engine that every user in the world knows about, anytime you introduce anything, there will be switchers. And you're right in saying that its onus is on us to build a product that retains more, and we'll keep trying until we sort of break through on it.
Long narrative56 words in one run. Rabbit trails like this are where outline coverage quietly dies.
Q. Microsoft's failure to make capital expenditures to support its search business wasn't because Microsoft lacked the financial resources to do it, right?A. The distinction that needs to be brought is two sets of things. What is the fixed cost capital expense and what is the cost of capital for serving the traffic. So one is what's the difference in scale, which you can use share as a proxy for it. And then the second is the fixed cost. And I felt that we were making, and over the years have made, and continue to make, all the fixed cost investments, with tradeoffs; we pick countries, we pick index size, we pick a particular angle. And then the variable expense comes with more traffic and share gains.
Long narrative104 words in one run. Rabbit trails like this are where outline coverage quietly dies.
Q. Well, did Microsoft have enough servers to support the search business it was forecasting?A. Oh, got it. So our capital investment in support of serving Bing traffic. Yeah, I mean, there are two sets of things even there, which is the fixed cost and then there's the variable cost. The variable one is based on, obviously, your search share in traffic, and then the fixed one is what you need to play. And in all these cases, we continuously keep having the argument about how to increase fixed cost and then how to be ready for any search gains -- or share gains.
Long narrative88 words in one run. Rabbit trails like this are where outline coverage quietly dies.
Q. This is the -- give us the distribution first and then we'll invest afterwards, right?A. Or coincident with it. So it's not, give us distribution, it's co-own distribution and use that as -- because you sort of need -- the only reason I bring up distribution is not that we just want somebody to hand us distribution for free. We are talking about pay for distribution, which then gives us some long-term guarantee that we will have traffic. And so, therefore, it makes economic rationale for us to invest.
Long narrative74 words in one run. Rabbit trails like this are where outline coverage quietly dies.
Q. You don't have any reason to believe that Bing had closed the gap on head count with Google on search engineering, do you?A. I don't -- I don't think so. In fact, the thing that we are trying to make sure is in a very focused way in core search be competitive in few segments with the option that get -- once we have more distribution, we can go accelerate other areas.
Long narrative47 words in one run. Rabbit trails like this are where outline coverage quietly dies.
Q. And, in fact, over the years, this head count gap, search engine years has only gotten worse for Microsoft, right?A. Yeah. I mean, just to make sure, the spirit of the way these head count things is, it's just our estimate. In fact, one of the big things I did was I shut down things like our book search and so on, because I sort of focused our energy with the head count we had on things that mattered, because you can be investing on a lot of things and then things that really, really matter and needed moving things. So, yeah, like, I have head count games inside large organizations like ours, I was playing it.
Long narrative97 words in one run. Rabbit trails like this are where outline coverage quietly dies.
How this was graded: the transcript excerpts above are public record — the official court-released trial transcript of day 14 of
U.S. v. Google (D.D.C., October 2, 2023), document 954 on the public docket, released by the court reporter in 2024 and mirrored by the Free Law Project's RECAP archive. We took one continuous stretch of the cross-examination (81 exchanges) and joined the Q&A pairs; the flags, counts, and coverage below are the unedited output of the Deposition Grader — a free, browser-based tool that reads Q&A text for the patterns examiners miss in the room. No human at Demur selected or edited these flags; the same input produces the same output for anyone, at
grader.trydemur.com. Also in this series:
Graded: the Pichai cross ·
Graded: the Bill Gates deposition ·
Graded: the Sackler deposition ·
Graded: the Eddy Cue examination.
Your deposition next
Run your last transcript through the same grading. Then imagine it live.
Free and private — everything runs in your browser, so nothing is uploaded. The Grader is the 10% version: Demur does this live, while the witness can still answer — against your outline and the record, every flag with its source page.