All questions have Show Video links that will fast forward to the appropriate place in the video.
Transcript Of The Office Hours Hangout
Click on any line of text to go to that point in the video
JOHN MUELLER: OK,
welcome, everyone,to today's Google Webmaster
Central Office Hours Hangout.My name is John Mueller.I am a webmaster trends analyst
at Google in Switzerland,and I try to connect webmasters,
publishers like you all,with our engineers and
make sure that informationis flowing in both directions.So we have a bunch of questions
that were already submitted,but if any one of you
who's in the Hangout nowwants to go ahead
and jump in and aska first question
or two, go ahead.
BARUCH LABUNSKI: I have
question about 301.
JOHN MUELLER: All night.
BARUCH LABUNSKI: So we all
know everything about 301sand so on, but I just
really wanted to ask you,I made a really big
change to this page,and I 301'd everything
correctly in the entire site,so all the URLs,
everything's fixed.So now I fetched Googlebot,
and it took over two weeks,so the URL is
finally showing up.But how long would it
take for all the juiceto pass from the old
page to the new one?And I know you've touched
on that many times,and I just wanted
you to clarify that.If it's possible.
JOHN MUELLER: So that's just
one page or a whole site?
BARUCH LABUNSKI: Yeah.Just a specific page.There's a lot of alteration done
to the page, and, again, notfor the search engine,
but for the user only.Like you guys wanted.
JOHN MUELLER: Yeah.So I guess that's something
where at least we'dhave to crawl it a few times
to kind of at least noticethat it changed.And after that, some of
our algorithms are fasterand some of them are
a little bit sloweron picking up those changes.So I imagine, for
the most part, you'dsee those changes within
a couple of days of us wecrawling those pages.
BARUCH LABUNSKI: OK.Because I see certain queries
show up-- the old pageis still showing up--
and then the new pageis showing up for
different queries.Why is that?
JOHN MUELLER: I
think that's probablyjust the way the
algorithms are picking upon some of the old things
and some of the new things.So that's kind of something
where users will make itto the final page anyway,
so that'll still be around.So I wouldn't really
worry too much about that.I mean, some of these things get
picked up a little bit faster,some of them take a
little bit longer,it's not that it's all
just, like, one crawland then suddenly everything's
over on the new URL.
But is it normalfor it to go down in rankings
and then go back to itself?
JOHN MUELLER: Usually
for individual pagesyou wouldn't see
big ranking changesjust if you're redirecting.If you're significantly
changing the contentof those pages, then of course
you'll see some fluctuations.Some of the things we'll be
able to pick up on a little bitfaster.Other things, they
kind of take awhileto understand the new page that
you're essentially showing.So if it has different links,
if the rest of the sitehas changed as well, all
of these things kind ofcome together, and sometimes
it takes a while for usto settle back down
with the previous state.
BARUCH LABUNSKI: OK.Thank you very much, John.
JOHN MUELLER: Sure.
MALE SPEAKER: Hey, John, quick
question about rel="canonical."
JOHN MUELLER: OK.
MALE SPEAKER: Let's assume we
have an ecommerce website whereproducts have different
variations based color.For example, so let's
say an eyeglasses store,and the frames themselves
have different colorsand there's a product URL
for each of these colors.Let's say there are three or
four variations per product.Would it be wise to
do a rel="canonical"given that most of the products
aren't really different exceptthat color?And would it be wise
to do a rel="canonical"to a single product or would
that be overkill and Googlewould decide what URL to give
to the users based on their suchword?
JOHN MUELLER: So what would
happen with a rel="canonical"is that we would essentially
drop or ignore those othervariations.So if you set the
rel="canonical"of your red page
to the blue page,then we won't really be able
to rank that page for queriesaround red sunglasses,
for example.So that's something where
if you're taking relatedbut slightly unique pages and
rel="canonical"-ing them backto one specific page, then we'll
kind of lose that informationthat makes that page unique.So sometimes that
might not matter.Maybe you have one
page that's, like,for this pair of sunglasses
in all variations,and that's your main page
for this specific modelof sunglasses.Maybe that's fine.Maybe it lists the
different colors;that could be a perfect
reason to kind of set that uplike that.But on the other hand, if you
have just one color and youhave different colors that
are actually kind of uniqueat the same level, then that's
something where I probablywouldn't use a rel="canonical,"
because those pages are uniqueby themselves.It would make sense
for us to sendpeople who are looking for
red sunglasses of that modelto that specific red page
and not to the blue page.
MALE SPEAKER: All right.So basically I should
think about whatthe users are
actually looking for.If they aren't really interested
in searching for actual colorvariation, then do
a rel="canonical."But if they are, then the fact
that all these variations havethe same product
description and price,for example, doesn't
really affect the websitefrom a negative
point of view, right?
JOHN MUELLER: Yeah.Exactly.Yeah.I mean, it's something
where you have to figure outdo you want these pages
to rank separately?Is there something unique on
these pages that make senseto show them in the
search results like that?Or are these things that you
can fold together and say,this is actually my
main page, and thisis the one I want ranking with
the content that's correctlyon it.
MALE SPEAKER: OK.That makes sense.Thanks.
JOHN MUELLER: Sure.All right, let's grab a--
GREG: About cycling
submitted for a second.
JOHN MUELLER: Go ahead.
GREG: When a website is
displaying site links,is that any indicator
of authority or trust?
JOHN MUELLER: Not directly.So I imagine to some
extent we want to show themfor pages where we
know it makes sense.So you could say that
something like trustis kind of flowing into
that, but primarily we'reable to understand the
structure of the site.We're able to understand
that if someone is searchingfor this specific brand
or this specific item,that maybe they go
to different pagesas well in the search results.So we'll kind of bubble that up
that way in the search results.So it's something
where we probablywouldn't show it for
any random website,but passed that it's
not a sign that you'vepassed some kind of
a magical thresholdand now your website
is good enoughto be shown with site links.It's primarily something
where we understand the sitestructure and know that
these other pages make sensefor the user as well, so
we'll show them as site links.
JOHN MUELLER: All right.We have a question
here, if a lot of usersare bouncing back to
Google from a website usingthe back button, isn't
that a strong indicatorof the quality of that
URL for a search term?Why would Google not use that
as a metric to rank a page?This is, I guess,
a tricky question,because to a big extent there
are lots of reasons why peoplego back and forth in
their search results,and it's not something
where we couldsay on a per-page level this
is an indicator of this beinga high-quality page
or low-quality page.Sometimes pages
have content on themthat users can read within
a couple of seconds,and they go back and
do something else.And that's perfectly fine.That's not a sign
that it's a bad page.Contrarily, the user was able to
get the information they wantedreally quickly, and
they're happy with that.We do use some kind of looking
at which results people areclicking on when we evaluate
our algorithms overall though.So if we're looking at specific
algorithm changes, and we say,well, people are clicking on
lower search results in generalacross this algorithm,
across millionsof queries that were
made, then maybe that'sa sign that the algorithm isn't
doing what it should be doing.That it's kind of
encouraging peopleto not find the results
that they are actuallylooking for in the
first couple of places.So on a very broad-level
it makes senseto use that information.On a very page-level,
focus-basis,it's a very noisy
signal that I thinkprobably wouldn't make
that much sense to keepand to kind of
track for ranking.It's also something that's
very easily gameable.So I don't think
it's something wherewe would say this is a
useful signal by itself.But across millions of
queries on an algorithm basis,that's probably something that
kind of settles down and givessome kind of useful information.
NEMANJA DIVJAK: can ask how
important is structured datathrough the web master tools?
JOHN MUELLER: For ranking,
we don't use structured data.So that's something where we use
it primarily for rich snippets.If we can pull out more
information about your pagesand we can bubble that
up as rich snippet,that's a great thing to do.That gives the user a little bit
more information about the pageand kind of tells him that this
is the page you're looking for.So from that point of view, from
a kind of a clickthrough ratepoint of view, that's something
that I think definitelymakes sense.We don't use it for
ranking, though.So if you have things
marked up on your page,we might be able to understand
the page a little bit better.But it's not that we'd
be able to say, well,this has structured data.This doesn't have
structured data.Therefore, this must
be a better page.This is something that's
kind of a technical elementon the page, and from
that point of view,I'd be cautious to say
this is an indicatorof a high-quality site.Because a lot of times
we'll see spammerscreate technically really
high-quality sites,because they have
a lot of experiencecreating lots and lots of sites.So their sites will be
technically perfect.And maybe they'll scrape
content from other sites;they'll markup the
structured data.But just because it's a
technically well-made websiteand uses structured
data, doesn'tmean that it's necessarily
a better website.
BARUCH LABUNSKI: These spammers.Ugh.
JOHN MUELLER: Well, I mean,
it's just one example.It's something that comes
up every now and then wherea mom and pop website will kind
of have this messy front pagewebsite that from a technical
point of view you'd look at itand say, oh, gosh, how can
this even show up in a browser?But from a practical
point of view,from a search quality
side, it's somethingthat we do want to show
in the search resultseven if it's kind of messy
code, if it's messy HTML,if it's an exported Word
page or something like that.That doesn't necessarily make it
a lower quality search result.All right.You mentioned that
Google can pick upon signals of
people recommendingyour website to other
people very directly.Can you elaborate on what
these signals might be?Social shares?Bookmarks?Et cetera?I'm not really sure
what specifically you'repointing at that I
might have said there,but I think one
thing to keep in mindis we don't use social
signals for search.So things like likes or
+1s, those kind of things,aren't necessarily something
that we'd be able to usefor search, because,
for the most part,we don't have that information.So that's something
that we can't reallyuse if we don't really have it.So from that point of view,
those kind of social signalskind of get lost.Anything else that essentially
looks like a web pagethat we can crawl,
that has links on it,that's something we can
treat as a web page,and we can kind of
pick up and reuse.
BARRY: John, with the exception
of personalized search?
Personalized searchis, I guess,
something different,because that's personalized.Like you said, that's
something wherewe do take into account who
your specific friends areor your circles or however
that kind of ties in.But that only works for the
information that we do have,and that's specific
to individual users.So that's not
something where if youhave a lot of friends
that are posting content,that doesn't mean
that that content willbe relevant to people who aren't
your friends, who aren't kindof in your circles or
have you in their circles.
JOHN MUELLER: All right.XML sitemaps.If the taxonomy overall
site structure is good,is there any need
for XML sitemaps?I've seen a lot of
examples of large siteswithout XML sitemap files
doing great for crawlingand indexing, et cetera.You absolutely don't
need XML sitemaps.So it's not a requirement.It's not that your website will
disappear from Google searchif you don't have one.The XML sitemaps help us a
lot with quickly changingcontent-- so content that
changes, content that getsadded to your pages--
because we don'thave to crawl your website to
realize that these pages arenew and have been updated.So if you have a blog, if
you have a news website,if you have a shop that has
maybe Christmas specials,those kind of things where you
want to new and updated contentpicked up as
quickly as possible,then that sitemap
is going to helpus to get there as
fast as possible.On the other hand, if you
have content you put up once,you keep it like
that for 10 years,and you're not going to change
it in the next 10 years,then it doesn't really matter
if you have a sitemap or not.We have that content already.We don't need to kind of
recrawl it every five minutes.We'll be able to kind
of figure out whatcycle make sense for us there.And if you happen to
change something there,it won't be picked
up immediately,but it'll be picked
up sooner or later.
LYLE ROMER: Hey, John,
real quick questionabout something we noticed
in Webmaster Tools recently.We were looking through
the Content Keywordsand noticed that for some reason
the word "logo" had come upvery high in the list there.And we discovered that
in a lot of our Alttext around the site where
we have logos of things,we would call it
"whatever logo."We were just wondering,
is that somethingwe should be concerned about?We went and redid the Alt
text to take them out,but in general should
we be concernedabout seeing
something like that?
JOHN MUELLER: No.So the Content Keywords
feature looks at the contentthat we see when
we crawl the pages,and we'll pick up
on things like logo.Or if we happen
to crawl and indexyour robots.txt
file or your sitemapand kind of treat it
as a normal HTML page,then we'll pick up
things like HTTP or WWWor whatever might be in there.So it's basically just a
signal that we found this textwhile we crawled your pages.It doesn't mean that it's
relevant for your website.It doesn't mean that it's
relevant for ranking.It's not that your
website is goingto rank for the term "logo"
or anything like that.It's just we found
the text; we thoughtwe'd let you know about that so
you're kind of aware of that.The main reason I'd
use that featureis to make sure that there's not
something completely irrelevantthat kind of pops up.So if your site is about cars,
and suddenly you see thingsabout casinos in
there, then chancesare someone has been
spamming your siteor hacked your site or
something like that.So that's a really good sign
that something really weirdis happening with your
website, and that'sworth kind of digging
into and figuring outwhat happened there.But if this is
based on somethingthat you have on your pages,
like the Alt text for your logoor when we crawl a sitemap
or robots.txt file,whatever, it happens.It's not going to change
anything for ranking.It's not going to
change anythingfor the rest of the words
that were found on your site.It's just, we found
this on your site.We thought we'd kind
of let you know about.
LYLE ROMER: OK, great.Thanks.
JOHN MUELLER: OK.Webmaster Tools and HTTPS.What's the difference
between HTTP and HTTPSin Webmaster Tools
Verified Properties?So for some of the
features in Webmaster Toolswe differentiate between
HTTP and HTTPS versions,and we'll show you
the data separately.So specifically, I think
this is around Index Status,this is around the
Search Queries feature,where if you're moving from
one variation to the other;you'll see the one
variation kind of go down,the other one kind
of go up again.And we primarily do
that separately sothat you can look
at these versionsseparately and figure out if
anything weird is happeningwith your transition, if
everything is being picked upproperly, if the search
queries are picking upover again, those
kind of things.Also, as with everything
else in Webmaster Tools,this doesn't affect
your rankingsif you have those
variations verified or not.But if you're doing a move
from one version to the other,I'd definitely make sure
that you have both of themin Webmaster Tools so that
you can kind of double checkthat technically
things are workingthe way they should be working.Another sitemaps question.Are there any advantages in
splitting up the XML sitemapinto smaller files, like 5,000
or 10,000 items per file?You can definitely do
that if you want to.What will probably
happen there is wewill have to
recrawl more sitemapfiles to pick up the changes.Usually that's not a
big load on your server.These are just text
files if you're notgenerating them dynamically.The advantage I see there
with smaller sitemap filesis, specifically
when you're lookingat the indexing of
those URLs, it'sa lot easier to kind of
pinpoint individual issues.So if you split your sitemap
file in a logical structure,like category pages, articles,
blog posts, products,something like that, you'll be
able to look at the IndexingStatus of each
individual sitemap fileand see, oh, for this type
of page, most of my pagesare being indexed.For this other type of
page, maybe half of my pagesare being indexed.And by knowing
which type of pageis indexed how well,
you can figure outis this a critical problem
or is this essentially fine.So if we're able to
pick up, let's say,half of your product
pages, but wecan pick up all of
your category pages,then probably your site
will still rank normallyand things will work
well in Search anyway.On the other hand, if
we're not able to crawlany of your articles or
any of your category pages,and your product pages
are being picked up,then that's a sign that
maybe something technicallyis wrong with the way your
category pages are set upor the way they're
maybe linked internally.Something like that.So by splitting it up, it's
usually a little bit easierto diagnose problems.If you don't have any
problems at all with yousitemap files, than
it doesn't reallygive you any advantage in
splitting them up or keepingthem be.
GREG: John, if I can jump in
real quick with a questionabout, not penguin per say,
but disavowing in general?
JOHN MUELLER: OK.
GREG: A lot of talk
about how great itwould be if Google could offer
some sort of notificationon when we download the latest
links in Webmaster Toolsto kind of identify which ones
have already been disavowedwhen we update the file.Is there any possibility of
that coming out in the future?It would make our jobs as
webmasters so much easier.
JOHN MUELLER: So if you would
see within the download whichones are kind of blocked by
your disavow file already?
JOHN MUELLER: I don't see
that happening any time soon,but it's an interesting idea.I think it's also something
that probably someone could makea simple Tool for, because
taking a disavow fileand comparing it
to a links downloadshouldn't be that complicated.
GREG: Sort of like
matching a one-to-one?
JOHN MUELLER: Yeah,
because I meanyour domains are things
that are easy to match.The individual URLs
are easy to match.It shouldn't be
that complicated.So if someone is looking
for a Christmas project,that might be
something to try out.
GREG: I'm going to hire Barry.
JOHN MUELLER: I don't know.Maybe if I have
time in between I'llput together a simple script
to kind of do that for you.
GREG: That'd be great.
JOHN MUELLER: But I
don't see that happeningin Webmaster Tools anytime soon.They're working on a variety
of features, and at the momentthe Links feature
isn't somethingthey're spending
too much time on.So I don't see them kind
of jumping in and tweakingthat just for that feature.
JOHN MUELLER: OK.Go ahead, Barry.
BARRY: Are you sure?OK.All right.Two questions.Two questions; one's fast,
one's kind of longer.So you said, John, in the past
that obviously hidden or tabcontents may not be
fully indexed or rankedin some sense; I'm not going
to go through all the details.But there was an argument
on Friday's Hangoutbetween webmasters,
because you said alsothat for many mobile
ranking factorsyou just look at
the desktop version.So when you change the CSS style
sheets to hide some navigationor content to make it
more mobile friendlyon a responsive
design, are you lookingat that on the mobile
side for hidden content,or you're still looking
at the desktop versionof the hidden content?That make sense?
JOHN MUELLER: We're looking
at the desktop version.That's the canonical
version, so if youhave the connection
between the two pages,if they're separate
pages, then we'dbe looking at the
desktop version,and that's the one we'd
be using for indexing.So if it's visible on
the desktop version,and you're kind of
folding it away,simplify it on the
mobile version, thenthat's absolutely fine.If we run across
situations wherewe see that people
are abusing this,that they're kind of, I
don't, showing cartoonson the desktop site and
showing adult contenton the mobile site,
then that's somethingwhere we might have to kind
of reconsider that and seewhat we have to do there.But for the most part, I
don't see that making sensethat you'd kind
of show somethingsignificantly different
on the mobile sitethan on the desktop
site, because thenon the mobile site, people
would be kind of lost.Like, what should
they be doing there?
BARRY: Cool.Thank you.And a very question,
blink twice if it's true,once if it's not true,
did you make an offerto Duane Forrester
to work on Google?
JOHN MUELLER: I think
he's working at Bing now.
BARRY: I know.I just wanted to know
between that time.
JOHN MUELLER: What do you mean?
BARRY: OK, I got you.Thank you.
JOHN MUELLER: No, I
didn't-- I got lost.OK.
BARRY: I'm just with you.
JOHN MUELLER: Oh.So if we made an offer
for Duane, I don't know.I didn't make anything
so-- it soundslike he got a lot of different
offers, but he's a good guy.I'm sure he'll find
something really interestingat Bing to do there so--
John, going furtheron Greg's questions regarding
Webmaster Group Features,if I can add a few suggestions
that I've noticed recentlythat would help me
manage my websites betterwith Webmaster Tools.Is it possible to add a
filter or something like thatso I can select
both web and mobile?For example, when I'm
looking at search queries,I want to see such queries
for both web and mobile.So for search and exclude
images, for example,because right now you
can search [INAUDIBLE]for web or for mobile, but
not both at the same time.Sure, you can see all, but
that would include alsoimages and everything
else that is there.And if it's also possible to
maybe add a local filter so wecan see queries that
show local results.Maybe that's possible?
JOHN MUELLER: OK.Let me just jot this down.
MALE SPEAKER: Maybe if I can?Instead of just
having a drop downwhere I can select Web
or Mobile for images,maybe just check boxes
so I can select multiple?Web and Mobile, but not Images?Also, do you know
if there's goingto be any better integration
with Google Analytics?There hasn't been any
changes with that.
JOHN MUELLER: I can
definitely pass the searchquery things along,
because I knowthey're working on
some fun stuff there.I don't know how far
they are with regardsto how much your
feedback will stillbe able to make it in there, but
I'm happy to pass that along.With regards to
Analytics, I'm notaware of anything
specific happening there.I know they still kind of
have the bucketed numbersin Analytics, so
that's somethingthat's definitely still open.I imagine that will get
fixed at some point,but I don't really know what
the Analytic side of thingsis planning.
MALE SPEAKER: OK.And is it possible
that we SCOs wouldhave like something
along the MCC accountthat PPC managers
have where they havemultiple clients
in the same accountand they can just ask a
website for permissionto handle their-- because when
we're getting new clients whoaren't really tech savvy, they
king of don't understand howto give access and that to users
and even create their webmasterto those accounts.So that's a lot of time
wasted at the beginning.Maybe there is going to be
better integration like the PPCpeople have with
the client center?
JOHN MUELLER: I don't
know how that actuallyworks on the outward side,
but in Webmaster Toolsyou can already delegate access.
BARUCH LABUNSKI: Yeah.
Maybe it's just toocomplicated from
what I hear from you.
MALE SPEAKER: By
delegate access youmean just add another user or?
JOHN MUELLER: Yeah.Yeah.You can add a
read-only user if youwanted to to kind of let
them look at Webmaster Toolswithout making changes.
MALE SPEAKER: Yeah, well, so
in the client center for PPCso you actually have an option
to actually request accessto a certain website without
actually telling that personcould you go ahead
and click that buttonand add this email and so forth.So actually, just
request an invitationand the user receives an
email or in the accounta notification, and he
just accepts or denies it.
BARUCH LABUNSKI: Yeah.It's basically like the same
kind of system as the My CCAccount, you know?I don't know.I mean, clients, basically,
I tell them where to go.They add me, and you
become an administratoror a read-only access
like you were saying.So, yeah, I mean--
[INAUDIBLE] just making itthe same like an easier way
like with the MCC, you know?
JOHN MUELLER: I have no idea how
that works with outward, so--
BARUCH LABUNSKI: I figured.
JOHN MUELLER: I will
take your word for itthat it's a lot easier.I'm not sure if
requesting accesswould be that much
of a good idea,because it obviously
spams the owner.But maybe there's a way to
do that in a reasonable way.So I'll definitely take
that up with the teamthat handles the verification.It sounds like an
BARUCH LABUNSKI: But, I mean,
it's fine the way it is now.Like--
JOSHUA BERG: John?Could we see some grouping of
the different domain types?Or is that in the
works by any chance?
JOHN MUELLER: I know that's
a commonly asked question,so I'm sure they're kind
of aware of this issue,but I don't know of any plans
of that changing immediately.So I imagine at some point
that'll kind of work outin a way that we can kind of
have these sites a little bitmore grouped, but
it's not somethingthat's coming out
today or tomorrow.
GREG: John, I know
you were saying--
JOSHUA BERG: I think
another popular one willbe to show, I mean,
I don't know if it'sgoing to be too much extra data,
but to show six months or soinstead of three months.You know, the default three
months that we've got nowis difficult to work with.Some people save that
data, but as webmasters,when we come along to
work on someone's site,they have not saved
any of that data.So all we've got to look
at is the three months.
JOHN MUELLER: Yeah.
BARUCH LABUNSKI: John, I--
JOHN MUELLER: That's something
I've heard before as well.Yeah.I know some of these
things seem easy to do,but sometimes they're a bit
harder than they initiallylook.So I know with the
data, that's somethingthat we've brought up
with the team as well,and they're aware
of this, but thereare a lot of sites out there
tracking a lot of data,so it's a lot of stuff they'd
have to kind of duplicateor triplicate depending on how
far back you'd want that data.
BARUCH LABUNSKI: John,
last time you touchedon mobile usability, and I said
if there's just a small option,can you still pass
feedback or no?
JOHN MUELLER: What do you mean?
BARUCH LABUNSKI: OK, so
for mobile usability,like, when it scans
the pages, right?So after you look at it and
you've fixed it along withthe developer or with
a team of 20 people--as in big Fortune 1,000s;
it's not that easy--and you've done that, you don't
have to come back to the sameproblem again.You just mark it as fixed
just as with crawlers.Is it hard for them to do that?Like, once you're done,
you fix the problem,check the page insights, because
it takes you there anyways,and then you've corrected the
problem, it's 100 out of 100,and you're dealing
with, like, 5,000 pages,you come back to
Mobile Usabilityagain all the errors
are there againand you're kind of confused.Just an easier-- just for--
JOHN MUELLER: I
don't know if markingas fixed would be the
best solution there,but for example another
idea might be just to say,we can recognize that you
fixed some of your pages.We can see that it's
all the same template.We'll just kind of recrawl
all of the similar pagesand reprocess them to make sure
that they're, like, OK now.
BARUCH LABUNSKI: Perfect.
JOHN MUELLER: So it's
kind of like requiringthat you mark them
individually as fixed.Maybe we could just
figure that out ourselves,but that definitely
sounds like somethingthat would make it a little bit
easier and a little bit clearerfor you to also show what work
you've done there on that site.Instead of saying,
OK, we'll justwait six months until
everything's recrawled,and then we'll see a
drop in the numbers.If you can see that change
a little bit faster,then that's probably useful.
BARUCH LABUNSKI: It'd make
a big difference, yes.Thank you so much.
JOHN MUELLER: All
right, looking usthrough some of the questions
that were submitted here,and then we'll get
back to Featurerequests for Webmaster Tools.Webmaster Tools-- back
to Webmaster Tools.How accurate is the Index
Count in Webmaster Tools?What about the Webmaster Tools
Count versus a site count?So I definitely wouldn't
use a site queryfor any kind of
diagnostics purposes.I generally just use that to
see if individual pages areindexed or not indexed.Webmaster Tools Count
is pretty accurate.It's based on the current
index data that we have,but it's sometimes
based on everythingthat we've kind of crawled
and indexed for your site.So instead of looking
at the aggregated numberfor your whole website of
all pages that are indexed,I'd recommend using
the Sitemap Fileand submitting a
Sitemap File, thenlooking at the Index Count
for that Sitemap File.Because then you'll
know, based on the URLsthat you've submitted,
the ones that you actuallywant to have indexed, how
many of those got indexed,and not how many
random pages didwe find on your
website and index.So both the Sitemaps Count
and the Webmaster Tools IndexStatus Count are pretty
accurate numbers.They should be updated
fairly frequently.The Sitemaps Count, I
think, gets updated daily.So that's something that gives
you very actionable feedbackon what we have indexed.Another sitemaps question.You mentioned last
year in a Hangoutthat XML sitemap files for
pages that are set to 404 or 410in order to speed up the process
is not the best approach.However, I've seen very
good results with it.Yes, I guess you can
do this in the sensethat if you update the
last modification date,submit the URL with the
sitemap, we'll recrawl that.We'll notice it's a 404
and we'll drop it outof the index a
little bit faster.What I wouldn't do is keep
them in there persistently.So if you've made a big
change on your websiteand you need to kind of drop
these pages fairly quickly,then submitting them in a
sitemap like that make sense.But keeping them there
for the long-run whereyou know that within your
sitemap file half of the URLsdon't exist anymore, that
just makes it very muchharder for you to
diagnose any real issues,because the index count
for that sitemap filewill be completely weird.Because on the one
hand, you submit URLsthat you know don't exist,
and on the other hand,you submit some that you
do want to have indexed.And that makes it really
hard for you to kind of takeaction on the number.But if you do this as a
one-time thing, and you say,I'm submitting them now.Google should recrawl
all these pagesand recognize that
they're gone, fine.That seems like
something you can do.
NEMANJA DIVJAK: I have to ask.We have more than 160,000
pages removed from Google,and we can remove, I
think, about 2,000 linksat this point per day.So does that affect the ranking
because of the big number?It will take, like, six months
for us to go to zero errors.
JOHN MUELLER: So what kind
of pages are you removing?
NEMANJA DIVJAK: have
websites similar to YouTube,so we've removed all
the [INAUDIBLE] content,we removed duplicate
content, canonical content,we removed more than 150,000
users from the websitethat they were link
building to their pages.So we removed
everything, so we haveabout 160,000 errors
at this point.
JOHN MUELLER: OK.And these are pages
that return 404?Something like that now?
NEMANJA DIVJAK: Yeah,
everything is all 404.
JOHN MUELLER: Yeah.That's fine.I would just let them be
recrawled and re-indexedlike that naturally.So I wouldn't use a URL
removal tool for that manually,and that's fine.It's not that it would
cause any problems for usif we see a lot of 404 errors.It's actually a sign that
your site is working,technically, correct,
because if we see 404 errors,then we kind of know
we can drop that out.Whereas if you try to hide
the 404 errors with redirectsor by serving a 200 result
code to make the error countlook lower, that's
something thatactually causes more
problems for us.So 404s are perfectly fine.And that's a perfect use case
for letting those pages dropoutnaturally.
BARUCH LABUNSKI: What if we--
NEMANJA DIVJAK: Sorry.It's not a problem
for dropping them;we wanted to remove
all the bad content.It's just that in the
Google Webmaster Toolswe went from 50,000 visits
from the Google search.We are hitting, like, 150
visits per day at this point.So it's been, like, three
years now going down,and we have removed--
about three months ago,everything is cleaned up,
and it's just waiting.When will happen?I think we got hit by the
penguin, some kind of penguin,but is there a chance to repair
that domain in the followingmonths?I know it's algorithm
and everything, but--
JOHN MUELLER: It's
really hard to saywithout looking at
the site, but if thisis something that
has been happeningover the course
of multiple years,then that sounds like we've
kind of picked up on this issuestep-by-step and said,
OK, this site overallis kind of lower quality.And it sounds like
we almost haveto reevaluate those
general decisions thatare algorithms are made.And if you've removed a lot
of really low-quality content,then that's something
that I thinkour algorithms will
pick up on over time.But it's not something where
you'd see a change from one dayto the next, or even from
one month to the next.So I imagine this
is something thatwill take a couple of months
to kind of settle down again.And if your page was raking
so well a couple years ago,a lot of things have changed
on the web, a lot of thingshave changed in our algorithms.I wouldn't expect that
it will automaticallypop back up to the old state.It might pop up to a
newer state that's higher;it might pop up to a newer
state that's lower than before.So things have changed on the
web over those couple of years.I think if you significantly
cleanup the lower qualitycontent and make sure it kind
of stays out of your site,maybe you revamp your site
to really look fantasticand have fantastic
content, then thoseare good prerequisites
to kind of showing upmore visibly in search.But it takes a lot
of time for thingslike that to kind of settle
down and pop back up.
LYLE ROMER: Hey,
John, if there's pagesthat the algorithm doesn't index
even though everything seemsright with them and
they're in the sitemaps,is that an indication
that the algorithm feelsthat that page is
of low quality?
JOHN MUELLER: Sometimes, yeah.It's really hard to say
without looking at the pages,because most of the time this is
more of a technical issue wherewe say, we crawled this page.We found no index.Or we crawled this
page, and it lookslike it's redirecting us to a
see; maybe they don't see that.But it's usually more
of a technical issue.It's fairly rare that
we would look at a pageand say, well, this
whole website looks OK.But this page is so low quality,
we're going to skip on it.So that's kind of
a rare situation.It can happen that our
algorithms look at those pages.Usually they'd look
at the whole siteand say, well, this whole
site is kind of low quality.It doesn't make sense
to spend too much energyon crawling and indexing
everything from this site.So that's usually the level
where the quality algorithmscome in with indexing.Most of the time,
if this site isbeing crawled normally and
indexed normally otherwise,then it's more a
matter that there'ssomething technically weird
with that individual pagethat it wouldn't get indexed.
LYLE ROMER: OK.Thanks.
GREG: John, if I can
jump in one last timewith a question
JOHN MUELLER: Sure.
GREG: Regarding subdomains, if
you have a blog that's hostedon a subdomain-- you know,
blog.website.com versuswebsite.com/blog having
it in the root domain--when you're doing outreach
and acquiring links backto the blog, wouldn't it make
more sense to have the bloghosted as part of the
root domain instead?
JOHN MUELLER: From a
search point of view,it doesn't really
matter that much.Sometimes for technical
reasons, sometimes for marketingreasons, you put it in a
subdomain or a subdirectory.That's essentially up to you.That's not something where we'd
say from a search point of viewit makes any big difference.The other thing
from your questionthat sounds more
tricky is you'regoing out and doing outreach
for link building, whichsounds like you might be doing
something sneaky with regardsto your links.But that's independent of where
you have those pages located.So that's more of a thing I'd
be kind of cautious about,or make sure you're handling
that in a way that you're not,I don't know, buying
links or trading links,those kind of things.I don't know.You're shaking
your head, so thatsounds like that's not
something you're doing.
JOHN MUELLER: OK.But from a technical
point of view,like subdomain or subdirectory,
is essentially up to you.What I would shy away from is
using wild card subdomains,so you have, like,
keyword.yourwebsite.com,because that makes it a lot
harder for us to actually crawlyour pages properly.But if you just have
blog and WWW, fine.That's no problem.
BARUCH LABUNSKI: John, just a
very quick on the homepage, OK?I don't remember
from all the Hangoutsthat I've been in that
we've ever touched on that.What would you recommend?So if brand, not a brand,
family, locally owned business,how much content should
be on the homepage?
JOHN MUELLER: At least 250
words or 700 characters.
BARUCH LABUNSKI: OK.Because I see a lot of
brands that don't have--
JOHN MUELLER: Sorry.Sorry.I just made that up.No, I mean, it really
depends on whatyou want to put
on your home page.
BARUCH LABUNSKI: No, I know,
because a lot of professionalsrecommend 750 and up, and
make big blogs and thingsabout it and so on and
talk about this forever.So I just thought--
JOHN MUELLER: It's not the
case that we go out and countthose words, and, like, our
algorithms have a thresholdand say, this is a good page,
because it has more words on itor fewer words on it.We try to find out what's
relevant for these pages,and a lot of times
that's in the text.So having enough text in
there to kind of help usrecognize those pages,
what they're about,what's actually relevant
on those pages, that'san important aspect there.But it's not that you
need a specific wordcount for that or
anything like that.
BARUCH LABUNSKI: OK.There's many blogs out there
that touch and say, 750,1,000 will help you.
JOHN MUELLER: Yeah.I really wouldn't worry
about the word count there.I'd really look at the
pages and look at the textand see if it's
actually relevant.If it's something
that you can pick upon where you can kind
of use that information.And also make sure that you're
looking at the actual textand not looking at
things like the images.What I still see a
lot of times is peoplewill kind of format
text in a fancy wayand keep that as an
image on their page.And then suddenly the logo
with, like, the subheading,that tagline, all of
that, is an image,and if we don't see that
at all in text form,then that makes it
really hard for usto see that this is actually
relevant for this page.We're getting better and
better at understanding pages,but kind of extracting
text from imagesand saying this is
a part of a page,is still a pretty hard problem.
BARUCH LABUNSKI: And
brands that don't have textat all on their
homepage, you recommendthat they do have text, right?I mean, I see still a lot of
brands that don't have text,and the person that owns
the site tells me, hey,we don't want text
because we like the style.And so we don't
care about the text.
JOHN MUELLER: Yeah.It's always tricky, because
some of these sites thatare really, really nice.They have really
minimalistic design,and it kind of matches
their marketing,the kind of general corporate
identity of the site,of the business, and sometimes
it just looks really nice.But if we don't have a lot
of text on these pages,it's really hard for us to pick
up which pages are relevant,which pages we should
be showing in search.So things like brand
queries or someonesearching for the company
name, we'll still usuallybe able to pick that
up and say, well,this website belongs
to this company.We'll show it in search.But if they're looking
for specific products,and we actually never
are able to findthat text on those
pages, then that'sgoing to be really
tricky and kind of hitand miss for us to actually
be able to use that.
BARUCH LABUNSKI: Thank
you for touching on that.Thanks.
JOHN MUELLER: Sure.
MALE SPEAKER: Doing a small
follow up on that question.We're working with a very
big automotive website.And as an automotive website,
they have a lot of photos.They usually do car
news and car reviews,and they have hundreds of
thousands of article for that.And for each article,
there's the gallery page,and of course for
each photo, there'sa separate URL
from that gallery.The idea is that
each photo URL onlyhas the actual photo and
links to all the other photosand then back to
the picture gallery.We did a rel="canonical" back
to the picture gallery fromthe picture URL, because we
thought it's too little contentand too similar from
all the other pages.But of course we sacrificed
the image trafficthat was-- the images
are not longer indexed.Is that something
that we should remove?Is it OK for us to have a
lot of image URLs with pagesthat only feature that image?
JOHN MUELLER: That's fine.I think that's a
decision you can make.And I know some sites don't
want their big images indexedin the image search, and
that's their decision to do.It's not something
where we'd say,this website is lower
quality because they'reblocking the images
from being indexed.That's essentially up to you.So if that makes sense
for you to kind of getthis kind of traffic, fined.If you want to
kind of concentratethings on your other
pages, that's fine, too.One thing you could do
is maybe treat, like,a sample of these
pages differently,and see which one actually
works better for your site.And you'll quickly see, does
it make sense to do this changeor does it not make
sense to do this change?And that's something
that should onlybe affecting those
specific types of pages.So that's something you can
fairly quickly kind of testto see are users really
picking up on this?Do they want to go to these
individual photo pages?Or are they happy
with the other kindof content on the
rest of your site?
MALE SPEAKER: But would
it be a problem for us--let's say we remove the
rel="canonical," and we have,as I said, a few hundred
thousand articles,but if you remove the
rel="canonical," we'll have,like, a couple of million
photo URLs suddenly addedto the index.Again, pages with the photo
and links to the other photosand back to the picture
gallery and the article.Would that be a problem?Would Google consider
it less of a quality?
JOHN MUELLER: That
shouldn't be a problem.That's something where I
think this is unique content,it's relevant content.If someone is searching
for that image,then that's the type of
thing that might be fine.I think you probably want to
just watch out for your servercapacity to make sure
that your server isable to handle this load.If Googlebot were
suddenly able to crawl,like, 100 times
the number of URLs,is that something that would
cause problems for your site?And if so, maybe it makes
sense to kind of holdthat back a little bit.But if your server can handle
that load without any problems,or if you can ramp
this up step-by-step,that's something to try out.I don't see any
problems with that.
MALE SPEAKER: OK.Thanks.
GREG: John, I have to follow up
on that last response you had.I'm dying on the inside, slowly.I just want clarify.Outreach and
content marketing isnot against Google's
guidelines, correct?We're not buying anything or
doing anything of that sort?
JOHN MUELLER: As long
as you're not going outand kind of creating artificial
links on other sites,I don't see a big
problem with that.If you're drawing
attention to your business,drawing attention to your
website, that's generally fine.If you're going out to hundreds
of directories and saying,hey, please put a link
to my site on your site,then that's something I
kind of shy away from,because I know that's something
our algorithms would pick upon as kind of a web
family type thing.And maybe the Manual Web Spam
team would also look at thatand say, well, it looks
like this guy is justdropping his link
everywhere; maybe weshould be more cautious
with the links to his site.
GREG: But when I only
do editorial-based linksthrough contacting different
websites and bloggers,not spammy stuff?
JOHN MUELLER: Yeah.I don't know what kind
of links you're doing,so I can't say everything
you're doing is perfect.But it sounds like you're
doing the right thing.It sounds like you're
kind of drawing attentionto your business,
to your website,and if people like what they
see and they link to it, fine.
BARRY: Thank you, sir.
JOHN MUELLER: All right.If a website main menu
their target URLs?Yes, they do.So we're getting
it's kind of like a menufor your website, then we'll
be able to pick up those linksand treat them
like normal links.So if you want to have them
treated as no-follow links,I'd just make sure that
as well, and then wecan pick that up and
use that appropriately.But if this is a normal menu
for within your website,then, in general, we should be
able to pick that up and crawlyour website based on that
like any other type of menu.
ARTHUR RADULESCU: John, can
I shortly follow up on this?
JOHN MUELLER: Sure.
ARTHUR RADULESCU: What if
when you mouse hover over it,it will pop up, and
then it will go back.What about this?Is this a problem from
Google's point of view?
JOHN MUELLER: That's fine.I mean, that's similar
to other type of menus.If you use CSS, if
can do something similarwhere on hover it kind
of displays things.Otherwise it highlights them.That's fine.I mean, a lot of websites
have that kind of navigation.
ARTHUR RADULESCU: OK.Thank you.
ROBB: John, can
I ask a question?
JOHN MUELLER: Oh, just let
me just go through this one,and I'll get you
right after that.
JOHN MUELLER: In
Webmaster Tools we'regetting server error
since the 9th of October,no response failure
rate is about 5%out of 200,000 requests per
day, server's not overloaded,no maintenance, no
firewall problems,no 50 times errors in logs,
Fetch as Google works fine.What could be the problem?I'd probably have to take a
look at that site specifically.So if you can post in
the forum and maybedrop a link from the
Hangout event page,I could take a
quick look at that.Sometimes this can be kind of
a misleading error in the sensethat, for example, if your
robots.txt file is unreachableonce a day, then that'll
affect a lot of potentialcrawls that we have
on your website.It's actually just one
request that's unavailable,but because it's
unavailable, we'llkind of delay crawling
of anything elseuntil we can fetch it again.So sometimes this
kind of situationcomes up with individual
requests that we can't get to.Sometimes it's a
legitimate problemthat we just can't
crawl everythingthat we try to request
from your server.And somewhere along the
line, in the firewall,maybe in the hosting
environment somewhere,those requests might
be getting blocked.But I can take a quick look if
you post your URL on the forum.
JOSHUA BERG: Is it OK
to limit the speed?
JOHN MUELLER: I'm sorry?
JOSHUA BERG: Is it OK to limit
the speed of the Googlebotindexing to slow it down?I've seen some people do that.I didn't think it
was the best idea.
JOHN MUELLER: So when we
run across a lot of errors,we'll generally slow
our crawling downbecause we think-- well, when we
run across a lot of connectionerrors or server
errors, we'll generallyslow the crawling down because
we think maybe Googlebot isthe reason why these
errors are showing up.Maybe we're doing too much.So that's something
that will happen.With normal crawl
errors, like 404s, 410s,we don't slow down, because
that's the normal serverresponse.But if we can't get
a normal response,then we'll slow
down a little bit.But if this is like a
stable 5% that we see errorswith the connection
problems like this, thengenerally that's not going
to cause any problemsfor indexing,
because we can stillcrawl more than enough
with the other requests.But I still like to kind of
resolve these kind of issuesas a webmaster,
because that way youknow that it doesn't
cause any problems.That way you know that
there's nothing kind of flakyin your network
leading to your server.All right.Robb?I think you're muted.
ROBB: Can you hear me now?
JOHN MUELLER: Yes.
ROBB: Is there any--
do search results varyaccording to the
time of day ever?We're a gift company, so we
see seasonal fluctuation, whichis obvious, but is there any
chance the search results couldbe influenced by
time of day as well?
JOHN MUELLER: I don't
think we do that.
ROBB: Or would that
be customer behavior?
JOHN MUELLER: Yeah, it
could be customer behavior.I know we've looked into
that a couple years ago.At least here,
where we were kindof analyzing what
people were searchingfor in various
countries to kind of seeif we need to tweak things
around a little bit and kind ofmake sure that the right
kind of content is visible.Also, with regards to,
like, weekend, weekday,maybe it makes sense to show
different search results.But as far as I know,
we don't do that.
ROBB: All right.Because we've seen a
lot of buying patternschange to the
evening, and I didn'tknow whether it was because
everyone's far more mobile.We started ten years ago, and
we saw that during the day,when people are at work and
had access to a computer,they'd buy, and then in
the evening they didn't.But obviously that's
now completely switched.But--
ROBB: This year in
particular, we'veseen evening purchases go up
a lot versus during the day.
JOHN MUELLER: I don't
think we do anythingwith the search ranking
there that we'd see.Evening we'll kind of
boost this type of site,and during the daytime, we'll
boost a different type of site.I think it might be an
interesting thing to experimenton, but it's very
tricky to try out.
BARUCH LABUNSKI: Would
you ever do that?
JOHN MUELLER: I mean, we run
experiments all the time.So I wouldn't say that
we'll never do that,but if this causes
more confusionand we notice that
users are kindof getting lost with
the search resultsbecause what they search
for in the morning whenat work doesn't match what
they search for in the eveningwhen they want to go buy
it, that can be confusing,you know?
BARUCH LABUNSKI: The morning
plumbers to the morning people,and then the 24-hour plumbers--
JOHN MUELLER: I don't know.I could imagine if we had,
like, lots and lots of data,that might make sense.Or if we knew that
people in evening timeswere more likely
to, I don't know,do informational
searches, then weshould interpret them a
little bit differently.But I think that's probably
not the case at the moment.I don't know, maybe it'll
happen in the future,but I don't think
that'll happen that soon.
LYLE ROMER: Hey,
John, could I justask one quick final question?In Webmaster Tools we've
been getting some errorson mobile usability.I just wanted to get an idea
of how much of a red flagwe should look at that at with
respect to ranking in general.
JOHN MUELLER: You probably see a
stronger impact from your usersdirectly.So if users are using your
site on their smartphone,and they can't get to the
content that they want,they're going to bounce
and go somewhere else.I think the statistic I had
from a recent conferencewas something like
60% of the users,when they run across a
site that doesn't work wellon their smartphone, they
avoid it in the future.So that's something that people
use their smartphones a lotfor search, and if they
can't get to the content,they're going to
go somewhere else.So from that point
of view, you'reprobably going to be
seeing a stronger effect.You probably also have trouble
seeing that in your logsthough, because if people
go to your site once,and they don't come
back on their phone,then you don't see
that 50% of your usersare using smartphones,
for example.Because they just go
there once, and theydon't go there a second time.From a search point
of view, at the momentwe show the little
mobile friendly labelin search results.I imagine, over time, users
will kind of get used to thatand focus on those
sites more whenthey search on a smartphone.At some point we're going
to look into whether or notit makes sense to rank
those sites differentlyfor smartphone users as well.So if your site doesn't
work well on a smartphone,maybe we shouldn't be showing
it in the search resultsas visibly for smartphone users.Maybe we should show it
normally for desktop, but notas visibility for
smartphone users.But that's still a bit off.So we still have a
lot of experimentsto do and to figure out
how we should handle thatappropriately.But I think that's going to be
a general direction that we'regoing to be heading for.But in the meantime,
you'll probablysee a much stronger
effect, really,just from the way users
are using our site.
BARUCH LABUNSKI: Like, I mean,
I tried to go to Adobe.com,and it's all, like, just
a black site on mobile.Like, I don't understand.Can you guys take
a look at that?
JOHN MUELLER: Yeah.I imagine something like
Adobe will be a bit different,because there are people
who are specificallygoing to that company's
website on purpose.But if you're looking
for something genericand you just want to
buy a pair of shoesand you're on your smartphone
and you're not on your laptop,then you're not going
to bother with, like,one site that doesn't work
well on your smartphonewhen you know that
other people offerthe same product in a way that
does work on your smartphone.So in cases like
that, users are goingto go somewhere
else where they canget that information
or that product.
BARUCH LABUNSKI: All right.Thank you for a wonderful year.
JOSHUA BERG: Thanks, John.
BARUCH LABUNSKI: Thank
you so much for everythingyou've done, John.
JOHN MUELLER: Yeah.Should we do another
one next week?Are you guys around?
BARUCH LABUNSKI: No.No.We're here.
ARTHUR RADULESCU: That would
be definitely a good thingto do another one.[INTERPOSING VOICES]
JOHN MUELLER: All right.Yeah.We have a couple
questions to wind up.
NEMANJA DIVJAK: Sorry.I said it seems that the
same people are always here.
We're always here.We never stop working.
JOHN MUELLER: That's
why I try to invitesome new people
from time to time.I think sometimes that
works; sometimes not so much.
ARTHUR RADULESCU: John, can
I propose to do a funny one?I mean, closer to the new year?It would be very, very
nice to have a funny one.
JOHN MUELLER: A funny one.OK.
ARTHUR RADULESCU: Yeah.Where you are
answering, directly,the questions and
with data and stuff.Real numbers.
JOHN MUELLER: Real numbers.Or maybe I'll just say yes
or no for all questions.We can go through more
questions that way,instead of having
to explain why.
BARUCH LABUNSKI: We'll
start with Barry.
JOHN MUELLER: OK.I'll set one up for next.I'm around as well, so we
can set that up and seeif we can do something a
little more light-hearted.
Thank you, John.
JOHN MUELLER: All right.Thank you all for joining.I wish you all a
great Christmas,if you're celebrating.And thanks for all the
questions, and maybe I'llsee you next week or next year.