$30 off During Our Annual Pro Sale. View Details »

International SEO - SMX Munich 2022

International SEO - SMX Munich 2022

My talk "Tomorrow, the world" from SMX Munich 2022 covering 15 tips for taking your SEO international, including content localisation, hreflang implementation and best practices, domain setup, geo targeting options and much more.

Bastian Grimm
PRO

November 07, 2022
Tweet

More Decks by Bastian Grimm

Other Decks in Marketing & SEO

Transcript

  1. Tomorrow, the world
    15 tips for taking your SEO international
    Bastian Grimm, Peak Ace AG | @basgr

    View Slide

  2. Let's start with the most basic question
    What do you want/need?

    View Slide

  3. What do you want/need?
    Multilingual, multi-regional – or both?
    multilingual
    e.g., a Swiss business with a
    German and French website
    multi-regional
    e.g., a website that explicitly
    targets users in different
    countries
    … or both
    e.g., a website might have
    different versions for
    DE & CH, and both DE & FR
    versions of the CH content

    View Slide

  4. Well… that's a shame! But we're here to help.
    You don't know which one you need?

    View Slide

  5. pa.ag
    @peakaceag
    5
    How do search engines determine your intended audience?
    You need to give search engines a few hints to let them know which audiences you're
    catering to. Here are a few pointers to explain this:
    Credits: https://pa.ag/2CqYwvB
    Content analysis
    In which language is the content
    written – and for which region?
    e.g., British English is different
    from American English.
    TLD hints
    In using a ccTLD
    (e.g., .de) search engines
    assume that you're targeting
    that country (e.g., Germany).
    Primary origin of links
    If a website gets most of its
    links from a specific ccTLD, it’s
    likely that it will be associated
    with the respective market.
    hreflang annotations
    Are you using hreflang to
    communicate which audiences
    you want to target with which
    URL?
    Content‐language attribute
    Bing, Baidu & others use the
    content-language HTML
    attribute for targeting hints
    instead of hreflang.
    Google & Bing config
    Properly configured language
    and/or geo targeting in GSC
    as well as geo targeting in
    Bing WMT.
    And others
    This is not a comprehensive
    list. There are most certainly
    other signals too.
    Local/business listings
    e.g., Google My Business
    listings and Bing's Places
    for Business.

    View Slide

  6. pa.ag
    @peakaceag
    6
    Google's algorithms can vary
    An algorithm can vary from country to country. The “US algorithms” are often the most
    advanced, followed by the UK and some EU countries, such as Germany:

    View Slide

  7. Less mature markets usually
    have higher search volume
    on generic keywords
    But generic keywords do have higher ambiguity in
    terms of user intent…

    View Slide

  8. pa.ag
    @peakaceag
    8
    30-second recap: What's search intent anyway?
    Search intent is the why behind a search query: why did the person make this search?
    Are they looking for information, to make a purchase, or for a specific website?
    Informational
    Navigational
    Commercial
    Transactional
    ▪ "Jason Statham movies"
    ▪ "Berlin Paris distance"
    ▪ "what are carbs"
    ▪ "peak ace address"
    ▪ "gmail"
    ▪ "instagram login"
    ▪ "Dubai winter temperature"
    ▪ "haircut near me"
    ▪ "best webinar software"
    ▪ "Audi rsq8 price"
    ▪ "champagne next day delivery"
    ▪ "BER CDG flights"

    View Slide

  9. pa.ag
    @peakaceag
    9
    Google is obsessed with “Intent”
    The current version of their Search Quality Evaluator Guidelines mentions “Intent” over
    420 times – the “Needs Met” section spans over 30 pages:
    Source: https://pa.ag/2W1qRCS

    View Slide

  10. pa.ag
    @peakaceag
    10
    Machine learning plays a huge role here as well
    Back in 2007, Microsoft published a patent suggesting that 87% of ambiguous queries
    can be identified and understood with supervised machine learning:
    Source: https://pa.ag/2XHdZTt
    We propose a machine learning
    model based on search results
    to identify ambiguous queries.
    The best classifier achieves accuracy as
    high as 87%. By applying the classifier,
    we estimate that about 16% queries
    are ambiguous in the sampled logs.

    View Slide

  11. pa.ag
    @peakaceag
    11
    Thanks to recent advances in ML, Google has
    made huge leaps ahead with getting search
    intent right - and they’re only going to get
    better at it.

    View Slide

  12. pa.ag
    @peakaceag
    12
    Can't get your head around automating at scale?
    Kevin Indig has got you covered! Go check out his two articles on the topic:
    Source: https://pa.ag/3u41oFj

    View Slide

  13. pa.ag
    @peakaceag
    13
    You're late to the party if you haven't figured this out yet:
    It’s of utmost importance right now
    to get intent mapping right; intent
    means relevance and therefore better
    rankings. Get this wrong, and you
    have no chance of ranking long term.

    View Slide

  14. Proper research is crucial
    Every customer wants something different. Make sure
    you're delivering the right content to the right people.

    View Slide

  15. Many businesses use a
    "rinse and repeat" approach
    But that ignores the unique features (e.g., cultural
    differences, etc.) of each market

    View Slide

  16. pa.ag
    @peakaceag
    16
    Updates are usually "global" (unless stated otherwise)
    However, their impact certainly can, and will, vary from country to country.
    Source: https://pa.ag/328cDRm

    View Slide

  17. pa.ag
    @peakaceag
    17
    ICYMI: Google doesn't adapt results according to their TLDs
    The results are localised nowadays, independently of the TLD of the Google. This means
    that the results would be the same on google.de and google.com in Germany.
    Source: https://pa.ag/2JMk71d
    […] the choice of country service will no longer be
    indicated by domain. Instead, by default, you’ll
    be served the country service that corresponds
    to your location. So if you live in Australia, you’ll
    automatically receive the country service for
    Australia, but when you travel to New Zealand,
    your results will switch automatically to the
    country service for New Zealand.

    View Slide

  18. Some of the most frequent issues and
    missteps we're seeing with international SEO
    15 common mistakes
    in international SEO

    View Slide

  19. pa.ag
    @peakaceag
    19
    #1 Not using different URLs for each international site
    Avoid using something like cookies for different versions of a site that use the exact
    same URL. In this example, only the English version will be indexed for MAERSK:

    View Slide

  20. Each language has unique characteristics,
    however tech is usually built in English first.
    Sort out special characters

    View Slide

  21. Don't forget about forms – accept them there!
    Ensure you handle them in URLs
    and file names (e.g., normalise)

    View Slide

  22. pa.ag
    @peakaceag
    22
    #2 Missing proper localisation of content, keywords, etc.
    American English vs. British English, terminology, synonyms, colours, national holidays…
    VS

    View Slide

  23. domain.es/en/ or domain.de/en-gb/
    Targeting the UK won't work, because the ccTLD
    is already geotargeted by default.
    #3 Using a single ccTLD to
    target multiple countries

    View Slide

  24. pa.ag
    @peakaceag
    24
    This shows the French version/SERP of Google, but
    returns all German results for New Yorker…
    #3 Using a single ccTLD to target multiple countries
    Fashion brand New Yorker uses the German .de to geo target globally…

    View Slide

  25. pa.ag
    @peakaceag
    25
    #4 Automatically redirecting users without giving options
    It's better to just let the user pick the suggested/relevant international website instead.
    VS

    View Slide

  26. Also: non-self-referencing canonicals can't have hreflang!
    #5 Canonicalising all versions to
    one to avoid DC

    View Slide

  27. Google doesn't support regional targeting on a
    continental level. Using "eu" in hreflang won't work.
    #6 Using malformed/invalid
    values in hreflang

    View Slide

  28. Google needs to (re-) crawl the entire pairing of URLs
    to understand your implementation!
    +8 of the most common implementation errors (coming up later)!
    #7 Not understanding how
    hreflang actually works

    View Slide

  29. Translate all the things!?
    What if that's not always the right approach?
    Translation vs. localisation vs. transcreation

    View Slide

  30. Transcreation?

    View Slide

  31. pa.ag
    @peakaceag
    31
    Translation
    The process of re-writing a sentence in a
    different language.
    Converting content "word for word" is
    technically already translating.
    This process strives to make the translated
    text as true to the original as possible.
    Key characteristics
    ▪ The language changes, but the words
    and the message stay the same.
    ▪ Language is translated and context is
    only given in editor's notes.
    Translation vs. localisation vs. transcreation
    Let's talk some definitions and characteristics, shall we?
    Localisation
    When you take translated content and
    edit it to reflect the culture of the target
    language, you are localising the content.
    This is key to online marketing, as user
    experience and readability, for example, are
    important factors that heavily impact the
    content format and length.
    Key characteristics
    ▪ The words change, but the meaning
    behind those words stays the same.
    ▪ The language is also adapted in a
    culture-sensitive way.
    Transcreation
    Transcreation is possible when the author of
    the source text (in language A) works
    together with the author of the
    translated/localised text (in language B).
    Both authors work together to tell a story to
    obtain the same effect in different
    markets.
    Key characteristics
    ▪ The content changes but the business
    goals stay the same.
    ▪ The content is developed anew in the
    local language.

    View Slide

  32. pa.ag
    @peakaceag
    32
    Translation vs. localisation vs. transcreation
    In other words…
    Translation would convert
    Super Bowl into the German
    Superschüssel.
    Transcreation would build a
    new, country-specific campaign
    around the ideas of fun and sport.
    Localisation would contextualise
    Super Bowl and probably use
    World Cup instead.

    View Slide

  33. If content is translated literally, it
    might not make sense in the new
    target market.

    View Slide

  34. The reason that we often localise rather than translate is
    that culture plays a key role in deciding on the tone of
    voice for a client.
    Localisation instead of translation

    View Slide

  35. Localisation considers details such as imagery, content
    visual placement, length – and more.
    Localisation beyond text

    View Slide

  36. pa.ag
    @peakaceag
    36
    For global campaigns, language matters a lot!
    On average, German contents needs 10% to 35% more space than its English counter-
    part? You need to plan for that – e.g. when working with your design team!
    English
    ▪ Relevant tips
    ▪ Employee satisfaction
    ▪ Sick note
    ▪ Financial services company
    German
    ▪ Sachdienliche Hinweise
    ▪ Mitarbeiterzufriedenheit
    ▪ Arbeitsunfähigkeitsbescheinigung
    ▪ Finanzdienstleistungsunternehmen

    View Slide

  37. pa.ag
    @peakaceag
    37
    To translate or not to translate?
    Some global strategies don't need full adaptation, because keeping the original
    language adds value for the end-user. Think of all the (unpronounceable) IKEA names.
    ▪ The world is full of creative examples, especially if we
    look at major multinational businesses that set the
    standards of localisation. But the question sometimes
    is: is localising even worth it?
    ▪ Sometimes keeping the original language adds value
    for the end-user, as it adds flair to a product. Just
    think of all the unpronounceable names at IKEA.
    ▪ How many memes and jokes have been made
    around the Swedish company?

    View Slide

  38. pa.ag
    @peakaceag
    38
    Translation, localisation - borders sometimes get blurry
    But maybe that doesn't really matter?
    ▪ The English adaptation of the Haribo slogan
    maintains the effect of the rhyme and the catchy
    playfulness of the German version.
    ▪ If translated back (“Kinder und Erwachsene lieben sie,
    die fröhliche Welt von Haribo”) the new word choice
    doesn’t convey the original slogan word for word.
    ▪ So, what makes a message stronger? An exact
    translation or a localisation of the slogan?

    View Slide

  39. pa.ag
    @peakaceag
    39
    What's your favourite team?
    Localisation and transcreation affect images just as much as the written content:
    An example I use a lot is the cover of FIFA,
    the video game. On the German cover there
    are players from German teams (say, Neuer
    for example) whereas on the Brazilian cover
    there are players from other teams.”
    Irene Morcilo San Jose, Peak Ace AG
    ▪ FIFA adopts a flexible approach where there's always one global game cover alongside the regional ones.
    ▪ Whether or not a global or regional cover is used in a market depends on internal research. A player like Lionel Messi,
    featured for 5 consecutive years, is well known in more than one country and doesn't need to be “localised”.

    View Slide

  40. pa.ag
    @peakaceag
    40
    At Peak Ace, we speak 20+ languages
    But why do we speak so many?
    If you talk to a man in a language he
    understands, that goes to his head.
    If you talk to him in his own language,
    that goes to his heart.
    Nelson Mandela

    View Slide

  41. And it gets even worse: localisation even exists (and
    certainly make sense) within single-language markets
    The importance of localisation
    cannot be stressed enough

    View Slide

  42. View Slide

  43. This kind of content speaks to them, catches their eye and
    ultimately “sounds” better than anything else.
    The end user is much more likely
    to respond to content that fits
    their culture

    View Slide

  44. pa.ag
    @peakaceag
    44
    A shoe to remember – or how to insult a whole country
    In 2011, Puma released a global marketing strategy that used the flag colours of each
    nation to honour diversity.
    Source: https://pa.ag/3MswEGX
    Someone at Puma clearly
    forgot to do their due
    research, as they showed a
    significant lack
    of cultural understanding in
    the United Arab Emirates.
    ▪ In UAE, anything associated
    with feet and the floor is
    considered dirty.
    ▪ UAE nationals were
    therefore not pleased to
    have their national flag
    turned into shoes.

    View Slide

  45. pa.ag
    @peakaceag
    45
    Translate, localise or transcreate – categorising the three:
    One could go as far as to categorise potential tasks as follows:
    Translation Localisation Transcreation
    Useful for technical content:
    manuals, safety warnings, legal or
    medical documents, etc.
    Useful for global online marketing in
    general, international SEO, UX
    elements, visuals etc.
    Useful for global campaign ideation
    and concepts, ad copy, global
    processes, etc.
    Focus on the source, technical
    knowledge required
    Balance between source & target
    language, technical knowledge with
    language support required
    Focus on the target language, deep
    knowledge of the new market
    culture & creative brief required
    Note: This is only a generic categorisation. You should
    evaluate which skillsets you need on a project-by-project basis.

    View Slide

  46. pa.ag
    @peakaceag
    46
    A practical example
    All methods in comparison:
    Source: Deutsche Welle
    Here's an excerpt from a Deutsche Welle article, published in Autumn 2021 for the German audience.
    Je kälter es draußen wird, desto mehr halten wir uns normalerweise
    drinnen auf. Und damit kann sich das Corona-Virus in geschlossenen
    Räumen wieder besser verbreiten. Laut Robert-Koch-Institut (RKI) war
    die Sieben-Tage-Inzidenz am Wochenende erstmals wieder dreistellig
    und stieg zu Wochenbeginn auf 110,1 Neuansteckungen je 100.000
    Einwohner. Vor einer Woche hatte die Inzidenz noch 74,4 betragen.”

    View Slide

  47. pa.ag
    @peakaceag
    47
    How could you adapt the excerpt?
    Market Translation Localisation Transcreation
    UK
    The colder it gets outside, the more we usually stay
    indoors; and the more time we spend indoors, the more
    easily Corona Virus can spread. According to the
    Robert Koch Institute (RKI) the seven-day incidence was
    in triple figures this weekend and rose to 110.1 new
    infections per 100,000 inhabitants at the beginning of
    the week. Just a week before, the incidence had been
    74.4.
    The colder it gets outside, the more we usually stay indoors;
    and the more time we spend indoors, the more easily COVID
    can spread. This is certainly something that has been seen in
    Germany. According to the German health authority, the
    Robert Koch Institute, the seven-day incidence was in triple
    figures this weekend and rose to 110.1 new infections per
    100,000 inhabitants at the beginning of the week. Just a week
    before, the incidence had been 74.4. In England, however,
    rates have remained lower at 19.7 in spite of fears of the
    Delta Variant (or, as it was known in Germany, the British
    Variant) of the Corona Virus spreading across the channel.
    The colder it gets outside, the more we usually stay indoors;
    and the more time we spend indoors, the more easily
    COVID can spread. The UK is no exception to this.
    According to the UK government, the case rate rose to
    56,000 new infections at the weekend. Just a week before,
    cases had been stable at 36,000. A clear increase in cases is
    here along with the colder weather.
    US
    The colder it gets outside, the more we usually stay
    indoors; and the more time we spend indoors, the more
    easily Corona Virus can spread. According to the
    Robert Koch Institute (RKI) the seven-day incidence was
    in triple figures this weekend and rose to 110.1 new
    infections per 100,000 inhabitants at the beginning of
    the week. Just a week before, the incidence had been
    74.4.
    The colder it gets outside, the more we usually stay indoors;
    and the more time we spend indoors, the more easily the
    Coronavirus can spread. This is certainly something that has
    been seen in Europe. According to the German health
    authority, the Robert Koch Institute, the seven-day incidence
    in Germany was in triple figures this weekend and rose to
    110.1 new infections per 100,000 inhabitants at the beginning
    of the week. Just a week before, the incidence had been 74.4.
    In the USA, however, rates have remained at the lowest
    they’ve been since July, with only moderate increases in
    recent days.
    The colder it gets outside, the more we usually stay indoors;
    and the more time we spend indoors, the more easily the
    Coronavirus can spread. The United States is no exception
    to this. According to the CDC, the 7 Day Average Moving
    Case Rate has risen by 10,000 cases since the beginning of
    November. However, this is still 80,000 less than what we
    saw at the height of summer 2021. Whether we will reach
    the same all-time high of 202,000 that we saw in January
    2021 this winter therefore remains to be seen.
    UK: COVID vs.
    US: Coronavirus
    UK would talk about the
    „Delta variant“ as this started
    to rise throughout Europe,
    US would talk about general
    infection rates being low.

    View Slide

  48. Some of the most common issues around domain usage
    and structure as well as geo-redirects
    Domains & Redirects

    View Slide

  49. pa.ag
    @peakaceag
    50
    Domain basics: gTLD vs. ccTLD
    There are two versions of top-level domains: generic TLDs (gTLD) and country code
    TLDs (ccTLD); ccTLDs have a fixed geo target (i.e., de = Germany).
    gTLD ccTLD
    .com .de
    .net .fr
    .org .co.uk
    Here are some of the most common domain setup
    options from an SEO point of view:
    ▪ multiple ccTLDs
    ▪ gTLDs in combination with subdomains
    ▪ gTLDs in combination with subfolders
    ▪ gTLDs with a combination of subdomains and subfolders
    Example: peakace.de + peakace.fr
    Example: de.peakace.com
    Example: peakace.com/fr/
    Example: es.peakace.com/es-mx/

    View Slide

  50. New TLDs are treated the same way as gTLDs.
    Nevertheless, there are some interesting differences once
    you look a bit closer.
    Some exceptions to the rule
    (of course…)

    View Slide

  51. Although these domains are associated with a geographic
    region, they are treated as gTLD (much like .com).
    This is currently true for: .eu & .asia
    How does Google treat generic
    regional top-level domains?

    View Slide

  52. Google treats some ccTLDs as gTLDs.
    This is currently true for .io (British Indian Ocean
    Territory), .cc, .co, .dj (Djibouti), .fm, .io, .la, .me
    (Montenegro), .ms, .tv, .tk, .ws and some others.
    Why do .io domains rank globally
    even though they're a ccTLD?

    View Slide

  53. .brand TLDs are treated the same as any other gTLD.
    Will a .brand TLD be given any
    more or less weight than a .com?

    View Slide

  54. Even if they look region-specific, Google treats them
    as other gTLDs.
    How are the new region city TLDs
    (like .berlin) handled?

    View Slide

  55. pa.ag
    @peakaceag
    56
    Peak Ace usually recommends one of two variants:
    One strong global gTLD for all languages or separate (cc)TLDs for every country.
    Variant Example Comments
    Separate domains for every country
    (either ccTLDs only, or a mix of both)
    amazon.com
    amazon.co.uk
    amazon.de
    ▪ Each domain is managed separately
    ▪ More difficult to establish a brand in new markets (starting from zero)
    ▪ Strong geotargeting signal for Google (for ccTLDs), GSC possible for gTLDs
    ▪ Often not possible for every market (domain availability)
    Multiple subdomains under a
    single global domain (gTLD)
    en.wikipedia.org
    es.wikipedia.org
    de.wikipedia.org
    fr.wikipedia.org
    ▪ Medium effort to set up and manage
    ▪ Subdomains will be treated separately by Google
    ▪ No benefit from domain authority
    ▪ Not great for CTR
    Subfolders on a single global domain
    (gTLD)
    netflix.com/de
    netflix.com/mx
    netflix.com/it
    ▪ Easier to expand to new markets
    ▪ Massive benefit from domain authority / external links
    ▪ Depending on CMS, easy to set up and manage
    ▪ Violating Google's guidelines could impact all versions (very rarely though)
    URL parameters
    instagram.com/?hl=en
    instagram.com/?hl=de
    instagram.com/?hl=es
    ▪ Geotargeting in GSC is not possible
    ▪ Google doesn't recommend it

    View Slide

  56. Those domains have default targeting which can't be
    “overruled“ through, e.g., Google Search Console
    You can't do
    geotargeting for ccTLDs

    View Slide

  57. pa.ag
    @peakaceag
    58
    You can't override GSC ccTLD country geotargeting
    Google will associate this .fr domain with France – it's not possible to change it.

    View Slide

  58. pa.ag
    @peakaceag
    59
    For gTLDs you can select a country targeting of choice
    Google will not have a default targeting enabled for this .eu domain – so they will use
    others signals until you provide the info:

    View Slide

  59. pa.ag
    @peakaceag
    60
    For gTLDs you can (& should!) also geotarget sub-folders
    It is possible to connect a (new) GSC property to a specific country folder on your gTLD.
    Of course, this needs to be done once per individual folder:

    View Slide

  60. Keep in mind, GSC settings are obviously only
    specifically available for Google
    We are talking Google-only!

    View Slide

  61. pa.ag
    @peakaceag
    62
    Geo redirects how-to (e.g., if you need to license)
    If the user is guided to a special language folder based on their IP
    , the redirect needs to
    be temporary (302 or 307), otherwise caching issues will come up.

    View Slide

  62. pa.ag
    @peakaceag
    63
    Geo redirects gone wrong (= don't do it!)
    Keep in mind: Google crawls primarily from IP addresses in the US.
    Source: https://pa.ag/3foCesW
    […] The client’s website didn’t have a
    section for the US, so instead all traffic
    from the US was redirected to a
    landing page explaining that their
    service was not available in the US.
    So that page was the only page that
    Google ever saw. […] All the other pages
    on domain.com were basically invisible
    to Google.
    domain.com
    domain.com/gb domain.com/fr domain.com/de
    redirect redirect redirect

    View Slide

  63. pa.ag
    @peakaceag
    64
    Don't do this, either…
    Disney wastes loads of link equity by relying on JS-redirects:

    View Slide

  64. pa.ag
    @peakaceag
    65
    Implement smart geo-selection possibilities
    Most importantly, don't force your users in a specific direction – leave them with the
    option to stay where they are, or navigate freely:

    View Slide

  65. Let Google know about localised versions
    hreflang & Geo Targeting

    View Slide

  66. pa.ag
    @peakaceag
    67
    hreflang intro: What does it look like?
    Language (required): ISO 639-1, country (optional): ISO 3166-1
    By using hreflang attributes, you can target all English-speaking users
    (hreflang=“en”), but of course, if you have a dedicated version for English-speaking
    people living in Canada, you can target them by using hreflang “en-ca”.
    i

    An example of hreflang attribute
    Language
    required
    Country
    Not required

    View Slide

  67. pa.ag
    @peakaceag
    68
    Use hreflang to let Google know about localised versions
    If you have multiple versions of a page for different languages or regions, let Google
    know about the variants:
    ▪ Each language version must list itself as well as other
    language versions
    ▪ If two pages don't both point to each other, the tags
    will be ignored!
    ▪ Alternate URLs must be fully-qualified
    ▪ e.g., do not forget the correct protocols (http/https)
    ▪ Implement a language as well as a country code to target
    various languages in one country (e.g., “fr-ch” & “de-ch”)
    ▪ Consider adding an x-default fallback for other
    unmatched versions
    ▪ hreflang="x-default" />

    View Slide

  68. pa.ag
    @peakaceag
    69
    Note: Not all search engines support hreflang
    Search Engine HTML HTTP Headers Sitemaps Alternatives
    *
    *
    *
    *
    * You have to use a HTML Meta Language tag.

    View Slide

  69. pa.ag
    @peakaceag
    70
    There are three different implementation options available
    Each has its advantages and disadvantages, so it's a matter of choosing what is best for
    you and your setup:
    HTML
    Vulnerable to errors when
    using many different TLDs
    Changes made to HTML
    code can lead to errors in
    parsing/processing
    Straightforward
    implementation in
    XML sitemap
    Allows implementation
    of changes quickly
    (centralised)
    Controllable- & CMS-
    independent setup
    Initial setup is slightly more
    complex than other
    methods
    -
    - -
    +
    Server header
    Complex implementation
    (server config)
    Non-HTML files (PDFs) can
    be integrated
    High maintenance effort
    (server config)
    -
    -
    + +
    +

    View Slide

  70. pa.ag
    @peakaceag
    71
    hreflang annotations within the HTML
    Straightforward to implement and easy to maintain, provided you don't have too many
    setup/parings and that they don’t get too complex!
    Examples:



    Disadvantages:
    ▪ Code bloat: We do not recommend implementing hreflang annotations for >10 in HTML . There is a
    risk that the code will bloat, which is especially bad for slower mobile connections.
    ▪ Maintenance: Consistently timed updates for each individual website are needed, otherwise hreflang pairings
    will be broken and can’t work.
    ▪ A slow recrawling of URLs at a deeper page level often causes pairings to be broken for longer.

    View Slide

  71. pa.ag
    @peakaceag
    72
    hreflang annotations as server headers
    hreflang annotations can also be implemented using server headers, but it's harder to
    both monitor and maintain. However, it pays off for non-HTML such as PDFs, etc.
    Examples:
    Link: ; rel="alternate"; hreflang="de-AT"
    Link: ; rel="alternate"; hreflang="de"
    Link: ; rel="alternate"; hreflang="en-GB"
    Disadvantages:
    ▪ It is more difficult to monitor changes and errors, due to annotations being "invisible" to the end user
    ▪ The implementation is very complex, and the effort is high maintenance, because directives need to be
    applied on server level, e.g., using Apache .htaccess / nginx conf (which often requires dev ops to be involved)

    View Slide

  72. pa.ag
    @peakaceag
    73
    XML sitemap (on a centralised server)
    The most practical solution for large-scale setups would be to implement hreflang via
    XML sitemaps:
    Example: sitemap AT



    https://www.domain.at/series/example-page.html



    Advantages of such a setup:
    ▪ No impact on actual website code/size, meaning there is no impact on web performance
    ▪ Entirely independent from individual websites, can be setup, run and maintained by a dedicated team
    ▪ Overall, it is easier to implement and update whilst allowing for faster refresh frequency than individual HTML
    pages (this means that pairings are usually always intact)

    View Slide

  73. pa.ag
    @peakaceag
    74
    Further advantages of a stand-alone sitemap server setup
    No impact on actual web infrastructure: Generating massive sitemaps with millions of
    URLs won't affect servers responsible for delivering the website:
    Sitemap
    server
    DE
    robots.txt
    UK
    robots.txt
    FR
    robots.txt
    Each TLD robots.txt file
    points to their sitemap on
    the sitemap server

    View Slide

  74. pa.ag
    @peakaceag
    75
    Make or buy? Tool recommendation:
    HREFLang Builder gives you everything you need, and at a very reasonable price. This
    makes it a serious alternative to building a custom solution – I am a huge fan!
    Source: https://www.hreflangbuilder.com/
    Features
    ▪ Localised URL mapping
    ▪ Complex URL mapping
    ▪ XML file hosting
    ▪ Error checking
    ▪ Automated updates
    ▪ Missing page identification
    ▪ Generating XML sitemaps

    View Slide

  75. pa.ag
    @peakaceag
    76
    Use the x-default tag for unmatched languages
    The x-default specifies where a user should be sent in the case none of the
    specified languages in your other hreflang links match the browser settings.
    Source: https://pa.ag/2W9WGpO
    Sometimes the x-default has been included by accident
    and the page is not a suitable fallback for the rest of the
    world. English is a good language to use as your fallback
    option, since so many people speak it, whereas German
    is not always a great choice (or only works for the DACH
    region), since fewer people speak it globally.
    i
    The website has content that targets users around the world as follows:
    ▪ http://example.com/en-gb: For English-speaking users in the UK
    ▪ http://example.com/en-us: For English-speaking users in the USA
    ▪ http://example.com/en-au: For English-speaking users in Australia
    ▪ http://example.com/: The homepage shows users a country selector and
    is the default page for users worldwide
    In this case you can annotate this cluster of pages using HTML like this:




    View Slide

  76. pa.ag
    @peakaceag
    77
    There is an endless number of things that can go wrong
    Sitebulb will check these issues against your hreflang implementation:
    Source: https://pa.ag/3KmVdDi
    ▪ Has invalid incoming hreflang annotations
    ▪ Has invalid outgoing hreflang annotations
    ▪ Has outgoing hreflang annotations to noindex URLs
    ▪ Noindex URL has incoming hreflang
    ▪ Has outgoing hreflang annotations to broken URLs
    ▪ Has outgoing hreflang annotations to canonicalised URLs
    ▪ Canonicalised URL has incoming hreflang
    ▪ Has outgoing hreflang annotations to disallowed URLs
    ▪ Disallowed URL has incoming hreflang
    ▪ Has conflicting incoming hreflang annotations
    ▪ Has conflicting outgoing hreflang annotations
    ▪ Has multiple self-referencing hreflang annotations
    ▪ Has outgoing hreflang annotation to multiple URLs
    ▪ Has outgoing hreflang annotations using relative URLs
    ▪ Invalid HTML lang attribute
    ▪ Mismatched hreflang and HTML lang declarations
    ▪ Missing hreflang annotations
    ▪ Missing reciprocal hreflang (no return-tag)
    ▪ Has outgoing hreflang annotations to redirecting URLs
    ▪ Has unsupported or misconfigured hreflang
    ▪ Has hreflang annotations using multiple methods
    ▪ Missing canonical URL
    ▪ Missing HTML lang attribute
    ▪ Has hreflang annotations without HTML lang
    ▪ Hreflang annotation also x-default
    ▪ Missing self-reference hreflang annotation

    View Slide

  77. pa.ag
    @peakaceag
    78
    You can't geo-target "EU" in hreflang
    Regions such as Asia, MENA, etc, are not supported by Google.
    Source: https://pa.ag/3fHjoxn

    View Slide

  78. pa.ag
    @peakaceag
    79
    There are a lot of false friends…
    hreflang attribute What you think it is What it actually is
    es-pa Spanish in Paraguay Spanish in Panama
    kr-kr Korean in Korea Kanuri in Korea
    cz-cz Czech in Czech Republic –
    cr-cr Croatian in Croatia Cree in Costa Rica
    EN-IR English in Ireland English in Iraq
    fr-Mo French in Monaco French in Macau
    ne-NE Nepali in Nepal Nepali in Niger
    It doesn't really matter whether you use uppercase or lowercase for the hreflang
    attribute value. It might make it easier to read – but for Google, it's all the same.
    i

    View Slide

  79. pa.ag
    @peakaceag
    80
    (Simple) hreflang generators
    Generators can help show you what the different country codes should look like. Pay
    attention to the fact that tags must be implemented reciprocally for each URL.
    Source: https://www.sistrix.de/hreflang-guide/hreflang-generator/ & https://www.aleydasolis.com/english/international-seo-tools/hreflang-tags-generator/

    View Slide

  80. pa.ag
    @peakaceag
    81
    hreflang validation
    When analysing hreflang codes without access to Google Search Console,
    free validation tools can help detect syntax errors.
    Source: https://technicalseo.com/seo-tools/hreflang/ & http://flang.dejanseo.com.au/

    View Slide

  81. pa.ag
    @peakaceag
    82
    Browser add-ons for quick checks
    Source: https://pa.ag/3HtgnxH

    View Slide

  82. pa.ag
    @peakaceag
    83
    hreflang error monitoring in Google Search Console
    The GSC section for error monitoring should be checked continuously. hreflang is
    complex enough without errors; incorrect implementation makes it even worse.

    View Slide

  83. pa.ag
    @peakaceag
    84
    If you're using them: hreflang & mobile subdomains
    Using mobile subdomains means that the hreflang must be implemented in the same
    way as it is for desktop: reciprocally and 1 to 1.
    Desktop
    English
    Desktop
    French
    Desktop
    German
    Mobile
    English
    Mobile
    French
    Mobile
    German
    alternate media
    rel=canonical
    hreflang hreflang
    hreflang
    hreflang hreflang
    hreflang
    alternate media
    rel=canonical
    alternate media
    rel=canonical

    View Slide

  84. pa.ag
    @peakaceag
    85
    However, hreflang can (and sometimes will) be ignored
    If Google receives "unusual" elements within the (i.e., syntax, which is not
    anticipated for the section), it will ignore everything below this section.

    View Slide

  85. pa.ag
    @peakaceag
    86
    No development resources available?
    Consider implementing hreflang without actually touching your website by using
    Cloudflare Workers:
    Source: https://pa.ag/3frUZfg

    View Slide

  86. pa.ag
    @peakaceag
    87
    Back in Sep. 2017, Cloudflare introduced their "Workers"
    These ultimately became publicly available in March 2018:
    Source: https://blog.cloudflare.com/introducing-cloudflare-workers/

    View Slide

  87. pa.ag
    @peakaceag
    88
    Using a CDN, all requests will pass through “edge servers“
    When we ignore DNS, databases etc for a minute, this is what it would look like:
    First request, ever.
    peakace.js is not cached
    on edge server yet
    Origin server
    Request: peakace.js Request: peakace.js
    peakace.js delivered
    from origin server
    Response: peakace.js
    peakace.js gets cached
    on edge server

    View Slide

  88. pa.ag
    @peakaceag
    89
    Using a CDN, all requests will pass through “edge servers“
    When we ignore DNS, databases etc for a minute, this is what it would look like:
    Origin server
    Request: peakace.js
    peakace.js delivered
    from edge server
    peakace.js is cached
    on edge server
    Second request
    (independent of user)

    View Slide

  89. Intercept and modify HTTP request and response
    URLs, status, headers, and HTML mark-up.
    Seriously though, this is WILD!

    View Slide

  90. pa.ag
    @peakaceag
    91
    hreflang annotations on the edge using server headers
    Tell Google about localised versions of your page – of course both ways (HTML or HTTP
    headers) work just fine:

    View Slide

  91. pa.ag
    @peakaceag
    92
    To finish off: My personal top-8 errors/mistakes
    If something seems off, these are the areas we always check first:
    You are too impatient: Google needs to (re-) crawl the entire pairing of URLs to understand your implementation
    Non-existent hreflang values: they need to be in ISO 639-1 for language & (optionally) ISO 3166-1 Alpha 2 for region
    Irrelevant hreflang mapping: e.g., "de-DE" has been mapped to an English URL
    Non-existent URLs: broken/mal-formatted URLs, missing protocols, destination returns 4xx/5xx, etc.
    Only using the country code: you can target a language on its own, but you can't target a country without a language
    (e.g., hreflang=“GB” doesn’t work!)
    Canonicalised elsewhere: an hreflang tag pointing to an URL that's canonicalising (or redirecting) elsewhere
    No-return hreflang tag: missing the actual hreflang tag pointing back to the origin
    Missing self-referencing hreflang tag: this is often overlooked, especially in larger hreflang groups

    View Slide

  92. Care for the slides?
    Any questions?
    [email protected] Take your career to the next level: jobs.pa.ag
    www.pa.ag
    twitter.com/peakaceag
    facebook.com/peakaceag
    Bastian Grimm
    https://pa.ag/smxm22
    [email protected]

    View Slide