For publishers and advertisers

Discussion in 'Mail Chat' started by DAgent, Feb 11, 2014.

  1. DAgent

    DAgent Moderator

    Joined:
    Mar 14, 2011
    Messages:
    452
    Likes Received:
    115
    Trophy Points:
    43
    Home Page:
    Dear community,

    I wanted to write about a couple topics that I've seen that are completely overlooked and they need to be the basics for every advertiser / network, and also some information for mailers that is not public but it has a huge impact in delivery.

    For the advertisers and publisher's mix:

    - Every offer that you want a mailer to send traffic to, is ultimately hosted at one IP (or set of IPs). The same is valid replacing IP by the word domain.
    - Bots will go through the links that are included on an email sent. They will record each step of the link chain (every domain involved, every IP involved)

    Have you noticed how an offer performs better when it starts and then it "slows down" day after day when the traffic comes from email? This is not a coincidence or market saturation (since you can tell that other traffic sources are relatively constant). It has to do with how filters learn not only about your publisher's (mailing publishers) domains, but also about the IP where your offer is hosted at, and also about the domain(s) your offer is hosted at. Here's how it works:

    The first domains / IPs that a filter "checks" is the one included on the email body. This builds/destroy the reputation on the sender side of things. Your sending IPs, your sending domains.

    The second domains / IPs are usually the publisher's traffic redirector (sometimes for geo-location based redirection, sometimes used for traffic control -- you name it). This builds/destroy the reputation of the publisher's overall operation if these IP(s)/domain(s) are the same for every mailing system / source they have in place.

    The following domain(s)/IP(s) in the link chain are usually on the advertiser side. These resources, depending on the advertiser / network, are cached by filters in order to determine what is the email being received about. If the final link (domain and/or IP) is the same, then every publisher will run with the same luck as others. There's nothing to be learned about the sender since they have more information about the offer, and a decision is easier to be taken of whether your email goes to junk or the inbox (if accepted at all)

    So what's my point?

    The IPs and domains used at the sender, are as important as the IPs and domains of where the offer you are sending is hosted at -- and all the intermediary links. It's more likely for you to have a server side block (5xx, 4xx SMTP response) if your sending IPs/domains have a bad reputation, and it's more likely to have a client side block (hit the junk folder or blackhole) if the domain/IP where the offer is hosted at, has a bad reputation. This last one is easier to spot, and easier to fingerprint by a filter. It's also more vulnerable, since not all the mailers sending traffic to that offer will follow the good practices that you will, yet you will hit the same client side filtering rules.

    What's the solution?

    Big companies that we have the pleasure to work with, ask their advertisers to allow them to host their landers (landing pages, the offer itself), where the recipient's input is strictly required, and no robot will be able to move forward. That's the final step of the chain for filters. If a mailer controls where this final lander stays, the mailer controls his reputation all over. This has obvious advantages for the advertiser as well, since their reputation remains intact from a client side perspective. If you follow this rationale and think about the networks that host all their offers at a single IP for more than a month, then you have an idea of how bad things can get from a publisher's perspective. There are networks who are well aware of this, and they obviously have a savvy approach: They "host" landers on a per publisher basis (IPs and domains). These are the networks that live longer than a 90 day period and have a decent publisher's retention rate.

    For publishers (mailers)

    Every mailer has gone through the setup process of their mailing infrastructure. Setting up domains, setting up IPs, pairing each other... and then: a test email goes out!

    There is one previous step that MUST go before the last: Recipient domains must know about your setup existence BEFORE you ask them to accept your email. It's an extremely simple procedure, yet overlooked by a lot of mailers that want to desperately test their IPs / domains after all the hard work setting things up. Before you send that test email, send an email to your from domain(s) from your seeds. This means that:

    The recipient domain you will send emails to, will have to check on your MX records in order to deliver that first piece of email that one of their users (your seed account) is trying to reach. From the filter's perspective, its own users are interested in having communications with your sending IPs and domains. This will put your sending IP/domain in a whole different group of senders: the ones that are important enough to be reached at.

    Questions are welcome, and those who already knew about this, please add everything I'm missing.
     
  2. mrlucky123

    mrlucky123 Member

    Joined:
    May 4, 2012
    Messages:
    182
    Likes Received:
    7
    Trophy Points:
    18
    Good info, just one quick question about the redirecting. How often would you recommend to rotate the IP/ domain of the redirect?
     
  3. DAgent

    DAgent Moderator

    Joined:
    Mar 14, 2011
    Messages:
    452
    Likes Received:
    115
    Trophy Points:
    43
    Home Page:
    Although your reasoning is valid, you want real AOL / Yahoo users to try to send you an email, since the fact of setting up an FBL has a negative impact in your reputation rather than positive. FBLs should come after the back and forth communication has been built between the recipient domain and the sender domain.

    SLD + TLD is the formula they will keep in order to determine the reputation. If they capture the MXs if your domain, every IP related to that domain runs the same luck.




    This would be something you must answer based on the results you are getting. If the filtering is happening on the client side, then it's most likely time for you to rotate. If it happens on the server side, then your sending IPs/Domains are toasted.
     
  4. reddorado

    reddorado VIP

    Joined:
    Jan 24, 2012
    Messages:
    86
    Likes Received:
    33
    Trophy Points:
    0
    Thanks for this post DAgent. What is your perspective on Javascript redirects?
     
  5. ca1980

    ca1980 VIP

    Joined:
    Jan 10, 2013
    Messages:
    26
    Likes Received:
    2
    Trophy Points:
    3
    Any thoughts on sniffing out bots in the redirect chain, and blocking them?
     
  6. DAgent

    DAgent Moderator

    Joined:
    Mar 14, 2011
    Messages:
    452
    Likes Received:
    115
    Trophy Points:
    43
    Home Page:
    Both last questions are slightly out of the topic, but the principle remains: It's about where they end up (either blocked or redirected, using javascript or whatever client side mechanism you want). Wherever they stop, is what they will record about you.
     
  7. DAgent

    DAgent Moderator

    Joined:
    Mar 14, 2011
    Messages:
    452
    Likes Received:
    115
    Trophy Points:
    43
    Home Page:
    Update,

    The new CSS/DBL issues are related to new networks of robots going through your links. Find them, and redirect them properly. Both listings will go off and never come back.

    Cheers
     
  8. kaavik

    kaavik New Member

    Joined:
    Apr 1, 2014
    Messages:
    1
    Likes Received:
    0
    Trophy Points:
    0
    @DAgent, appreciate the post - particularly the differentiation of reputation control throughout the link chain.
     
  9. phdesign

    phdesign Active Member

    Joined:
    Dec 12, 2011
    Messages:
    123
    Likes Received:
    32
    Trophy Points:
    28
    Thanks for posting this DAgent - I hope to prove it useful... I've had a separate CSS issue in another thread that was TLD related. However, I recently had a bunch of severs I was running GI on get hit with CSS listings. It was mind boggling.

    I run two MTAs for GI. On one of them I have redirects in place for known bot IPs that seem primarily to be ISPs scanning the emails (I have googled a lot of the IPs). The servers running that MTA have no CSS listings (I just double checked them to be sure).

    However, I have 5 servers running a different MTA without the same redirects that all have CSS listings. I put up 3 new servers with clean IPs a couple of weeks ago and only ran my clickers (that were IW filtered), and still got listed on the first send.

    I suppose I'm now going to have to prove this by putting the redirects on a few new servers with the MTA that keeps getting listed and see if this keeps the IPs clean.

    If this is true then it seems that the same IPs that have previously been scanning the emails that I have already redirected could be a part of this network of bots and perhaps it was just an update?
     
  10. DAgent

    DAgent Moderator

    Joined:
    Mar 14, 2011
    Messages:
    452
    Likes Received:
    115
    Trophy Points:
    43
    Home Page:
    Correct. This is an addition of nets to the same filters
     

Share This Page