Future of deliverability: 3. The role of domain reputation
Latest posts | Feed | | By Mark Brownlow
Part 1: User interactionPart 2: Authentication
When organizations look at incoming email, they use a set of criteria to decide what to do with it.
For many such organizations, particularly the big webmail services, the reputation of the sender is a very important criterion determining whether that email should go to the inbox.
This sender reputation is itself built out of various factors, such as how many spam complaints the sender gets or how many defunct addresses they are trying to email.
It's a pretty good way of regulating email, but problems arise through the definition of the "sender" part of sender reputation.
To date, this reputation has largely been tied not to the sender in the traditional sense of the word, but to the sending IP address: the original "connection" to the net that initiated the email transfer.
This gives rise to various difficulties.
For example, if different organizations send email through the same "connection" (sharing an IP address at an email marketing service, for example) then they also share a common sender reputation.
So a "good" email sender can find their reputation dragged down by the others. And vice versa.
Even if you have a dedicated IP address all to your own, you can face problems if you move to a new one. Your sender reputation isn't portable: you have to start from scratch.
If we could associate reputation with an actual domain (e.g. news.email-marketing-reports.com), then reputation would become independent of the system/location used to send out that domain's emails.
It wouldn't matter where my emails are sent out from, because the reputation factors associated with those emails would be tied to the news.email-marketing-report.com domain name.
I wouldn't have to worry about naughty senders putting out their email through the same "connection" as me: it wouldn't affect my domain-based reputation.
This concept is called domain-based sender reputation and it's already impacting the email deliverability world.
Unfortunately, the reality is not quite as simple and positive as the above theory suggests. Nevertheless, the potential benefits are clear.
Deirdre Baird, President & CEO of Pivotal Veracity told me:
"...domain-based reputation will be of big help in addressing the pain-points associated with sharing IP addresses...furthermore, large mailers will have the benefit of being able to add new IPs and/or switch IPs (or ESPs) without the painful and blocking-fraught 'IP warm up process'."
Domain-based reputation works both ways, of course, as George Bilbrey (President) and Tom Sather (Professional Services Director) of Return Path explain:
"If you are a good mailer, sharing IP space with less-good mailers, you no longer pay the penalty for their practices. But if you are a bad mailer sharing IP space with good mailers, no more free ride."
Is domain-based reputation relevant today?
Understandably, ISPs and others are cagey about providing details, but domain reputation is already a factor at some organizations and likely to spread rapidly.
Baird reveals:
"...many leading providers are already in the process of moving to domain-based reputation. Yahoo is already enabling DKIM-compliant senders to benefit from domain-based reputation portability. AOL is switching later in Q4/early Q1."
Bilbrey and Sather confirm that the impacts are already being felt, noting that the top-tier ESPs they work with have started to see some domains get treated differently off the same IP addresses.
Jeremy Saibil, Director of Deliverability at Campaigner offers further confirmation:
"I believe we'll be seeing domain-based reputation "kick in" almost immediately. In fact, I have suspicions it has already started; certainly "warming up" new DKIM authenticated IPs has been much easier of late with certain ISPs."
These quotes highlight the importance of authentication, particularly DKIM (see Part 2 for details).
Domain reputation only works if the receiver can verify whether the domain claiming to be the sender is indeed the source of the email.
The implication is clear. If you want to benefit from domain-based sender reputation, then authenticate your messages. Equally, since not everyone will do so, "traditional" IP-based reputation will continue to play a role in deliverability (more on that later).
Domain-based reputation is not a "get out of jail free" card
Of course, the advantages of domain-based reputation only accrue if you have a good one (a point often overlooked).
Bilbrey and Sather note that the metrics that make up the current IP reputation systems are going to be the same for domain reputation: complaints, unknown users, spam trap hits, sending consistency, "this is not spam" votes, opens and clicks:
"Bottom line: if you have a bad IP reputation because of poor practices, domain reputation is not going to improve your inbox placement rates."
They also warn that the mix of factors and thresholds for these factors will also likely change as ISPs adapt their approaches:
"What defines a "good" complaint rate, for example, is not static and never will be."
So marketers need to stay on top of new developments.
Will it replace IP-based sender reputation?
A question that's hard to answer is the extent to which domain-based reputation will truly replace the current role played by IP-based reputation.
One thing that's certain: the two will inevitably coexist for the foreseeable future. Chris Wheeler, Director of Deliverability at Bronto, says:
"...IP reputation will still be used by many ISPs while domain reputation gains momentum. Also, IP and domain reputation heuristics will not be mutually exclusive."
Instead, he says, domain-based and IP-based reputation will blend together in determining the final destination of delivered email.
Baird also believes IP-based reputation will remain important:
"...the switch to domain-based reputation is just beginning, and many smaller ISPs haven't implemented DKIM or SPF yet. Additionally, we don't expect IP-based reputation to go away entirely as it is used when no domain-reputation exists even at the largest ISPs."
Bilbrey and Sather add:
"The truth is that ISPs are using both domain and IP reputation. And that is not going to change anytime soon. ISPs will still be looking at reputation metrics for IPs, IP-ranges, URLs and more."
So domain-based reputation isn't a global panacea to the issues surrounding shared IPs. It makes you more accountable for your actions, but you won't become completely independent of other senders' email activities if you share an IP address with them.
Bilbrey and Sather continue:
"A high-performing marketer with a stellar domain reputation that is sharing an IP with bottom-of-the-barrel spammers is still likely to see issues getting to the inbox, even with domain reputation in place at both ends of the pipe."
For the individual marketers, it's important to know if your ESP is using DKIM authentication. If not, as Bilbrey and Sather note, today's rules continue to apply because...
"...in the absence of DKIM, ISPs are going to fall back to IP reputation."
Another issue is exactly how DKIM is implemented. Saibil explains:
"...it will be interesting to see how ESPs roll out DKIM/DK...For shared IP environments that support hundreds of thousands of small customers, will it be feasible for each customer to sign as well as the ESP? Perhaps eventually, but I suspect the majority of those emails will be signed by the ESP only for some time to come."
Should different email types be sent from different domains?
Domain-based sender reputation brings accountability, which is a welcome development if you're a good sender. But even the best senders can run into temporary deliverability problems.
If your domain reputation takes a hit, then all email from that domain might struggle to get to inboxes.
Does it make sense, then, to allocate different message types to different (sub-)domains?
Saibil notes that those using an ESP for marketing email likely already have this domain split:
"Using an ESP generally requires a separate domain entity so the ESP can do reply processing."
...but he adds:
"Splitting corporate, transactional and marketing mails at a domain level is something I recommend, however, we usually suggest using sub-domains for these splits."
Bilbrey and Sather go into further detail:
"There is a string in the DKIM record, d=, which is what reputation is tied to. So you can use different d= strings (examples: d= marketing.returnpath.net, d=sales.returnpath.net, d= service.returnpath.net) to differentiate various streams even though the recipient still just sees "returnpath.net"."
Baird warns, however, that while email authentication can help make the process of separating mail streams more efficient and clearly defined, marketers should not rely on authentication alone.
After all, as we just learned, domain-based reputation isn't the only factor affecting deliverability.
"There should be physical separations between corporate mail and all other mail. If a marketer's mail stream comes under fire and you're blocked, your corporate mail may be the only line of communication to a filtering body/ISP. At the very least it's important to leave yourself a life raft in case the whole ship goes down."
Wheeler also points us to this detailed post on domain-based reputation, where he recommends auditing all your outgoing email to identify any potential conflicts:
"If Bob over in coupon sales is killing it with blasting recipients twice daily and aggravating them, Sally over in order processing trying to get shipment confirmation emails out will suffer if both mail streams are coming from the same domain."
Part 4: Email certification
More on deliverability
Permalink | October 28, 2009 | 2 comment(s) - add yours!
Get posts like this: as an RSS feed | biweekly email | via Twitter
Twice a month, free, packed with email marketing advice and all the posts from this blog.
2 Comments:
Hi Mark....
Great post.
I understand that deliverability is also tied to volume.
So a smaller sender with their own IP will not rank as high as a larger sender with their own IP.
This is not relevant for companies that send millions of messages a month, but how does a company that sends 10,000 messages a month begin to build reputation and compete for ISP approval if they are being scored on their volume of sends as well.
For them being in a shard pool helps a lot. If they are small fish in a big pool of delivery then their volume reputation will also go up.
I think dedicated IP’s will cause some deliverability issues for smaller fish.
On that note, what are you your thought about a pay4 system. Kind of like the way postage stamps work?
Jeff Ginsberg
Chief eMail Officer
www.theeMailguide.com
By Chief Email Officer, on
30 October, 2009
That's a key point Jeff. Shared IPs - assuming their reputation is managed properly - allow small senders to have reputation in the first place.
A similar issue might apply to domain-based reputation. If you're not sending enough email, you won't generate any reputation. Which is why the shared concept I guess will remain important.
As for email stamps. Not sure, but see Part 4 on certification for expert opinion.
By , on
30 October, 2009



