A few months ago, Google made a splash in the political press and the email marketing space when they asked the FEC the following question:
May Google launch a free and non-partisan pilot program to test Gmail design features, which will be open to authorized candidate committees, political party committees, and leadership political action committees, where spam detection as applied to messages from a pilot participant on direct feedback from the recipient rather than standard spam detection, and each pilot participant will receive information regarding the rate of emails delivered into Gmail users’ inboxes, as long as the pilot will rely predominantly participant is in compliance with the program’s requirements?Google’s letter to the FEC (.pdf link)
The letter is actually worth a read as many of the general press reports about the request focused on Google asking the FEC to allow politicians to spam freely. I mostly avoided discussions about it because it seemed that the popular takes were misrepresenting… well, everything.
The press continued its political spam is coming frenzy when the FEC determined that this pilot program was allowed. Again, much of what was actually reported was mostly about how this was going to result in tons of spam for recipients.
I mostly stayed off the bandwagon then, too, because so much of the press wasn’t clear what was really going on. In their desire for clicks, they didn’t (to my mind) correctly report what Google was asking. Then they failed to report what the FEC said.
The Commission found that, while Google would be modifying its service only for certain political committees, it would be doing so for commercial as opposed to political reasons. […] Similarly, the Commission found Google’s pilot program would serve its commercial interests in protecting its brand reputation and obtaining valuable data on how to enhance its product. Also, based on the results of the pilot program, Google may extend it to others, including government entities, entities involved in providing government services, senders of class-action notices and non-profit organizations.FEC Advisory opinion
As with most things in the email deliverability space, the devil is in the details, so I decided to wait until Google announced their program rules before drawing any conclusions or making any predictions. This week Google announced the Verified Political Sender Program. There are multiple pages and FAQs that I encourage folks to read if you’re interested in all the details. (Program rules and eligibility requirements, Program FAQs)
Reading between the lines, Google is making some filter modifications that will rely more heavily on their user specific filters rather than their global filters.
Brief Segue into Filter Theory
Some of you might have heard me talk about my current mental model of filters at consumer domains. I’m primarily talking about Gmail, but I think it applies to Microsoft and Yahoo as well. Basically there’s two levels of spam filters.
The first is a global spam filter that looks at the overall reputation of a mail stream and decides if the message is in one of two categories: more likely wanted and more likely unwanted. That filter is used to deliver the message for recipients where the ISP doesn’t have any, or much user interaction data. Maybe it’s the first time you mailed a user, or maybe that user doesn’t use the web interface so there’s less information about the user’s preferences. This is the filter we, as deliverability folks, are influencing when we’re repairing reputation. Our advice sets senders up so that the default decision on their mail is: deliver to inbox.
The second is a user specific spam filter. This filter looks at the email usage profile of the final recipient and has picture of their preferences for email. Maybe this user has marked mail from this sender as spam in the past or they always delete this sender’s mail without reading. Or this user has a specific filter directing the mail to a folder or the sender is in their address book. These filters can override the global decision. Mail that might otherwise go to spam may go to the inbox of some users while mail that might otherwise go to the inbox may go to spam for different users. This is a filter that relies on sending mail that the end recipient actually wants to see and is influenced by good marketing (rather than deliverability) practices.
Back to the Gmail Political Mail Program
What is explicit about this program is: Google will be managing the user choice to opt-in or out of mail from a very specific group of senders that are registered with the FEC and approved by Google. These senders will need to meet specific criteria for authentication and security.
To determine eligibility for the program, Google is piggybacking on the authentication and rules of the FEC. Only authorised candidate committees, political party committees, or leadership PACs are eligible. Despite the sturm und drang about how this was going to be open to all political senders, that’s not what’s happening here. It’s open to candidates and their committees not every sender that defines themselves as political. It’s not open season on inboxes the way some of my colleagues and the press were screaming was going to happen.
Additionally, they’re imposing some pretty strict requirements on the security of the domain and the authentication of the messages.
- All messages must be protected with DMARC policy of quarantine or reject and both SPF and DKIM must be aligned to that domain. This is a fairly high bar and one that will have some ESPs scrambling to update their ability to support full alignment.
- They’re requiring protection against forgery of non-existent subdomains. If a committee doesn’t want their primary domain to have strict DMARC policy, they can set up a sp= policy that applies to subdomains.
- They’re requiring TLS 1.2 for all mail.
- All emails must have a list-unsubscribe header.
Participants in the pilot program are limited to 5 domains that are owned (in whole or part) by the committee and are under the committee’s control. Domains also must be registered for a minimum of 30 days. Confirmation of registration is being sent to the email address on the FEC registration forms, meaning that Google is using the FEC registration process to determine who is eligible.
All of these requirements are on top of the standard Google best practices. This includes getting consent from users, setting expectations, and confirming recipient’s email addresses before sending.
Finally, Google states that a participant with a complaint rate higher than 5% will have all of their domains suspended from the program for a period of one week. Now, to me, this is a very generous complaint rate. I have rarely seen any sender with a complaint rate that high at Google. In fact, I start talking to my clients when I see their complaint rates are higher than 0.3% for more than a few days. Even my current “bad” clients are showing complaint rates at below 2%.
These authentication rules don’t seem too onerous, but how many political senders have published p=reject or p=quarantine policies? Of those that haven’t how many are in a position to do so? Is it even a good idea for them to do so? DMARC is a big step that can affect all mail, not just marketing mail, and is one that political committees might not want to take 60 days before the midterms.
What do senders get?
In return for meeting all these requirements, what do senders get? The answers are buried in the FAQ. My interpretation is that senders get exempted from the global spam filters I talked about above. Instead of going through Google’s standard delivery process, all delivery decisions are going to be user specific. I’m basing this on reading lots of stuff from Google on the problem, including the answers in the FAQ.
How do I prevent my emails from being sent to spam?
If your Committee is approved for the Gmail Verified Sender Program Pilot, when following the message-based delivery requirements and Gmail’s terms of service, your Committee’s emails to Gmail users will not be affected by some forms of spam detection to which you would otherwise be subject.
You can also check the Postmaster Tools dashboard for deliverability metrics that inform how your emails are filtered in Gmail.Verified Program Participant FAQ
This is the piece that leads me to believe that all mail from registered senders will be treated globally as wanted mail. The only filters to be applied will be those based on user engagement.
How will Google determine user engagement with the message?
This is where it gets a little interesting. Buried in the FAQ under the question: Can users unsubscribe or resubscribe to a verified sender’s email? is this little tidbit:
In the official Gmail app/client, a banner will be shown to a user on the first message they read from each pilot participant’s domain. The banner will prompt users to choose to continue receiving emails from the sender or to unsubscribe. If they choose to unsubscribe, they may also choose to report as spam.
Pilot participants are required to honor unsubscribe requests within 24 hours. If the user chooses to mark the email as spam, the user will no longer receive emails in their inbox from that domain, but instead in their spam folder. If the user subsequently marks a message as not spam, the user will immediately begin receiving emails from that domain in the inbox going forward. Users will also be able to use Gmail’s existing tools to mark a message as spam or unsubscribe.Verified Political Program FAQ
As I read this, Gmail will be asking every recipient mailed by participants in the program if they want to continue receiving mail from that participant. If the user unsubscribes, Google expects the sender to honor that in 24 hours. If the user chooses to report the mail as spam, then future mail from that sender will go to the spam folder.
Google has now inserted themselves in the opt-in process for registered senders. Instead of making guesses about whether or not mail from a sender is wanted or not, Google is asking each and every recipient if they want to receive mail from them. If they don’t, then the user has the option of reporting the message as spam. Any future delivery of mail to the spam folder will specifically be because the user has told Google the mail is spam.
My analysis of the program
Here’s what I’m currently thinking about what is going on here. I talked above about my current mental model of spam filters.
- Google is letting a select group of senders, registered with the FEC, opt out of their global spam filters.
- In return, Google is asking every single recipient of that mail if they want to continue receiving the mail, unsubscribe from the mail or unsubscribe and report the mail as spam.
- Future delivery decisions during the pilot will not be mediated by Google’s global filter, but rather by the individual user filters. For those users who are using a 3rd party mail client they will continue to receive the mail in their inbox (subject to filters in their personal mail client).
What I’m thinking is that for domains registered in the program, global reputation will be calculated. But instead of that reputation being used to determine delivery, all delivery will be determined by the user specific rules.
Google is pretty carefully not mentioning what tab the mail will be delivered to. Mostly they talk about mail not going to spam rather than being delivered to the inbox. We know all tabs are considered the inbox by Google so I have to wonder if there are plans for a special tab just for this mail?
What about after the pilot program?
There’s another little tidbit buried in the FAQ that make me want to pull out my Admiral Akbar gif. One of the questions is Will my domain reputation be impacted if I participate in the pilot?
Spam markings during the duration of the pilot will impact domain reputation following the conclusion of, or the sender’s removal from, the pilot. We recommend following Gmail’s guidelines for bulk senders to ensure email messages are delivered as expected to Gmail users.Verified Political Program FAQ
As I’m reading this and mentally modelling the filters, it’s seeming more and more that Google is basically going to route all the mail from registered domains to the inbox for a one time re-confirmation mediated by Google. Delivery after that point will be wholly dependent on whether or not the recipient opted in or out of the mail from that sender.
Overall, this strikes me as a sensible, possibly even crafty, move by Google.
- They pulled the rug out from politicians that were misrepresenting the results of a study on algorithm bias (.pdf link) in consumer mailbox filters. I actually encourage folks to go read the study, it’s quite well done and there are some interesting insights unrelated to the bias question.
- They get a constrained research project looking at how well their filters reflect what the users want. I have zero doubt that the emails in this project will be scored through the standard Google algorithm and then delivered based on the pilot program conditions. Now Google can see what complaints and user engagement look like for mail that they think is spam but have delivered to the inbox anyway. Even better? They can even monitor unsubscribe rates through the List-Unsubscribe header clicks. This is a treasure trove of data and the scientist in me is envious of the folks who get to look at this data.
- They moved the authentication ratchet one more click by requiring both SPF and DKIM alignment as well as strict DMARC policies.
I expect that we’re going to see multiple published papers on the results of this, at least I hope we do. I also expect that the true impact of “let the users decide” is going to surprise a lot of program participants.
We have 2 weeks before the end of the month fundraising push that the RNC says performs badly due to Google putting all their mail into spam. That’s enough time for the RNC to get registered and set up. I can hardly wait to see what happens when their normal end of the month blast goes to the inbox of all those Google recipients. Is the RNC right? Will they see a massive increase in their donations compared to other months? Or will they hit the 5% complaint level and get themselves suspended from the program?
Only time will tell what the outcomes are. I have my suspicions, I’m sure Google has well informed opinions that drove some of their policies here.