I was humbled to be able to interview Sri Harsha of the Gmail Anti-Spam team at length recently, visiting him in person at the Google Campus in Palo Alto, it gave me the opportunity to ask a great deal of questions and get clarifications in a face-to-face setting.
I thank Google for allowing Sri to take the best part of a day to show me around and chat about Gmail and its approach to anti-spam, the ESP Coalition (ESPC) for affording me the opportunity to present Sri to the membership.
The primary purpose of the interview was to present Gmail to the ESPC and to resolve technical questions around the recently introduced feedback loop which is currently in Beta. Much useful information came out of my discussions with Sri, and there will be a number of posts around the information shared and my findings.
If you have not had the chance already you can read my original blog post on the interview. This particular post will look to dispel a few myths that have grown to have some voice online in respect of the Gmail feedback loop, with direct quotes from one of the individuals driving the FBL trials.
It is with pleasure that this information from the Gmail AntiSpam team can be shared here exclusively so that we can clear up some misconceptions and encourage wider adoption amongst the ESP community.
Myth # 1: Cannot share the information
There have been those that felt the Gmail FBL was some kind of secret agreement that was shrouded in mystery. This is not true, the program is simply in beta currently. Once an ESP is enrolled in the program they are free to share the information that is provided by Gmail if it is for the purpose of improving the quality of email emanating from their networks. As Sri said “we are absolutely fine with you funnelling that data back to your clients as evidence ”.
I got clear clarifications that any data that is shared with an ESP by Gmail can be shared with your clients for the purpose of improving the email program of your clients and/or fighting spam.
Myth # 2: No Complaints, Means my program is fine
The analogy of the FBL is like a smoke alarm “the fact the smoke alarm hasn’t gone off does not mean the given house is safe”. A client is not doing everything right just by virtue of the fact there is a lack of complaints. In Sri’s words “do not use the fact you have not got an FBL report to indicate you’re doing all the right things”, the lack of FBL reports does not mean your program is 100% safe.
Myth # 3: There are specific thresholds that I can game
When is data shared? The thresholds are dynamic, they keep changing and are modelled on past spam data using 100s of different signals, and the weight of each signal is not determined by the team, but by the algorithm. When asked what 3 things a marketer could do to improve their chance of inbox placement, Sri turned the question back on us.
“What are the top two reasons for a user marking your email as spam” he asked, explaining this is essentially the same question that a spam filter will use to make its decisions.
Myth #4: FBL enrollment means ESP is approved by Google or has an advantage
Absolutely no such thing should be implied or understood. Sri indicated that whilst Gmail would attempt to prevent abuse of the program, sooner rather than later there are definitely going to be spammers who will find their way into the program. Be this as the result of a rogue ESP or just rogue senders that are the client of an unknowing ESP. Enrollment in the program does not give an ESP an advantage in and by itself. However, not enrolling in the Feedback Loop ensures an ESP is kept in the dark as they have no visibility to complaints generated.
Myth #5 Not every report is an issue
Every FBL spam report should be a call to action, it is something that needs to be pursued. This is not a standard ARF based FBL where you get complaint reports every time a user indicates a message is spam. Gmail will only alert ESPs to senders that are causing significant issues, in the words of Sri when you receive an FBL report “you’re on the verge of real deliverability disaster” and therefore every Gmail FBL report is a real, and an urgent call to action.
Myth #6: you must sign Dual DKIM
There is still resistance from some Email solution Providers to embrace the Gmail Feedback Loop, citing a variety of technical issues or incompatibilities with their own systems. One such objection is the requirement to utilise dual-DKIM signing. In fact, there is no such requirement unless your client is signing with their own DKIM and you want that client enrolled in the FBL. You can opt to only enroll those clients that do not sign with their own DKIM keys and never be required to sign Dual DKIM. Almost all current commercial MTAs support Dual DKIM signing.
Myth #7 Reverse PTR Records
Another misconception is that domains used in email headers must all match or similar technical requirements, and that these make it impossible for many ESPs to meet the technical requirements of signing up. In my experience the technical requirements are minimal in implementing the FBL, and am proud to ensure hundreds of thousands of senders are fully enrolled. To clarify, the only requirement in respect of PTR records is this: all sending IPs must resolve to a valid hostname and have a valid PTR record, as Sri clarified “we do not insist that the PTR records resolve back to the DMARC domain”.
Myth #8: All or Nothing
ESPs are not required to enroll every single sender in the FBL program, they are free to only enroll a small element of their senders if there are technical barriers or other reasons a full implementation across the entire client base is not appropriate. If you are an email marketer and you have not been enrolled, do ask why?
Myth #9: The complaint rates make no sense
There is a common confusion with recipients of FBL reports, typical questions from clients that are contacted revolve around the high complaint rates often quoted, when compared to aggregate complaint rates, or even per ISP complaint rates. One thing that is truly unique about the Gmail Feedback Loop is that it is recorded complaints versus the number of messages delivered to the inbox (across all tabs).
This means if you send 10,000 messages, and 9000 are delivered to spam, and the balance of 1000 are delivered to the inbox, and you generate 10 complaints, this will be represented as a 1% complaint rate. This is a very useful metric, and without the ability to know exactly what your inbox penetration at any given ESP it could be seen as the most useful way of measuring your complaints. Knowing you generated 10 complaints at Yahoo, but with no indication of inbox placement means the metric can be extremely misleading.
Users cannot generally report spam when a message is delivered to the spam folder, this is true across all web based email clients. It is also generally fair to assume that if there were greater levels of inbox penetration, the complaint rate would remain at a similar level.
Come back for further posts on the feedback loop and how it is of benefit to legitimate email marketers.