Unfortunately, we do not have the concrete answers you are looking for in this request, as the "speed" messages are delivered at are highly situational. Mailbox providers don't measure messages in a speed, per se, but rather in the number of concurrent SMTP connections they allow at any given time, and the number of messages they accept over that connection. The number of connections a mailbox provider will accept, and the number of messages they will accept over that connection are based primarily on your reputation as a sender. For senders with a good reputation, they can eventually "max" out the processing volume of a single IP address when volumes reach greater than ~500k messages per hour. Our largest customers process at peak about 250,000 messages an hour to Gmail using 25 simultaneous connections on a single IP address. That is a little less than 3 msgs/sec per connection.
Since SocketLabs is managing all of the outbound connections for messages injected to our platform, it is our responsibility to ensure that the messages are being delivered as quickly as recipient domains will accept them. We recommend customers get messages over to our platform as quickly as possible, and let us manage how quickly they process out. We will never process messages from a technical perspective in a means that would negatively impact deliverability or sender reputation. If you suddenly inject 2 million messages in an hour, we will traffic shape delivery to optimize deliverability, even if that means it takes a few hours to process all of the messages. In regards to SMTP connections to our inbound gateway - we recommend opening 30 simultaneous connections and processing 50 messages per each connection. If you find yourselves maxing out the 30 connections and hitting SMTP errors like 421 4.4.5 Too Many Connections, then please let me know and we can make some custom arrangements for increased settings.
There are few scenarios where we would hold back delivering messages:
- Connection limits being reached at a particular recipient domain.
- Reputation issues that cause a provider to limit the connections they will accept from you, we automatically back-off using intelligent error handling. We call these our Smart Delivery rules, and we've been curating them for over 15 years as part of the "secret sauce" that allows our platform to out-perform an off-the-shelf MTA product.
- The Warm Up Period
There isn't necessarily specific speeds that you would want to be cautious of in regards to sending messages, but there are some guidelines for total hourly/daily volume. Drastic volume changes can be a red flag for providers, as this is a common sign of compromised mail systems. Reviewing server 4015, for the past 90 days, this server has averaged an hourly peak of 60,000 messages. To maintain your current sending patterns, I would put the upper bounds of a recommended injection rate at around 120,000 messages per hour (about 200% of average hourly peak from the last 90 days). If your volumes grow and you start injecting mail to us more rapidly, you could easily increase this over time with the quality of your reputation.
Message content is usually driven by decisions unrelated to message delivery speed, so this is also hard to quantify. The actual message size (in terms of raw bytes) impacts delivery speed simply from a bandwidth perspective. Some providers (i.e. office 365) are known for being stingy with the bandwidth they allot to individual connections. Beyond message size, the number of other items often used in spam analysis can theoretically impact delivery speed. For example, it is safe to say that messages with a lot of different URLs contained within the HTML can take longer to be scanned by the recipient mail system than a message that contains a single URL. Attachments that tend to be spammy may go through more rigorous scanning as well so having attachments like .xls may take longer for the messages to be scanned for spam and delivered to the end recipient. I'm not sure your going to see any kind of noticeable increases in delivery speed by modifying content of messages unless you are significantly reducing overall size.
Reputation is obviously the biggest factor in reaching the highest delivery rates. Reputation today is mostly dictated by engagement and recipient reaction. Recipient reaction is slightly harder to quantify. Open and clicks can be easily quantified, as well as unsubscribes and complaints, and we have tools that allow you to monitor these reactions. That said, spam detection systems are evolving, using machine learning to identify what is or isn't desired mail. These systems are being taught by recipients what mail they want to receive, and what they don't want, through both action, and inactions. We discuss this at length in our blog, found here:
Other practices would be to follow list management best practices (removing bad addresses, segmenting lists so you are sending to customers who are engaged with the brand, etc.) and acceptable address acquisition methods.
Additionally, configuring our custom authentication features, Advanced DKIM Signing and Custom Bounce Domains, can help avoid deliverability issues by white labeling and authenticating SocketLabs as an authorized sender of a particular domain's messages.