2011 to 2017: sequencers turn follow-up into a system, Gmail starts grading the sender

Between 2011 and 2017 sales engagement software turned follow-up into a system: write the steps once, enroll a list, and the software sends step two to everyone who did not answer step one. In the same years Gmail began grading senders by their history and by what each recipient did with their mail, so from this era on, what your domain did last month shapes where today's email lands. This part covers both halves, and why the current versions of the authentication standards you are now required to publish all date from this window.

What did a sequencer change about cold email?

Before sequencers, a follow-up was a rep remembering to write one. Outreach started in 2014, when a struggling Seattle recruiting startup turned the tool it had built for its own prospecting into a product. Apollo was founded in 2015 and went through Y Combinator the following winter. Woodpecker was created in 2015. Together with Reply, they sold versions of one idea: the sequence. A sequence is a set of timed steps, each one an email template with merge fields, sent until the person replies, opts out or reaches the end.

For a sales team the gain was real. Follow-up stopped depending on memory. A manager could see every step, test two versions of step one against each other, and hand a new rep a working sequence on day one. One rep could keep hundreds of people in motion at the same time.

For the inbox, the same feature looked different. Picture Dana Reyes, head of sales at Northwind, on a Tuesday morning in 2016. Four vendors' sequences reach step three that day. Each one is "bumping this to the top of your inbox." Each one was written once and sent to thousands of people. A message repeated at that scale is exactly what a learning filter is built to notice.

How did Gmail start grading the sender?

In May 2013 Gmail introduced the tabbed inbox, which sorted mail into categories and put deals and offers in a tab of their own. A large provider was now deciding, on the reader's behalf, which commercial mail deserved the first screen.

In July 2015 Google described the next step, and it changed what a sender is up against.

For starters, the spam filter now uses an artificial neural network.

The same post said filtering would now reflect individual preferences: one person's welcome newsletter is another person's junk. Every click on "Report spam" or "Not spam" trains the filter for the future. And it launched Postmaster Tools, which gives high-volume senders data on delivery errors, spam reports and reputation. Google also put numbers on where it stood: less than 0.1% of the email in the average Gmail inbox was spam, and wanted mail landing in the spam folder was under 0.05%.

Read those three changes together. The filter learns from what each recipient does. Placement can differ from one reader to the next. And the provider now keeps a record on the sender and shows it to them. From 2015 the question was no longer only "is this message spam?" It was also "what has this sender done before?"

Were SPF, DKIM and DMARC around in this era?

Yes. The current versions of the standards that Gmail and Yahoo now require from bulk senders were all published in these years, and every one of them was optional at the time.

A sequencer could send without any of it, and many senders skipped it. Nine years after DMARC was published, the same standards became the entry requirement for anyone sending at volume to Gmail and Yahoo users. Part 5 covers how that happened and what it means for a small team.

What era two means for a head of sales in 2026

• Reputation belongs to the domain, not the campaign. A sequence you switched off last quarter is still in the record the provider keeps. A new campaign inherits it.
• A follow-up is a new decision. Step three to someone who ignored steps one and two adds to the history the filter reads, and so does every spam report it earns. Send it because something changed, not because day nine arrived. Part 11 covers how to learn the follow-up rhythm per audience instead of copying a default.
• Placement is personal. Since 2015 the same email can land in Dana's inbox and her colleague's spam folder. An open rate is an average of many private decisions, which is one reason it is a weak measure.
• When replies drop, look at the sender first. The reputation layer sits underneath the message, so the diagnostic runbook in part 12 starts with sender and domain health before anyone rewrites copy.

A note on what anyone outside a mailbox provider can know. Google publishes which signals it measures and which dashboards senders can see. It does not publish how its models weigh them. Everything this series says about filters is built from what the providers have said in public, not from guesses about their internals.

Next in the series

Part 3 picks up in 2018, when enrichment and contact data plugged straight into the sequencer, and mailbox providers answered by leaning on identity: domain, IP, authentication and recipient feedback, read together. Part 1, on how sending at scale began, sets up the pattern. The series page lists every part in order.