Martin Lanus
 

Email is as old as the Internet.
 

Actually that’s not true. In a more basic form, it predates the creation of the Internet and was partly responsible for its birth.  Along with FTP, Newsgroups and other now obscure services, it was one of the first services created to exchange messages and information between users.

Throughout the years and despite the rise of the World Wide Web, instant messaging, Web 2.0, social media and mobile; email remains one of the most widely adopted and used services online.

In recent years, Social media threatened to become the main channel of communication in terms of marketing and B2C information. But this notion started to vanish as reach in this networks started declining requiring brands and advertising to pay in order to reach their “acquired” audiences.
 

After all of this, email remains one of the most efficient and cost-effective tools in the modern marketer toolbox to engage and convert users and customers.

Because of all this, it’s no surprise that about a year ago, when I took the marketing reins at Taringa! (the largest spanish speaking community in Latinamerica), I was put in charge of activating an email marketing channel to engage with its 25+ million users.

One million emails per day

We had a blank slate to start with and only a simple goal in mind: Send 1 million emails per day in 30 days or less.

Issue #1: Reputation

The task presented several problems, the first of which was reputation. Email providers monitor senders and build a reputation based on the quality of what’s being sent out (open rates, click rates, spam complaints, etc.). Without a reputation, no email provider would allow us to reach anywhere near 1 million inboxes. Moreover when there’s a rather small window of a few hours to cram our outgoing campaigns in order to ensure a high position in the user inbox. Usually you can start with 2, 5 or 10k outgoing emails a day and start building a reputation from scratch, unless someone else has already done so and is willing to handle your outgoing mail (think of services like SendGrid or Mailchimp). But for us, borrowing on someone else’s reputation wasn’t an option, since the projected volume would make the SaaS solution too expensive in the long run.
 

Issue #2: User base

To make matters a bit more complicated we had a large portion of our users who had not been active in recent months and, if written to, we would risk complaints and low engagement number which could lead to an email provider flagging our mails as spam. So we had to be able to maintain a high engagement rate as the number of daily emails started rising. This wasn’t difficult with the first initiall 5k or 10k, but proved a little more complicated when we approached six figures numbers.

Issue #3: Content Excess

Last but not least, we had to think of the Content. What content would engage better with users? How to curate and segment content? How to get it into our newsletters and to our users inboxes?

Taringa!’s content is 100% user created and a large community ensures a constant flow of articles that generates more than 5,000 new articles every day. And that’s without counting short form content  (images, links and tweet-like texts) which is several multiples over the number of articles created.  So the issue was more of an excess of content and how to find “the needle in the haystack” so to speak.

Solution #1: The Trojan Horse

The key to deliverability (the ability to reach people’s inboxes), as stated above, is reputation. And to build a good reputation you must show email providers that your users are engaging with your content (good open and click rates, low bounces and spam complaints). We assumed that some of your users would engage with our content and another good part would engage poorly or not at all. But the problem was how to identify them. We also wanted to give inactive or dated registrations the chance to engage with our content, but that would surely entail very high bounces/spam complaints with low engagement rates.

So we had to resort to an age old military tactic: the trojan horse. That is, mix our lower engaging users with more loyal and engaging ones. But we still had to be able to identify these groups and predict engagement rates.

In the weeks prior to our launch, we tested several different things using a SaaS Email Marketing Platform. Among these we tested different segmentation rules to group our entire database based on estimated engagement rates. Several variables were used for this segmentation but what mattered most was a combination of recency of activity and registration date. Basically we could expect a higher “quality” from long dated registrations who have been active recently. Then it was only a matter of ordering the groups in terms of this two values.

When sending out the campaigns, we would initially focus on higher quality groups and started dosing the lower quality as the campaigns numbers grew. We also separated our campaigns in different groups based on actual engagement rates. As users engaged with our campaigns, we would bucket them into higher frequency campaigns and would lower frequency on emails which didn’t engage (eventually moving them out of our lists). Al off this summed provided a growing base of highly engaging campaigns which we could use to dilute our lower quality users while maintaining good overall engagement rates, solving our two first and most important issues: Reputation and user base.
 

Solution #2: Content Rating and Curation

As stated above and opposite to most cases, content quantity was a problem. Since personalization wasn’t possible at the moment, we had to find a way to select a small group of content from over 5,000 the new pieces of content created every day by our users.  

Testing was a primordial tool in this steps, since many of our guesses into which type of content would we more engaging failed. After several iterations of the format and type of content of the campaigns we find the solution by mixing implicit user responses to content on our site, segmenting based on categories affinity to the different user segments and some manual curation to pick the right “cherries”.

Today we have a strong reputation and have reached volumes of 3 or 4 million emails in one day without sweating a drop. Moreover, we can activate high volume campaigns on top of that in blink without hurting our reputation and being slapped by email providers. So our focus has turned to acquisition and further segmentation or personalization. But none of this would be possible on a large scale without deliverability, reputation and good engagement rates on a high scale.