Jacob O'Bryant
The Sample: solving discovery for newsletters

24 February 2021

I recently launched The Sample. Like Findka Essays (my main project until now), it’s a newsletter curated by machine learning. You sign up and then an algorithm decides what to send you. But instead of links to essays, you get the latest issue from some newsletter.

I’ve set up an inbound-only mail server so that I can easily generate unique email addresses (like abc123@sample.findka.com) and write code to handle any emails sent to those addresses. To get things rolling, I did some Googling and compiled an initial list of about 80 newsletters to import. I sign up for each one manually with a unique address, and then all the emails from that newsletter get stored in The Sample’s database.

After you sign up, we ask what topics you’re interested in. That gives us a good starting point for what to send you, but the algorithm continues to adapt. It pays attention to which newsletters you engage with (by clicking links), and there’s a link at the top of every newsletter which lets you give a 1-to-5 star rating.

Of course, none of this will matter much unless The Sample can get lots of subscribers. Will this actually grow? I think so, for a few reasons. First, the format is extremely convenient for consumers. It’s just another newsletter, and signing up for newsletters is becoming a habit for lots of people. (Newsletter directories don’t share this advantage: only power users will go out of their way to look through a directory).

On top of that, The Sample was designed for cross-promotion. It turns out that one of the highest-converting ways to promote a newsletter is to get a shout-out from another newsletter. If you’ve signed up for one newsletter already, you’re a person who signs up for newsletters, and that’s a big factor in how effective an ad will be.

Traditionally, when cross-promoting, you have to find another newsletter that has about the same number of subscribers and has a related topic. So the set of newsletters with which you can effectively cross-promote is constrained. But The Sample gets around these constraints, partially.

If newsletters were Pokemon, The Sample would be Ditto. The algorithm can create different audience segments on the fly, without human direction. When cross-promoting some other newsletter, the algorithm can create a segment just for that newsletter.

For example: say The Sample has 2,000 subscribers, and 600 of those people are interested in marketing. If you have a marketing newsletter with 400 subscribers, then we can cross-promote: the algorithm will create a segment of 400 people who are interested in marketing, and it’ll forward the latest issue of your newsletter to them. No matter how large The Sample grows, it can still effectively cross-promote with smaller newsletters. In fact, the more subscribers The Sample has, the higher the chance is that it’ll be able to create the perfect segment for your newsletter.

This is in fact the main reason I decided to make The Sample instead of continuing to work solely on Findka Essays: it can soak up cross-promotion like a sponge.

View discussions