Tools for Online Speech logo
by Jacob O'Bryant

The Sample publisher report

Notice for anyone using Revue: someone had an issue last week with importing their subscribers CSV from The Sample. Revue froze their account because they thought it looked suspicious. It's fairly normal for email service providers to do this, especially for new users. However this person isn't a new Revue user, and after several days—including some back-and-forth with a support person (!)—his account is still frozen.

(I set up a Revue account myself once just to check it out. After importing a CSV with exactly one subscriber on it, they froze my account. I assumed this was mainly because it was a brand new account, but maybe they're trigger-happy when it comes to CSV imports in general. They did unfreeze my account after three days without me needing to talk to support though.)

So for now, I recommend that anyone using Revue should not import their subscribers CSV from The Sample (or at least don't do it right before you were planning to send off an issue!), and I am going to write an integration for Revue ASAP so that 1-click subscribers can be imported automatically. I'm assuming that this will look less fishy to Revue than doing a CSV import.

The person who had his Revue account frozen asked if I had any recommendations for other newsletter platforms, since apparently Revue has had a bunch of deliverability problems over the past year (?) on top of this whole freezing mess. I do have a recommendation actually! Beehiiv. I tried it out several months ago while they were still in beta. It was clunky at the time, but I just played around with it again, and it looks great. It's free for up to 2,500 subscribers, and then they have some flat-fee paid tiers (e.g. the next one up is $30/month and lets you have up to 10,000 subscribers). Plus they've built a tool called Pollinate which auto-imports new subscribers from Revue into Beehiiv (it also can import into Ghost, ConvertKit, MailerLite and Mailchimp). So if you do switch to Beehiiv, you can keep using Revue for the Twitter integration.

Hey, and for anyone on Substack, feel free to check it out too, wink wink. You wouldn't have to import subs from The Sample manually any more.

(Just to be clear, I have no affiliation with Beehiiv. I just have an interest in promoting newsletter platforms that don't make things unnecessarily difficult for me/all of you who are using The Sample. Maybe going forward I'll add a little "Recommended tools" section at the bottom of these emails where I'll stick links to Beehiiv, Refind, and anything else interesting that I run into).


Other than that, we've had one significant update last week: my cofounder has developed an almost-completely-new recommendation algorithm, and so far it's beating our current one in an A/B test. So hopefully we'll switch to that fully soon. The new model is also much easier to tinker around with. It's easier to throw in different kinds of data and let the algorithm figure out what it should use and what it should ignore. For example, the current algorithm is not very intelligent about how it handles topics (i.e. the checkboxes people select when they sign up). The new algorithm might do a better job of that. We could also do things like e.g. tell the algorithm "this is a Ghost newsletter, this is a Substack newsletter, ..." and let it figure out if certain users tend to prefer Ghost newsletters over Substack newsletters or vice-versa. That kind of thing.

Also. Last week I discussed our issue of not-super-great retention and possible ways to either fix it or deal with it. Since then I think I've settled on the following for our roadmap, in addition to all the FB ad stuff + algorithm improvements:

  1. Stay focused on optimizing the product we have, without making any drastic changes. There's enough stuff in this bucket to keep me busy for a while. The main thing will likely be experimenting with different default sending frequencies (e.g. send new users one or two emails per week, with multiple newsletter recommendations batched into each email).

  2. Once we've finished doing everything there that seems worth doing (vague, I know), I'd like to make a "low-touch cross-promo directory." (See my original explanation on Discord. You might need to join the server first.) The business goal here would be to get more newsletters to join The Sample, to help The Sample cross-promote with more of those newsletters, and maybe to help me start doing newsletter ads with positive ROI.

  3. After that directory has been sufficiently polished, maybe expand The Sample's scope to include more than just discovering newsletters. If we're growing sustainably by this point, I'll probably skip this, but if we've hit a dead-end, then this could be a way out. If we are growing, I might still do this eventually, but probably not for a loong time.

Let me explain what I have in mind for that last point. I tend to think of content consumption as being in one of three different stages: discovery (i.e. you're being introduced to a new content source, like a newsletter), filtering (you consume some of the items from a content source, but not all), and subscription (you consume all, or at least most, of the items from a content source).

Most social media sites like Twitter are good at discovery and filtering but are bad at subscription. Email clients are great for subscription but aren't good at discovery or filtering. It'd be great to have a setup which handled all three stages well.

The Sample attempts to improve discovery for email. The idea I'm getting at in point #3 above is to expand The Sample to cover filtering as well. Since email clients already handle subscription wonderfully, we'd have all three stages covered. If I were to implement this, I might do it like so:

  1. Make it so each email contains a list of several different newsletters (5?) instead of just one. (As mentioned, I'll probably do this anyway in an attempt to improve retention).

  2. Add a button for each newsletter that says "keep sampling" or something. If you click it, the newsletter will be added to a list of newsletters which you are sampling (i.e. they're in the "filter" stage... I'll need to decide on some consistent terminology for this). A percentage (20%? 50%? 75%?) of your newsletter links are dedicated to newsletters in this list.

  3. Add a bookmark/read-it-later button for each newsletter issue we recommend. A certain percentage of the links we send you are dedicated to specific issues you bookmark. This is like #2, except instead of sending you new issues from a particular newsletter which you're "sampling," we keep sending you the same issue that you bookmarked until you mark it as read.

  4. Let people add bookmarks, newsletters, and other content sources from outside The Sample. We could give everyone a unique "abc132@inbox.thesample.ai" address; then you can subscribe to newsletters you find elsewhere and have them go on your sampling/filter list. We could also let people put blogs, RSS feeds, or Twitter accounts on their filter list. We could import bookmarks from Pocket and other services. etc.

The hypothesis beneath all of this is "maybe discovery alone isn't the most important problem to solve in information overload—maybe people need more help with managing all the stuff they've already discovered." A big reason I'm interested in this idea is because that statement is definitely true for me. I don't even use The Sample as a reader because I know that even if I subscribe to a newsletter I'll never read it (I already subscribe to a bunch!). So it is not all that surprising to me that our retention is kind of low, and it seems plausible that minor product tweaks won't be enough to fix it.

There's a little bit of prior art. I've already built the bookmarks part of this idea for my own use and it's been really useful, but the filtered-subscriptions part is a big missing piece that I've been pining for. Mailbrew kind of does the filtered-subscriptions thing, but it didn't quite solve the problem for me. They model everything as different blocks, e.g. you have a newsletter block, a Twitter block, a Reddit block, etc. Each block has its own settings. There were too many knobs and dials for me—I'd rather have all my content sources be treated uniformly and thrown into one big bucket, and then I could apply settings to all of them at once. But at least a lot of people do seem to get value from Mailbrew, so that's some validation, and I may be able to provide a take on the same problem that appeals to different people like myself.

The biggest downside is dilution of focus, and that's why I've decided to postpone this idea. Another risk I just thought of—what if filtered-subscriptions work too well and the business gets pulled towards that and away from the original goal of converting people to full subscriptions? I guess as long as we continue to charge only for 1-click subscribes that shouldn't be an issue. Or maybe I'm getting too far ahead of myself. 🤷‍♂️

In any case, it is not clear to me yet if this should be part of The Sample or if it should be a separate product made by someone else. I won't build it until I'm confident the answer is "yes, it should be part of The Sample."

Published 24 Jan 2022

I write an occasional newsletter
about my work and ideas.

RSS feed · Archive

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.