Substack: What's in a Recommendation?

Insight: On Creating Awareness and Leveraging Platforms.

Photo by Juni Shi on Unsplash

Do you remember the days of seeking letters of recommendation from those who know you best for your undergraduate or graduate school applications? Or, perhaps, when employers asked for those mandatory references at the end of successful job interviews?

Well, if some of you are under the impression that's what Substack means by "Recommendations," I'd like to share a different view with you.

Since I've read notes, comments, and posts, advocating strongly for and against recommendations, I thought this post might help to clarify the strategy.

The What, Why and Why Not

The Recommendations feature on Substack is a marketing tool designed to increase awareness of your product—your publication. Recommendations don’t have to be only from those writers who know and can vouch for your publication.

Its sole aim is to promote writers within the community of readers and other writers.

The feature applies an algorithmic bias to recommend specific publications to you and, in turn, recommend yours to others.

The bias is driven by business logic that may operate across multiple views as below.

Key Attributes Considered: Number of Subscribers, Monetization Levels, Genre, Topic Trends, Celebrity or Bestseller Status (e.g. Stephen Fry), Follows, Frequency of Posts, Engagement levels etc.

Disclaimer: This is my conjecture on a possible strategy; the Substack algorithm may be entirely different.

A Reader View

Let's take a sample Reader 1 of Publication D. Consider her engagement data:

1. She reads publications A, B, and C, which respectively cover Science Fiction, Spirituality, and History.

2. She also reads and engages heavily with publications D, E, and F, all within the Literature genre.

3. Weighting it, it may conclude that she mostly reads D, E, and F (70%) and sometimes, reads A (20%), with B and C receiving equal attention (10%).

4. Therefore, find similar publications that are: a) trending, b) active, c) monetized (paid subscribers), d) open to recommending others, and e) falling under categories D, E, and F. It may then apply a weightage of 70% to these categories, distributing the rest among A, B, and C based on her reading habits data.

A Writer View

Consider the data from a writer’s perspective:

1. This publication currently boasts 200 subscribers, with 30% of the audience also engaged in Publications X, Y, and Z at high levels. Another 20% are reading A, B, and C with moderate engagement, while the remainder are exploring random publications suggested during signups.

2. So, let's group all of the writer’s existing recommendations, organize them according to the Writer View criteria (Step 1), and apply a random generator to propose new publications to potential new subscribers of this publication.

3. Suggest to the writer to expand their list of recommendations. This will enable the software to improve its predictive capabilities, and hopefully, increase the number of monetized publications, thereby, increasing the odds of paid subscriptions.

   Use case: Considering this writer already enjoys high monetization levels, akin to 'Letters from An American', and boasts impressive conversion rates, it would be prudent to suggest this publication every time someone expresses interest in 'Politics' or is perusing random political publications.

4. By executing this strategy effectively, subscribers will encounter a diverse array of writers and topics, thus bolstering their inclination to become paid subscribers and remain active on the platform.

5. To aid writers in making informed choices, let’s encourage them to seek recommendations of fellow writers. Writers, if you noticed, don’t have to provide justifications - it is optional. This is to prevent the delay in the execution of algorithm-based predictions in real-time, which could otherwise be blocked by the failure to write a blurb in time.

6. If a user writes an endorsement, they may be rewarded by a boost their algorithmically derived score by 10 points perhaps. This ensures that publications with a solid 'Why' behind them have additional social proof, factored into the final bias calculation.

Follows in Recommendations

Substack has also just enabled follows in recommendations, as seen in your notifications. This effectively suggests to a subscriber of A, that ‘A follows C,’ so you may want to ‘follow C too.’ This ‘recommendation’ of whom to follow then allows the transitive nature of an attribute (A→B→C), and the known psychological factor of social FOMO (fear of missing out) do its job. It also reinforces credibility in the system - if you (A), like writer B, and writer B follows writer C, then perhaps there is value in following writer C?

Yes, social media-driven biases may creep in over time unless the algorithm is monitored and user observes some discretion in following others.

What does this all mean to you as a writer?

Well, Substack has its own internal secret sauce of recommendations, which could be a version of the above explanation or something entirely different, given their access to vast amounts of data about you.

Since you are a writer looking for monetization (hopefully) now, or down the road, you should review the recommendations you are being asked to endorse in your Dahsboard→Settings->Recommendations view.

If you have time, then read the recommended publication’s About section, and some of their free articles, to get a sense of their alignment with your views, beliefs, and recommendation standards.

Ideally, most recommended publications should be in a neutral state, focused on their niche or general areas of interest, with no intent to rock the boat.

Endorse them if you're comfortable with their writing style and content by selecting ‘Add'.’ If you're pressed for time, endorse them, and wait for any red flags to surface - something is bound to sooner or later.

Remember, the algorithm is a random generator of your recommendations list. If it is handled well at the business logic level, as I outlined above, a larger number of subscribers or future readers will be shown your publication over time to follow, read, or subscribe based on where they are in their workflow on the platform.

Think of the user engaged states this way:

• Follower: Browsing, may read or not. May just see your Notes.

• Subscriber: Curious, reads, engages, or is indifferent and may leave.

• Paid Subscriber: Reads, supports, engages, and seeks exposure to all paid benefits.

• Founding or Pay Anything Subscriber: Invested and acts as a sponsor.

Visualizing this as a marketing funnel with four different-sized buckets can help you better understand and navigate your recommendation and engagement strategy.

Remember, once they become a subscriber, you have additional opportunities to demonstrate value to them through your writing.

Reader Funnel for Publications ©Jayshree Gururaj

Note, free subscribers can also directly convert to founding members, so technically this should show three layers, but it illustrates the principle better as paid subscribers, after a year, may convert into founding members.

What does all this mean for me?

If you are on Substack to grow fast and monetize, then the smartest approach would be to build the largest possible funnel of readers interested in your work.

That entails leveraging any support available from the Substack platform to help create this funnel for you through recommendations.

Beyond the Substack platform, you'll need to invest in digital advertising, including Google Ads, social media outreach, and collaborations, to drive traffic to your site and expand your funnel.

So what should the approach be?

Tier your recommendations for other publications as the starting point.

Tier 1: If you're familiar with certain writers and appreciate their work as peers, or have benefited from it, then consider them your Tier 1 set and write them an authentic recommendation blurb. Upto three blurbs, can be used to offer testimonials to potential subscribers displayed on at signup. Below are the ones I display on my welcome page. You can rotate the blurbs as well so it is nice to have multiple blurbs.

Thanks

Red Solo Traveler

Jane Hutcheon

and

Projectkin Ringleader

and the dozens of others who have written a blurb, recommending my work. I am grateful for your kind words.

Tier 2: Group the remaining publications into Tier 2 until you can promote some publications to Tier 1 as you become more acquainted with their work. Recommend all publications in Tier 2 by adding them to your ‘recommend’ list.

Tier 3: Exclude writers and publications that you don't align with from Tier 3. Simply leave them be on the list, i.e.. – don't ‘Add’ or recommend them if you feel they shouldn't be promoted to your network.

Tier 4: Block writers and publications in Tier 4 from ever appearing in your feed, recommendations list, etc., if you oppose them on specific grounds.

Writers do reciprocate the recommendations and blurbs if there is alignment, and they have a similar approach. So it is a win-win.

By following this mental model, over time, the algorithm should become smarter and serve you better recommendations - more names in Tier 1 and Tier 2, while reducing Tier 3, and eliminating Tier 4.

Success Measures

These are quite simple:

1. The number of publications recommending you significantly influences your subscriber acquisition. In my recent stats, 119 subscribers were acquired from recommendations across 41 publications (which is 15% of all recommended publications). This equates to an average of approximately 2.9 subscribers per recommendation. My other publication,

   helloIndia!
   has a higher ratio so this can vary based on factors outlined above.

   

Thanks to the 41 publications recommending

TechMadeSimple

2. As Substack doesn't currently provide sufficient data on followers, focus on calculating conversion ratios from free to paid subscriptions. This involves tracking how many subscribers are converting to paid subscribers, and to founding members.

3. Substack offers detailed recommendation stats, including subscribers acquired from different recommended publications. If you notice that some recommended publications haven't generated any free subscribers for you, consider removing them from your 'Recommend' list.

   

   This may suggest a lack of audience overlap or interest, and you can replace them with others more likely to drive engagement.

4. Keep experimenting and tweaking your approach. You can also add a publication manually. Simply navigate to 'Manage' under 'Recommendations' settings and choose ‘Add Publication.’

Caveats

There's always the risk of unknowingly recommending someone whose writing you vehemently disagree with. This is unavoidable upfront unless you invest the time to vet every single publication in the list shown by the Substack algorithm. A better approach is to watch for red flags after the recommendation to ensure any issues are addressed promptly.

I haven't encountered this as a significant issue because I use interests to control the feeds, unfollow, or block users to remove their publications from showing up in my recommendations.

Conclusion

Awareness is the initial step in any marketing strategy to pave the way for monetization. I may share a simple marketing strategy cheat sheet in a future post (let me know if this interests you).

Bottom line? Recommendations serve as the awareness engine for Substack, reducing your overheads. Use it effectively.

---

Was this useful? Do you have other tips you’d like to share?

Please consider pledging your support for this publication as it helps me continue providing valuable content and insights. Thank you!