You touched on a major issue that I feel Substack has not addressed: when the business model is based on paying to read long-form publications, there is an inherent limitation to the number of subscriptions one can afford or simply have time to keep up with. It’s not like social media with the indefinite free scroll of content, but they are kind of assuming that it is going to work that way. I have already gone beyond my limit and can’t afford to subscribe to any more writers, and don’t have the time to keep up with the ones I do subscribe to. We definitely need to be focused more on bringing new readers to the platform, versus expecting writers to subscribe to dozens of other writers, which is what Notes promotes.
Yes. I have an unfair advantage because I have a lot of content and had 25,000 subscribers, so I have been able to put 15+ books behind a paywall and serialize one of my books weekly. I'm hoping that in the arms race of finding paid subscribers, offering more bonus content means something, but I'm not sure that it does. I did it for my email subscribers mostly.
Unfortunately, the winners will win more in the future as the network effects mean more and more, and the people struggling to gain purchase will struggle more because there is not enough money to go around.
If you are interested in this topic, my latest article that launched yesterday talks about personalized marketing, and presents a new archetype system for author marketing my business partner and I think will help people do this work for longer with more joy, and I have another post coming in the next couple of days about how to use sections to give more back catalog work to paid subscribers.
It's very hard out there, and while free growth will likely continue to explode for people (which is great), paid growth will likely stagnate or go backward.
“At the end of the day, being a writer is work, and work often sucks” made me laugh. So many truth bombs in this one. Brilliant reply from Giannis Antetokounmpo.
I am a writer AND happy, but that’s genetics and environment and luck making me happy, not necessarily writing, although writing mostly makes me happy too.
Chase the joy. Optimize the happiness.
I’ve also tried to explain the Zen parable of a finger pointing at the moon. The point is not to look at the finger. The point is to look at the moon. Much as writing is my calling, the point is not beautiful writing. The point is a beautiful life, or as beautiful as possible while acknowledging the wretchedness and hard work.
Against my better judgment, I couldn't help myself but address what I view as a misunderstanding of normal distributions. This specific misunderstanding is a pet peeve of mine. For context, I have worked as a data engineer in sales reporting and statistical modeling for a decade at big tech companies.
Let me start by saying I agree with your actual point. Collecting more data allows you to understand the distribution, and small sample sizes make it difficult to estimate the true distribution. Good stuff.
I must address that your distribution of expected sales for a population will not be normally distributed. Normal distributions are found in nature when inertia prevents outliers. A common example is height. You cannot be 3 standard deviations taller than the average person. Human biology does not work that way.
When you are talking about sales, the distribution will never be normal. Let's take subscribers to this publication as an example. They may pay $0, $20 (on-sale sub), $50 (regular price sub), $300 (dragon sub). According to this article, 90% of people pay $0. Already it should be obvious the distribution is not normal (the statistical term is zero-inflated).
Even if we eliminate the free subs then the average expected outcome (or average lifetime value (LTV)) would be somewhere closer to $50 than $300. Now in theory, someone could gift an unlimited amount of money. Unlike nature, there is nothing preventing statistical outliers. Someone could spend $0 or $1,000 (infinitely more than 0, 50x more than $20). A distribution of sales follows an exponential curve, it is right skewed (AKA positive skew) and zero-inflated.
OK, done. Sorry, idk why I did this. I am currently binging content from this publication and enjoying it. Thanks!
Okay. I have been doing this work for 15 years and this model has always worked for me, so I will keep using it both in my own forecasting and in how I teach it. I am sorry if that doesn’t work for you. It absolutely does work for me and to teach this concept.
I have a degree in demographic sociology and I am pretty sure that is not true. A bell curve standard distribution comes when you approximate the population through having enough randomly distributed data on a sample of the population to approximate it. Perhaps we are talking about different concepts across disciplines, but you couldn’t do polling if you couldn’t approximate a population with sample data.
Your misunderstanding lies in applying demographic sociology to the domain of modelling sales. They are not the same.
Like I was explaining, for data to be normally distributed it must follow natural laws. The reason we use normal distributions in the sciences is because natural phenomenon have friction that prevent statistical outliers. The common example is height. You physically cannot have a human that is 1 inch tall or a human that is 20 feet tall. The friction preventing outliers is what causes the normal distribution to occur. The same applies to demographic sociology, there are physical limitations that allow normal distributions to occur.
When it comes to modeling the distribution of sales, or anything in finance, there is no headwind preventing very large and very small numbers i.e. outliers. Most people will pay $0 and some people will pay $300. The people who pay $300 are paying exponentially more than the people who pay $10. There is no friction preventing one person from paying exponentially more than someone else.
All data will have some underlying distribution, but in the case where there is no inertia preventing extreme outliers, the distribution will instead follow a power law.
I would recommend you check out the Incerto Series by Nassim Taleb if you are interested in learning more about statistical distributions as they exist outside of the natural sciences i.e. in the real world of business.
Don't believe me? Try to make a distribution of the LTV for your own subscribers, with frequency on the y axis, and LTV on the x axis. Do you really think that distribution will be normally distributed? By your own admission 90% of your subscribers give you $0.
You touched on a major issue that I feel Substack has not addressed: when the business model is based on paying to read long-form publications, there is an inherent limitation to the number of subscriptions one can afford or simply have time to keep up with. It’s not like social media with the indefinite free scroll of content, but they are kind of assuming that it is going to work that way. I have already gone beyond my limit and can’t afford to subscribe to any more writers, and don’t have the time to keep up with the ones I do subscribe to. We definitely need to be focused more on bringing new readers to the platform, versus expecting writers to subscribe to dozens of other writers, which is what Notes promotes.
Yes. I have an unfair advantage because I have a lot of content and had 25,000 subscribers, so I have been able to put 15+ books behind a paywall and serialize one of my books weekly. I'm hoping that in the arms race of finding paid subscribers, offering more bonus content means something, but I'm not sure that it does. I did it for my email subscribers mostly.
Unfortunately, the winners will win more in the future as the network effects mean more and more, and the people struggling to gain purchase will struggle more because there is not enough money to go around.
If you are interested in this topic, my latest article that launched yesterday talks about personalized marketing, and presents a new archetype system for author marketing my business partner and I think will help people do this work for longer with more joy, and I have another post coming in the next couple of days about how to use sections to give more back catalog work to paid subscribers.
It's very hard out there, and while free growth will likely continue to explode for people (which is great), paid growth will likely stagnate or go backward.
I love your archetype framework! I think I am a Forest.
so excited to show you more :)
“At the end of the day, being a writer is work, and work often sucks” made me laugh. So many truth bombs in this one. Brilliant reply from Giannis Antetokounmpo.
I am a writer AND happy, but that’s genetics and environment and luck making me happy, not necessarily writing, although writing mostly makes me happy too.
Chase the joy. Optimize the happiness.
I’ve also tried to explain the Zen parable of a finger pointing at the moon. The point is not to look at the finger. The point is to look at the moon. Much as writing is my calling, the point is not beautiful writing. The point is a beautiful life, or as beautiful as possible while acknowledging the wretchedness and hard work.
I love it.
Thanks for this. Really helpful.
You’re welcome! If you liked this one then I recommend this one. https://authorstack.substack.com/p/you-dont-have-to-like-me-as-long
I don’t think it’s behind the paywall yet. Also this weeks is about personalized marketing which seems to have been very helpful to people :)
Good advice!
Thank you!
Keep chasing joy! 😉
trying :) Thanks!
Great article, thanks for writing it.
Against my better judgment, I couldn't help myself but address what I view as a misunderstanding of normal distributions. This specific misunderstanding is a pet peeve of mine. For context, I have worked as a data engineer in sales reporting and statistical modeling for a decade at big tech companies.
Let me start by saying I agree with your actual point. Collecting more data allows you to understand the distribution, and small sample sizes make it difficult to estimate the true distribution. Good stuff.
I must address that your distribution of expected sales for a population will not be normally distributed. Normal distributions are found in nature when inertia prevents outliers. A common example is height. You cannot be 3 standard deviations taller than the average person. Human biology does not work that way.
When you are talking about sales, the distribution will never be normal. Let's take subscribers to this publication as an example. They may pay $0, $20 (on-sale sub), $50 (regular price sub), $300 (dragon sub). According to this article, 90% of people pay $0. Already it should be obvious the distribution is not normal (the statistical term is zero-inflated).
Even if we eliminate the free subs then the average expected outcome (or average lifetime value (LTV)) would be somewhere closer to $50 than $300. Now in theory, someone could gift an unlimited amount of money. Unlike nature, there is nothing preventing statistical outliers. Someone could spend $0 or $1,000 (infinitely more than 0, 50x more than $20). A distribution of sales follows an exponential curve, it is right skewed (AKA positive skew) and zero-inflated.
OK, done. Sorry, idk why I did this. I am currently binging content from this publication and enjoying it. Thanks!
Okay. I have been doing this work for 15 years and this model has always worked for me, so I will keep using it both in my own forecasting and in how I teach it. I am sorry if that doesn’t work for you. It absolutely does work for me and to teach this concept.
I have a degree in demographic sociology and I am pretty sure that is not true. A bell curve standard distribution comes when you approximate the population through having enough randomly distributed data on a sample of the population to approximate it. Perhaps we are talking about different concepts across disciplines, but you couldn’t do polling if you couldn’t approximate a population with sample data.
Your misunderstanding lies in applying demographic sociology to the domain of modelling sales. They are not the same.
Like I was explaining, for data to be normally distributed it must follow natural laws. The reason we use normal distributions in the sciences is because natural phenomenon have friction that prevent statistical outliers. The common example is height. You physically cannot have a human that is 1 inch tall or a human that is 20 feet tall. The friction preventing outliers is what causes the normal distribution to occur. The same applies to demographic sociology, there are physical limitations that allow normal distributions to occur.
When it comes to modeling the distribution of sales, or anything in finance, there is no headwind preventing very large and very small numbers i.e. outliers. Most people will pay $0 and some people will pay $300. The people who pay $300 are paying exponentially more than the people who pay $10. There is no friction preventing one person from paying exponentially more than someone else.
All data will have some underlying distribution, but in the case where there is no inertia preventing extreme outliers, the distribution will instead follow a power law.
I would recommend you check out the Incerto Series by Nassim Taleb if you are interested in learning more about statistical distributions as they exist outside of the natural sciences i.e. in the real world of business.
Don't believe me? Try to make a distribution of the LTV for your own subscribers, with frequency on the y axis, and LTV on the x axis. Do you really think that distribution will be normally distributed? By your own admission 90% of your subscribers give you $0.