I have a problem with one part of your article, here:
"Interestingly, this dynamic is exactly what US law wants: a hypercompetitive market that is amazing for consumers. That’s why the US government has had such a hard time regulating aggregators. They realize something is askew, but the end result is great for consumers, and in the US the definition of monopolies depends on whether consumers are better off. "
This is true, and indeed the law as currently formulated focuses only on monopolies that are demonstrably bad for consumers. The issue is that the market power that makes Facebook and Google great investments is also the market power that they 1) wield against competition on their platforms and 2) abuse in ways that damage society. There are stronger protections against both in Europe, which is why Facebook is in such trouble there. But it is evident here in the damage Facebook does to society in pursuing the profit motive exclusively and amorally.
I went to Stanford and many of my closest friends there are in tech or VCs. I did work in the field myself, but I am a management consultant now, and some of my projects have been with companies at the very frontier of cyber. These issues are blind spots for my friends, and there's definitely a solipsistic and libertarian strain in Silicon Valley. But these are issues that society and the law will have to wrestle with, because it's crushing the middle class and destabilizing democracy. Some is starting to happen with things like the growing traction of trends like ESG and stakeholder capitalism, but society moves much slower than tech, and I fear we will not put a cap on it before we lose the formula for what is important - our people, rather than solely our profits.
By the way, I am very much a capitalist. But I have seen too much having lived in the tech world for decades of the unbridled, but amoral enthusiasm for the "next great business model" without a pause to consider the damage that's being done. I am against a heavy hand from the government, but "power corrupts" has never seemed more true than today with the rise of global, unaccountable elites. Capitalism is absolutely the best system at it's core, but there are absolutely market failures, and they have never been more consequential than today.
The exceptions appear when aggregators blatantly exploit their power in a way that makes it worse for consumers, such as the App Store forcing payments through Apple and their 30% cut (which reduces alternative payment systems and takes such a heavy tax that many businesses are simply not viable because of it), or when Google sends searchers to worse results pages because it gives an advantage to its own results, like what happened with Google Shopping.
And hence why I want to cover in a future how to break the power accrued through network effects in a few companies.
I do think the negative impact of these networks is not as much their network effects (all social networks converge in the same direction) as it is the fact that humans haven't worked out the right rules of engagement with them.
Yup - and thinking about my reaction, it was more emotional, as that discussion is out of place in an article on network effects generally, and more than worthy of a full article of its own.
To wit, I was thinking the same as your second paragraph. What would be really ground breaking forward thinking would be to work through some of the solutions to the market failures inherent in network effects that are more pervasive, rapidly evolving and fast-moving than ever before. YET - to preserve the great value in economic efficiency that they drive. That's truly a hard nut to crack, since the driver of both is the same at its core.
I think open platforms and other technical elements can be part of the solution (difficult to implement themselves), but part may require policy and social work that goes far beyond what we've been used to. But I tend to think democracy won't survive very long under our current societal architecture - lies travel way faster than truth, mainly because social networks have become extremely efficient at hacking human attention.
The problem is network effect businesses are only valuable at scale.
But they also are only dangerous at scale.
Thus, the natural old-school laws of competitions don't strictly apply.
I do think this point is overemphasized as platforms like to plead the I'm special card.
But it contains a kernel of truth.
I think eventually they will be regulated the way natural monopolies currently are.
I also think a blockchain based network that allowed easy and effective verification would be vital: right now, that's very unlikely to happen because the capability just isn't there.
With regard to hacking attention, it's a business model problem.
It's not just about switching to paid subscription and all the problems will go away as many naively claim.
Spotify premium is adfree and it's just as guilty of misinformation.
It seems to me that the real problem is recommendation; if a business model is based on algorithmic recommendation, then it must hack attention: there's no other way to succeed.
That won't go away.
It may, however, be mitigated as people learn to deal with it: these businesses are still very young and we as societies and communities have not fully evolved with them as of yet.
The first thing is some points about network effects are inaccurate.
Metcalfe's law is kind of true mathematically but is untrue realistically.
That's because every new member is not valuable to the entire network: no one's social circle has that sort of range.
So, even if I could talk with 4 million people on the telephone, I certainly don't want to talk with even a fraction of them: possibility is not the same as demand.
Even mathematically, the law is suspect. The growth in value should really be N² - N rather than N² since each node can't communicate with itself.
I'm surprised as to how very few people point either of these out.
Two, network effects are a very successful business model whose time, I suspect, will soon begin to set under the sun.
The reason is because platforms and to a lesser extent, aggregators, give up vertical control for the benefits of controlling supply and demand in one place.
The problems are twofold: one, it only works if there's a massive pool of supply, like with information (Google) or social relationships(Meta) or applications( Apple app stores). Uber and Airbnb, companies already dealing in huge markets, owe a considerable part of their success to creating new sources of supply: many people who had not drove a cab or put their apartment up for rent in their lives began to do so.
This is why it's difficult to have a platform business with carwashing or massages or something similar: both the demand and supply aren't high enough to make the marketplace truly liquid.
The second problem is the issue of vertical control. If you give up vertical control, you allow some bad eggs onto the platform. There's no other option.
That's fine when the bad egg is just a troll or an ideologue.
With other industries, it's different.
Industries like healthcare cannot be truly network-scalable because of this problem. And also because these are not one to many networks either, e.g a doctor can't treat thousands at once.
Network effects are great at transactions. They are not great at relationships.
So relationship businesses like law and accounting have remained fairly impervious to software's world-sized appetite.
Three, aggregators are kind of a patch. Aggregators bring the relationships back. That's what they do: bringing back the relationship with the writer ( Substack), the relationship with the retailer( Shopify), etc.
And so, they represent a sort of evolution.
I would add Stripe to this list too.
The companies of this decade and the next are going to be the ones who provide empowerment: those who own the infrastructure of the new relationships of the digital economy.
And on this, Tomas, as usual, is spot-on.
I look forward to seeing the related posts on these matters.
Yes, valid point, but check out Reed's Law, which is a counterfactor that increases the value of network effects. The overarching point is that this is a nonlinear effect.
2. Agreed that there's not billions of network effect businesses. But there are still many to be created!
3. I'm confused at the healthcare example. You mention the importance of not getting it wrong, and then the capacity of handling so much demand. Those are two very different things.
On the bad eggs, there's plenty of bad doctors. Reviews and other types of feedback on them can give signals to the market on who is good and who isn't.
The capacity of tackling demand is no different than most other physical businesses...?
4. Network effects can be great at relationships. That's what Windows did with software, as a simple example. It's what Shopify is doing. I think you mean aggregators are bad at creating relationships, which I agree and mention in the article. That's a choice though.
The problem you mention about "relationship businesses" is that of disintermediation. When you find a good lawyer, you want to keep her, and bypass the platform. Note that this is not the case for notaries, for example, since the need for them is much more sporadic and their services less differentiated.
I think you're mixing platforms with aggregators. Platforms are the ones that facilitate relationships—as defined by Ben Thompson.
I don't think Substack can be considered as either today (I consider them a SaaS), but they definitely want to go the platform route, you're right. I have an article coming on that. Shopify though is truly a platform.
I can't judge Stripe. You might be right. I am not acquainted enough with the details of their business model. Why do you propose them as one?
Very interesting conversation! Thanks for sharing your thoughts.
1.) First off, I deeply appreciate your meticulousness.
2.) Reed's law is certainly more precise than Metcalfe's law and the real point is about nonlinearities as you've stated anyway.
3.) We are in full agreement that there are many platform businesses yet to be created. They just won't be as profitable as the first set of platforms.
4.) Please allow me to clarify your third point: for me, the two problems are interwoven.
There are more risks to having bad eggs on an healthcare platform so there will be more vertical control: thorough certification checks, thorough background checks, and certainly more regulatory oversight.
Vertical control is expensive and will eat into margins.
5.) To the point of user ratings, user reviews and ratings are seriously broken.
They are the proverbial hammer of online businesses. They represented a fairly effective way to solve the trust problem in the beginning when Ebay pioneered them but now, it's not the case.
Apart from the fact that many are fake, and we can't tell how many because we only know what we know, there are two issues:
a.) Some services are more mutable than others.
Driving you from Oakland to San Francisco is not really a different task from driving me from Oakland to San Francisco.
So if you rate the driver well or poorly, it is very useful information for me, as long as you are being authentic.
With certain domains like healthcare, it's very different.
I have a different body, physiology, biochemistry, and immune system to yours.
So even if you rate the doctor well, it's not very valuable information.
Even if it's the same doctor and the same medical condition.
b.) User reviews work when the consumer can judge the service properly.
For instance, you don't have to be a good driver to know if someone is driving recklessly.
And you don't have to be a furniture expert to know if a chair is too fragile to sit on.
These domains have low opacity of information.
With fields like law, accounting, and medicine, it's different.
I'm not qualified to judge all the actions of my doctor and I'm certainly less qualified to know whether his success was due to him, luck, or something else.
So my review again has little value.
Many people are in the same boat and it's the dreaded agency problem: unless the agent is truly honest, and thankfully, many are, the principal has to understand the actions of the agent to an extent, so as to avoid being cheated.
So ironically, the principal-agent relationship must be somewhat redundant and unnecessary to be successful.
6.) It's a wise and lovely point as usual about disintermediation.
I will certainly be adding it to my toolbox of insights.
Perhaps, we may say that when the relationship is of very high value, both participants have an incentive to export it out of the platform and capture all of the value themselves.
Also, with law, things are different.
We can all share the same great driver but we can't all share the same great lawyer: there may be conflicts of interest.
And when there may be conflicts of interest, there will be conflicts of interest.
7.) It's also a very lovely insight about Substack as a SaaS business.
Now that I think of it that way, it really adds up: it definitely explains why they release features the way Taylor Swift releases albums.
I made the mistake of judging a business by how it makes its money( here, Substack is a platform) rather than how it actually operates( here, as you've wisely written, it is a SaaS.)
They sometimes don't overlap.
8.) Finally, I include Stripe cos in many cases, the consumers don't even know or have to know they are using Stripe for payments.
In a world where we will paying for everything digitally, Stripe would be an incredibly valuable business.
In that sense, it is empowering businesses by taking care of things like fraud prevention that these businesses can't do very well on their own.
The only big caveat is if crypto becomes the currency of the future or if everyone has a wallet with the central bank.
For reasons of my own, I don't see any happening any time soon.
Do pardon the length.
And thank you for the insights and the pleasure of the conversation.
Very interesting, thank you. I especially like your section around reviews and how to manage trust in industries where the product is harder to qualify (which is why regulations exist in these industries in the first place).
Re Substack it’s a SaaS also in payments. Subs don’t pay Substack, which pays us. Subs pay writers, and the writers give a rev share to Substack.
Re Stripe I understand they’re valuable. I still don’t understand their network effects :/
The terminology around platforms business models & strategy is not consistent. Most definitions of platform would look at "aggregators" as simply one type of platform. Ben Thompson is a smart guy, but his "aggregation theory" construct is a unique vocabulary and is not used by almost any other analysts.
Nice article - I think the one thing missing, is the lesson from the early internet days. Prior to the 90's, there were platforms - AOL, Compuserve , Prodigy, GEnie, Dialcom, GeoNet where you could send people email but only if both people had accounts with the same service.
In the '90s we built common protocols, and pretty soon anyone could email anyone else - the Network effect applied across everyone (much like interconnected phone networks) so overall it built the lock-in services.
That's still the case with email, and Web1.0 was designed on the same principles, but most of us have moved to messaging products that lock you in - we are back to a world where I need a Facebook account to send a message to X and a Telegram account for Y but X cant message Y, and I can't do a three way conversation.
The question is for how long will that persist, before we get so fed up with having a dozen messaging apps, that we demand the ability to cross-message. The unfortunate thing is that unlike email, we aren't the ones paying for the service (that's the advertisers), so there is less cost to the service if I take my business elsewhere, so its harder this time to see a change coming.
This is the crux of the problem: interconnectivity eliminates the capture of network effects at the cost of fixing protocols in time, and hence slowing down all innovation in that network. So new networks appear to replace them, which are not interconnected. It’s the balance between speed and concentration on one side, or slowness and decentralization on the other.
I get the impression from the end of your article that platforms are gaining in popularity vs. aggregators. What I would be interested to know is:
- Is that actually the case, or is it just my impression? I know lots of aggregators e.g. Amazon have done well in the pandemic, but then again so has Shopify.
- Do you think there are growing business reasons to pursue more of a platform model vs. an aggregator model? I know that big aggregators e.g. Google, Facebook and Amazon are all getting more regulatory attention and bad press than they did previously, and Milennials and Gen Z are more likely to want to work and invest in businesses that they are aligned with philosophically...are these risks enough to overcome the siren song of the stronger network effects that you get from being an aggregator? I understand that this could vary greatly depending on the industry, but would love to hear your thoughts.
I'm coming back to this article and would like your opinion Tomas. It could as well be another article :)
How do you see the platform landscape and model changing when AI comes into play? We can see community-based platforms like Brainly, Stack Overflow, DeviantArt etc. being pressured by LLMs taking over their market space. Some fight it, some join it, some choose shady path to maximise short-term profit. Regardless, I believe AI will impact this whole business model.
What are your thoughs?
Also, a sidenote - it's damn hard to find a specific article from you (or another author that I follow) on substack. I had to use Google to find this one, since I didn't find an easy way to search just your articles on this platform.
Great article. Few questions regarding it. I've seen companies that still succeed even without network effects for example: all utility category like Avira or Norton - those companies do not have strong network effects, what they do have is kinda strong tech and good historical organic traffic. Same goes for example Kape Tech - they have bunch VPN's and they don't have strong network effects either.
My point is not that the only successful companies have network effects. My point is that they have an outsized growth potential compared to other types of businesses.
I don't know the businesses you mention, but from a 10s googling session, are they all SaaS? If so, stand by for the next article.
I have a problem with one part of your article, here:
"Interestingly, this dynamic is exactly what US law wants: a hypercompetitive market that is amazing for consumers. That’s why the US government has had such a hard time regulating aggregators. They realize something is askew, but the end result is great for consumers, and in the US the definition of monopolies depends on whether consumers are better off. "
This is true, and indeed the law as currently formulated focuses only on monopolies that are demonstrably bad for consumers. The issue is that the market power that makes Facebook and Google great investments is also the market power that they 1) wield against competition on their platforms and 2) abuse in ways that damage society. There are stronger protections against both in Europe, which is why Facebook is in such trouble there. But it is evident here in the damage Facebook does to society in pursuing the profit motive exclusively and amorally.
I went to Stanford and many of my closest friends there are in tech or VCs. I did work in the field myself, but I am a management consultant now, and some of my projects have been with companies at the very frontier of cyber. These issues are blind spots for my friends, and there's definitely a solipsistic and libertarian strain in Silicon Valley. But these are issues that society and the law will have to wrestle with, because it's crushing the middle class and destabilizing democracy. Some is starting to happen with things like the growing traction of trends like ESG and stakeholder capitalism, but society moves much slower than tech, and I fear we will not put a cap on it before we lose the formula for what is important - our people, rather than solely our profits.
By the way, I am very much a capitalist. But I have seen too much having lived in the tech world for decades of the unbridled, but amoral enthusiasm for the "next great business model" without a pause to consider the damage that's being done. I am against a heavy hand from the government, but "power corrupts" has never seemed more true than today with the rise of global, unaccountable elites. Capitalism is absolutely the best system at it's core, but there are absolutely market failures, and they have never been more consequential than today.
Agreed. From the article:
The exceptions appear when aggregators blatantly exploit their power in a way that makes it worse for consumers, such as the App Store forcing payments through Apple and their 30% cut (which reduces alternative payment systems and takes such a heavy tax that many businesses are simply not viable because of it), or when Google sends searchers to worse results pages because it gives an advantage to its own results, like what happened with Google Shopping.
And hence why I want to cover in a future how to break the power accrued through network effects in a few companies.
I do think the negative impact of these networks is not as much their network effects (all social networks converge in the same direction) as it is the fact that humans haven't worked out the right rules of engagement with them.
Yup - and thinking about my reaction, it was more emotional, as that discussion is out of place in an article on network effects generally, and more than worthy of a full article of its own.
To wit, I was thinking the same as your second paragraph. What would be really ground breaking forward thinking would be to work through some of the solutions to the market failures inherent in network effects that are more pervasive, rapidly evolving and fast-moving than ever before. YET - to preserve the great value in economic efficiency that they drive. That's truly a hard nut to crack, since the driver of both is the same at its core.
I think open platforms and other technical elements can be part of the solution (difficult to implement themselves), but part may require policy and social work that goes far beyond what we've been used to. But I tend to think democracy won't survive very long under our current societal architecture - lies travel way faster than truth, mainly because social networks have become extremely efficient at hacking human attention.
The problem is network effect businesses are only valuable at scale.
But they also are only dangerous at scale.
Thus, the natural old-school laws of competitions don't strictly apply.
I do think this point is overemphasized as platforms like to plead the I'm special card.
But it contains a kernel of truth.
I think eventually they will be regulated the way natural monopolies currently are.
I also think a blockchain based network that allowed easy and effective verification would be vital: right now, that's very unlikely to happen because the capability just isn't there.
With regard to hacking attention, it's a business model problem.
It's not just about switching to paid subscription and all the problems will go away as many naively claim.
Spotify premium is adfree and it's just as guilty of misinformation.
It seems to me that the real problem is recommendation; if a business model is based on algorithmic recommendation, then it must hack attention: there's no other way to succeed.
That won't go away.
It may, however, be mitigated as people learn to deal with it: these businesses are still very young and we as societies and communities have not fully evolved with them as of yet.
This is a lovely article.
It demands a measured and considered response.
I will try to do that here.
The first thing is some points about network effects are inaccurate.
Metcalfe's law is kind of true mathematically but is untrue realistically.
That's because every new member is not valuable to the entire network: no one's social circle has that sort of range.
So, even if I could talk with 4 million people on the telephone, I certainly don't want to talk with even a fraction of them: possibility is not the same as demand.
Even mathematically, the law is suspect. The growth in value should really be N² - N rather than N² since each node can't communicate with itself.
I'm surprised as to how very few people point either of these out.
Two, network effects are a very successful business model whose time, I suspect, will soon begin to set under the sun.
The reason is because platforms and to a lesser extent, aggregators, give up vertical control for the benefits of controlling supply and demand in one place.
The problems are twofold: one, it only works if there's a massive pool of supply, like with information (Google) or social relationships(Meta) or applications( Apple app stores). Uber and Airbnb, companies already dealing in huge markets, owe a considerable part of their success to creating new sources of supply: many people who had not drove a cab or put their apartment up for rent in their lives began to do so.
This is why it's difficult to have a platform business with carwashing or massages or something similar: both the demand and supply aren't high enough to make the marketplace truly liquid.
The second problem is the issue of vertical control. If you give up vertical control, you allow some bad eggs onto the platform. There's no other option.
That's fine when the bad egg is just a troll or an ideologue.
With other industries, it's different.
Industries like healthcare cannot be truly network-scalable because of this problem. And also because these are not one to many networks either, e.g a doctor can't treat thousands at once.
Network effects are great at transactions. They are not great at relationships.
So relationship businesses like law and accounting have remained fairly impervious to software's world-sized appetite.
Three, aggregators are kind of a patch. Aggregators bring the relationships back. That's what they do: bringing back the relationship with the writer ( Substack), the relationship with the retailer( Shopify), etc.
And so, they represent a sort of evolution.
I would add Stripe to this list too.
The companies of this decade and the next are going to be the ones who provide empowerment: those who own the infrastructure of the new relationships of the digital economy.
And on this, Tomas, as usual, is spot-on.
I look forward to seeing the related posts on these matters.
Thx! A few reactions:
1. Metcalfe's Law
Yes, valid point, but check out Reed's Law, which is a counterfactor that increases the value of network effects. The overarching point is that this is a nonlinear effect.
2. Agreed that there's not billions of network effect businesses. But there are still many to be created!
3. I'm confused at the healthcare example. You mention the importance of not getting it wrong, and then the capacity of handling so much demand. Those are two very different things.
On the bad eggs, there's plenty of bad doctors. Reviews and other types of feedback on them can give signals to the market on who is good and who isn't.
The capacity of tackling demand is no different than most other physical businesses...?
4. Network effects can be great at relationships. That's what Windows did with software, as a simple example. It's what Shopify is doing. I think you mean aggregators are bad at creating relationships, which I agree and mention in the article. That's a choice though.
The problem you mention about "relationship businesses" is that of disintermediation. When you find a good lawyer, you want to keep her, and bypass the platform. Note that this is not the case for notaries, for example, since the need for them is much more sporadic and their services less differentiated.
I think you're mixing platforms with aggregators. Platforms are the ones that facilitate relationships—as defined by Ben Thompson.
I don't think Substack can be considered as either today (I consider them a SaaS), but they definitely want to go the platform route, you're right. I have an article coming on that. Shopify though is truly a platform.
I can't judge Stripe. You might be right. I am not acquainted enough with the details of their business model. Why do you propose them as one?
Very interesting conversation! Thanks for sharing your thoughts.
1.) First off, I deeply appreciate your meticulousness.
2.) Reed's law is certainly more precise than Metcalfe's law and the real point is about nonlinearities as you've stated anyway.
3.) We are in full agreement that there are many platform businesses yet to be created. They just won't be as profitable as the first set of platforms.
4.) Please allow me to clarify your third point: for me, the two problems are interwoven.
There are more risks to having bad eggs on an healthcare platform so there will be more vertical control: thorough certification checks, thorough background checks, and certainly more regulatory oversight.
Vertical control is expensive and will eat into margins.
5.) To the point of user ratings, user reviews and ratings are seriously broken.
They are the proverbial hammer of online businesses. They represented a fairly effective way to solve the trust problem in the beginning when Ebay pioneered them but now, it's not the case.
Apart from the fact that many are fake, and we can't tell how many because we only know what we know, there are two issues:
a.) Some services are more mutable than others.
Driving you from Oakland to San Francisco is not really a different task from driving me from Oakland to San Francisco.
So if you rate the driver well or poorly, it is very useful information for me, as long as you are being authentic.
With certain domains like healthcare, it's very different.
I have a different body, physiology, biochemistry, and immune system to yours.
So even if you rate the doctor well, it's not very valuable information.
Even if it's the same doctor and the same medical condition.
b.) User reviews work when the consumer can judge the service properly.
For instance, you don't have to be a good driver to know if someone is driving recklessly.
And you don't have to be a furniture expert to know if a chair is too fragile to sit on.
These domains have low opacity of information.
With fields like law, accounting, and medicine, it's different.
I'm not qualified to judge all the actions of my doctor and I'm certainly less qualified to know whether his success was due to him, luck, or something else.
So my review again has little value.
Many people are in the same boat and it's the dreaded agency problem: unless the agent is truly honest, and thankfully, many are, the principal has to understand the actions of the agent to an extent, so as to avoid being cheated.
So ironically, the principal-agent relationship must be somewhat redundant and unnecessary to be successful.
6.) It's a wise and lovely point as usual about disintermediation.
I will certainly be adding it to my toolbox of insights.
Perhaps, we may say that when the relationship is of very high value, both participants have an incentive to export it out of the platform and capture all of the value themselves.
Also, with law, things are different.
We can all share the same great driver but we can't all share the same great lawyer: there may be conflicts of interest.
And when there may be conflicts of interest, there will be conflicts of interest.
7.) It's also a very lovely insight about Substack as a SaaS business.
Now that I think of it that way, it really adds up: it definitely explains why they release features the way Taylor Swift releases albums.
I made the mistake of judging a business by how it makes its money( here, Substack is a platform) rather than how it actually operates( here, as you've wisely written, it is a SaaS.)
They sometimes don't overlap.
8.) Finally, I include Stripe cos in many cases, the consumers don't even know or have to know they are using Stripe for payments.
In a world where we will paying for everything digitally, Stripe would be an incredibly valuable business.
In that sense, it is empowering businesses by taking care of things like fraud prevention that these businesses can't do very well on their own.
The only big caveat is if crypto becomes the currency of the future or if everyone has a wallet with the central bank.
For reasons of my own, I don't see any happening any time soon.
Do pardon the length.
And thank you for the insights and the pleasure of the conversation.
Very interesting, thank you. I especially like your section around reviews and how to manage trust in industries where the product is harder to qualify (which is why regulations exist in these industries in the first place).
Re Substack it’s a SaaS also in payments. Subs don’t pay Substack, which pays us. Subs pay writers, and the writers give a rev share to Substack.
Re Stripe I understand they’re valuable. I still don’t understand their network effects :/
Re Substack, I deeply agree
Re Stripe, if Stripe becomes the dominant infrastructure for digital payments, then there would be no need to use other platforms.
Everyone I want to transact with would be using Stripe already anyway; other platforms will be redundant.
The more merchants and creators it signs up, the more value for consumers who already use Stripe.
It's a kind of imperceptible one-sided network effect.
The terminology around platforms business models & strategy is not consistent. Most definitions of platform would look at "aggregators" as simply one type of platform. Ben Thompson is a smart guy, but his "aggregation theory" construct is a unique vocabulary and is not used by almost any other analysts.
I hear you, but it's the most complete language I have seen, so I'm using it. If a better taxonomy appears, I might adopt it.
Nice article - I think the one thing missing, is the lesson from the early internet days. Prior to the 90's, there were platforms - AOL, Compuserve , Prodigy, GEnie, Dialcom, GeoNet where you could send people email but only if both people had accounts with the same service.
In the '90s we built common protocols, and pretty soon anyone could email anyone else - the Network effect applied across everyone (much like interconnected phone networks) so overall it built the lock-in services.
That's still the case with email, and Web1.0 was designed on the same principles, but most of us have moved to messaging products that lock you in - we are back to a world where I need a Facebook account to send a message to X and a Telegram account for Y but X cant message Y, and I can't do a three way conversation.
The question is for how long will that persist, before we get so fed up with having a dozen messaging apps, that we demand the ability to cross-message. The unfortunate thing is that unlike email, we aren't the ones paying for the service (that's the advertisers), so there is less cost to the service if I take my business elsewhere, so its harder this time to see a change coming.
This is the crux of the problem: interconnectivity eliminates the capture of network effects at the cost of fixing protocols in time, and hence slowing down all innovation in that network. So new networks appear to replace them, which are not interconnected. It’s the balance between speed and concentration on one side, or slowness and decentralization on the other.
Great article.
I get the impression from the end of your article that platforms are gaining in popularity vs. aggregators. What I would be interested to know is:
- Is that actually the case, or is it just my impression? I know lots of aggregators e.g. Amazon have done well in the pandemic, but then again so has Shopify.
- Do you think there are growing business reasons to pursue more of a platform model vs. an aggregator model? I know that big aggregators e.g. Google, Facebook and Amazon are all getting more regulatory attention and bad press than they did previously, and Milennials and Gen Z are more likely to want to work and invest in businesses that they are aligned with philosophically...are these risks enough to overcome the siren song of the stronger network effects that you get from being an aggregator? I understand that this could vary greatly depending on the industry, but would love to hear your thoughts.
This is the narrative I have in my brain:
Whenever there are network effects, the first wave of companies always monopolize those network effects.
Network effects are native to the Internet.
Therefore, massive companies owning these network effects were meant to appear on the Internet.
However, the cycle of innovation is also much faster over the Internet. New competitors can easily appear. Some will be successful.
This is especially true if these competitors are not as greedy, as in the case for Shopify or Ankorstore (vs. Amazon), or Substack (vs. Medium).
I really enjoyed this article.
Wonderful as usual, Tomas. Thanks.
I'm coming back to this article and would like your opinion Tomas. It could as well be another article :)
How do you see the platform landscape and model changing when AI comes into play? We can see community-based platforms like Brainly, Stack Overflow, DeviantArt etc. being pressured by LLMs taking over their market space. Some fight it, some join it, some choose shady path to maximise short-term profit. Regardless, I believe AI will impact this whole business model.
What are your thoughs?
Also, a sidenote - it's damn hard to find a specific article from you (or another author that I follow) on substack. I had to use Google to find this one, since I didn't find an easy way to search just your articles on this platform.
Great article. Few questions regarding it. I've seen companies that still succeed even without network effects for example: all utility category like Avira or Norton - those companies do not have strong network effects, what they do have is kinda strong tech and good historical organic traffic. Same goes for example Kape Tech - they have bunch VPN's and they don't have strong network effects either.
My point is not that the only successful companies have network effects. My point is that they have an outsized growth potential compared to other types of businesses.
I don't know the businesses you mention, but from a 10s googling session, are they all SaaS? If so, stand by for the next article.