Excellent article. The time for thin marketplaces is over. Go deep and make your experience AI native while you still can. Assume LLMs will become formidable at much of what you do so you aren't surprised later if they do get there.
I can give some context about the travel industry. I previously worked for Booking.com. I see history repeating itself. Google has been trying to become the one place for all your travel needs since ~2015, but it couldn't go beyond the travel search experience, which is another ads (CPC) product.
Same for Amazon. It also tried to get into the travel business multiple times. Definitely both of these companies don't lack financial resources or technical expertise.
My diagnosis:
1. They both couldn't navigate the operational complexity of the post-purchase experience (24/7 everywhere on earth), and immediate need to react at scale during disasters.
2. Travel is a separate industry that has its own regulations that an aggregator needs to comply with (A single transaction has regulations for travelers, and regulations for hotels/homes)
3. Travel is a big cost transaction. Consumers spend significant time on researching how to get the best value for their money. FOMO is influencing their process. The big travel players are in continues process of inventing ways to source
a. Exclusive data about hotels/homes that de-risk the transaction (data that hotels themselves might not have)
b. Source exclusive discounts and bundle them in creative ways to get the best ROI
c. Optimize how to present the data and pricing to each customer
LLM based search (My money is on Google AI, not ChatGPT) will have a leg up on what they know about the customer, but will face the same challenges that faced Google and Amazon before. ROI wise, it will be way more profitable for LLM players to adopt an Ads (CPC) model, and let the travel OTAs deal with the messy travel business.
The question is how to do ads without hurting customer trust. This is Google's bread and butter, and customers are used to seeing Ads on Google. ChatGPT is new to this game and customers are not used to seeing ads on ChatGPT.
That's why I believe the near future of travel is not going to change much. Same players at each step of the funnel but some new tech.
Thank you for the context! The operational complexity and regulatory hurdle is very real. If anything, this slows down the timeline people are projecting for how quickly LLMs are going to be able to move into new industries.
Tend to agree. The complexity of the travel industry. Absolute pricing, and Absolute availability are hard to grasp for the rest of ecommerce world. You can't just ship it next week when it comes back in stock. I think the LLM search will end up doing like Google and becoming a ppc layer. They are generalists.
This framing really clicks, LLMs don’t just threaten marketplaces, they selectively squeeze where value is thinnest. The supply legibility + management lens explains why some platforms get hollowed out while others actually get stronger.
Curious how you see “trust, risk, and fulfillment guarantees” evolving as the real moat when search and transactions get abstracted away.
Generally marketplaces with a peer to peer dynamic have (1) harder to aggregate supply and (2) need to invest more heavily in trust and guarantees. This would push these models to being more defensible against LLMs.
Travel industry is more complex than it might appear from the outside.
Connecting hotel supply to LLMs such that they can book a stay is not straightforward. Hotels use channel managers to distribute supply real time to multiple channels (OTAs, own site etc.). This is essential to avoid overbooking which terrifies hotels and is a huge customer issue.
OTAs like Booking have spent over a decade building hundreds of integrations with these channel managers which is a heavily fragmented ecosystem. Outside the big chains, hotels are also a very fragmented supply base globally. (US is an outlier).
So while the top of the funnel is changing (Search ->LLM), it will take a lot of effort to make meaningful dents to the transaction layer. I don't see LLMs spending that kind of energy / money to replicate this, esp. if there are bigger opportunities (i.e. going after Ad dollars).
Nevertheless, OTAs can't sit still and need to give customers a strong reason to book directly. Investing in loyalty programs will be a good start.
Great article and agree with your rationale! Where would you say leading online classified companies such as Zillow, CarGurus or Rightmove rank? Many thanks
Interesting highlevel view. Actually Booking is the largest hotel marketplace. They're incidentally also a rentals marketplace. Booking lives on SEM. They absolutely rule that space. Tripadvisor isn't a marketplace. They do reviews mostly and sell clicks to drive traffic to hotel and OTA websites. Getting inventory from hotels is painful and prices change every day or multiple times a day. LLMs are going to struggle with that. Booking and Expedia might become the plumbing for LLMs. But I doubt the LLMs will want to have hundreds of integrations to hotels for inventory and rates. FYI I just published the top 50 "marketplaces" in the hotel space: https://martinsoler.substack.com/i/184632493/great-distribution-starts-with-finding-the-right-guests
Yes Booking is the largest hotel marketplace but their homes business is large and growing to the point it is rivaling Airbnb in size
I would call TripAdvisor a Lead Gen marketplace where they facilitate search only and not transactions (far left of the framework)
What the LLMs are doing with their commerce protocols (eg Google Universal Commerce Protocol) is meant to solve the integration problem you are referring to. Agree it’s a tricky one but I think it’s more solvable for hotels than it is for e.g. e-commerce
Nice reply!!! :-) Booking are competing with Airbnb on rentals. But the Airbnb market is quite small compared to hotels. The main business of Booking is hotels. Booking is larger than Airbnb in the rental space. Which is crazy.
I agree that between MCP and UCP (hopefully) will reduce the barrier to entry to get inventory and rates. But my theory is that most of the existing marketplaces will need to pivot to becoming the plumbing layer for the data to LLM search. It took Google more than a decade to get access to rates and availability of hotels, and still today, they have a thin and unstable API layer that fails. Booking, Expedia etc have actual inventory and rates. Expedia is already starting to pivot towards plumbing. They will resist it as much as they can. Look at Amazon blocking bots.
Excellent one. How would you rate marketplaces like olx / poshmark etc that is used goods market ? Not a very frequent purchase but AI can do (grading to negotiation) which traditional processes can’t. So an opportunity to multiply the TAM, grow CM …or is it just a marginal opportunity?
Excellent analysis. However I would personally add one other idea. First, I think its worth keeping in mind that the pureplay marketplace will likely always be able to correctly argue that their supply coverage is at least somewhat better than that of the LLM. That's because in practically all marketplace settings, getting the final chunk of supply is really challenging; and unlikely to be on the critical path for the LLM. And because we know for now LLMs aren't perfect either.
Whether that matters depends on two things. First as you point out, quite how fragmented the supply actually is. Clearly here hotels are less fragmented in the US than takeaway outlets for example. However, I think there's another variable; the actual importance of the transaction itself. The proxy for that would be ticket size. So its the degree of fragmentation (F) * ticket size.
I would argue that the high ticket size provides additional protection for the likes of Zillow, and arguably flight aggregation too (which is done by Amadeus, Sabre today and then resold by Expedia etc), whereas consumers are unlikely to be too bothered for small tickets (food delivery, ride-sharing). Put another way, the additional hassle factor of leaving ChatGPT - which knows more about me - may not be worth it for that last 1 or 2 burger joints in the SERP - but probably is if you are searching for a house or the last flight home for Christmas.
Excellent article. The time for thin marketplaces is over. Go deep and make your experience AI native while you still can. Assume LLMs will become formidable at much of what you do so you aren't surprised later if they do get there.
"thin marketplaces" is such a good way to put it
Been watching too many poker analysis videos.
Great 2*2s for framing the problem
Thanks for this… great insights!
Glad you enjoyed it!
Superb 👌 insights
Thank you!
Amazing article!!
thanks!
Great article!
I can give some context about the travel industry. I previously worked for Booking.com. I see history repeating itself. Google has been trying to become the one place for all your travel needs since ~2015, but it couldn't go beyond the travel search experience, which is another ads (CPC) product.
Same for Amazon. It also tried to get into the travel business multiple times. Definitely both of these companies don't lack financial resources or technical expertise.
My diagnosis:
1. They both couldn't navigate the operational complexity of the post-purchase experience (24/7 everywhere on earth), and immediate need to react at scale during disasters.
2. Travel is a separate industry that has its own regulations that an aggregator needs to comply with (A single transaction has regulations for travelers, and regulations for hotels/homes)
3. Travel is a big cost transaction. Consumers spend significant time on researching how to get the best value for their money. FOMO is influencing their process. The big travel players are in continues process of inventing ways to source
a. Exclusive data about hotels/homes that de-risk the transaction (data that hotels themselves might not have)
b. Source exclusive discounts and bundle them in creative ways to get the best ROI
c. Optimize how to present the data and pricing to each customer
LLM based search (My money is on Google AI, not ChatGPT) will have a leg up on what they know about the customer, but will face the same challenges that faced Google and Amazon before. ROI wise, it will be way more profitable for LLM players to adopt an Ads (CPC) model, and let the travel OTAs deal with the messy travel business.
The question is how to do ads without hurting customer trust. This is Google's bread and butter, and customers are used to seeing Ads on Google. ChatGPT is new to this game and customers are not used to seeing ads on ChatGPT.
That's why I believe the near future of travel is not going to change much. Same players at each step of the funnel but some new tech.
Thank you for the context! The operational complexity and regulatory hurdle is very real. If anything, this slows down the timeline people are projecting for how quickly LLMs are going to be able to move into new industries.
Tend to agree. The complexity of the travel industry. Absolute pricing, and Absolute availability are hard to grasp for the rest of ecommerce world. You can't just ship it next week when it comes back in stock. I think the LLM search will end up doing like Google and becoming a ppc layer. They are generalists.
This framing really clicks, LLMs don’t just threaten marketplaces, they selectively squeeze where value is thinnest. The supply legibility + management lens explains why some platforms get hollowed out while others actually get stronger.
Curious how you see “trust, risk, and fulfillment guarantees” evolving as the real moat when search and transactions get abstracted away.
What about real estate and automotive portals?
I don’t see a mention of the peer-to-peer aspect of some marketplaces, eg Airbnb. Would love to hear your take why it will matter vs not matter.
Generally marketplaces with a peer to peer dynamic have (1) harder to aggregate supply and (2) need to invest more heavily in trust and guarantees. This would push these models to being more defensible against LLMs.
Thanks Dan 🙏🏼
Would you consider Zocdoc local services? Local, considered, and relatively low frequency. But some medical services are more heterogeneous.
Interesting article.
Travel industry is more complex than it might appear from the outside.
Connecting hotel supply to LLMs such that they can book a stay is not straightforward. Hotels use channel managers to distribute supply real time to multiple channels (OTAs, own site etc.). This is essential to avoid overbooking which terrifies hotels and is a huge customer issue.
OTAs like Booking have spent over a decade building hundreds of integrations with these channel managers which is a heavily fragmented ecosystem. Outside the big chains, hotels are also a very fragmented supply base globally. (US is an outlier).
So while the top of the funnel is changing (Search ->LLM), it will take a lot of effort to make meaningful dents to the transaction layer. I don't see LLMs spending that kind of energy / money to replicate this, esp. if there are bigger opportunities (i.e. going after Ad dollars).
Nevertheless, OTAs can't sit still and need to give customers a strong reason to book directly. Investing in loyalty programs will be a good start.
are you long on any of these marketplaces? Or short?
Great article and agree with your rationale! Where would you say leading online classified companies such as Zillow, CarGurus or Rightmove rank? Many thanks
Interesting highlevel view. Actually Booking is the largest hotel marketplace. They're incidentally also a rentals marketplace. Booking lives on SEM. They absolutely rule that space. Tripadvisor isn't a marketplace. They do reviews mostly and sell clicks to drive traffic to hotel and OTA websites. Getting inventory from hotels is painful and prices change every day or multiple times a day. LLMs are going to struggle with that. Booking and Expedia might become the plumbing for LLMs. But I doubt the LLMs will want to have hundreds of integrations to hotels for inventory and rates. FYI I just published the top 50 "marketplaces" in the hotel space: https://martinsoler.substack.com/i/184632493/great-distribution-starts-with-finding-the-right-guests
Thanks for this! My take:
Yes Booking is the largest hotel marketplace but their homes business is large and growing to the point it is rivaling Airbnb in size
I would call TripAdvisor a Lead Gen marketplace where they facilitate search only and not transactions (far left of the framework)
What the LLMs are doing with their commerce protocols (eg Google Universal Commerce Protocol) is meant to solve the integration problem you are referring to. Agree it’s a tricky one but I think it’s more solvable for hotels than it is for e.g. e-commerce
Nice reply!!! :-) Booking are competing with Airbnb on rentals. But the Airbnb market is quite small compared to hotels. The main business of Booking is hotels. Booking is larger than Airbnb in the rental space. Which is crazy.
I agree that between MCP and UCP (hopefully) will reduce the barrier to entry to get inventory and rates. But my theory is that most of the existing marketplaces will need to pivot to becoming the plumbing layer for the data to LLM search. It took Google more than a decade to get access to rates and availability of hotels, and still today, they have a thin and unstable API layer that fails. Booking, Expedia etc have actual inventory and rates. Expedia is already starting to pivot towards plumbing. They will resist it as much as they can. Look at Amazon blocking bots.
Excellent one. How would you rate marketplaces like olx / poshmark etc that is used goods market ? Not a very frequent purchase but AI can do (grading to negotiation) which traditional processes can’t. So an opportunity to multiply the TAM, grow CM …or is it just a marginal opportunity?
I think they are relatively defensible. Incredibly hard supply to aggregate and make legible, it’s very unlikely that LLMs go after that soon.
Excellent analysis. However I would personally add one other idea. First, I think its worth keeping in mind that the pureplay marketplace will likely always be able to correctly argue that their supply coverage is at least somewhat better than that of the LLM. That's because in practically all marketplace settings, getting the final chunk of supply is really challenging; and unlikely to be on the critical path for the LLM. And because we know for now LLMs aren't perfect either.
Whether that matters depends on two things. First as you point out, quite how fragmented the supply actually is. Clearly here hotels are less fragmented in the US than takeaway outlets for example. However, I think there's another variable; the actual importance of the transaction itself. The proxy for that would be ticket size. So its the degree of fragmentation (F) * ticket size.
I would argue that the high ticket size provides additional protection for the likes of Zillow, and arguably flight aggregation too (which is done by Amadeus, Sabre today and then resold by Expedia etc), whereas consumers are unlikely to be too bothered for small tickets (food delivery, ride-sharing). Put another way, the additional hassle factor of leaving ChatGPT - which knows more about me - may not be worth it for that last 1 or 2 burger joints in the SERP - but probably is if you are searching for a house or the last flight home for Christmas.