Booking: The $170B+ A/B Testing Machine
How A/B testing turned two Priceline acquisitions (totalling $294m) into the world's most valuable travel company ($170B+ market cap).
Thanks for subscribing to SatPost.
Today, we talk about how Booking Holdings mastered A/B testing to build a $170B+ online travel beast.
Also this week:
50th anniversary of Jaws
LA Lakers sold for $10B
…and them fire posts (including a new LEGO set)
So, my kid is in school and we are officially on his schedule now.
That means no more random vacations on off-months like April or September.
We have to wait for summer break and do the annual jujitsu match known as “get your eyes gouged out by booking flights and hotels online for travel in June, July and August.”
I’ve spent the past few weeks doing a sweep of hotels on Expedia, Agoda, Kayak and Booking. The latter three are all under the Booking Holdings umbrella, which was previously The Priceline Group (the name change took place in 2018, and was more than a decade after Priceline laid the groundwork by spending $294m on two of the best travel acquisitions ever: Booking.com and Active Hotels).
Before I talk about the jujitsu on these online travel platforms (OTAs), let me ask you a question: do you remember Alec Baldwin in the film Glengarry Glen Ross? Of course you do. He plays a real estate sales guru named Blake in maybe the most iconic film cameo ever.
There’s a line during his immortal 8-minute rant when Blake says “a guy don't walk on the lot lest he wants to buy. They're sitting out there waiting to give you their money. Or you gonna take it?”
I can’t confirm but am guessing that employees at Booking Holdings send each other that quote on Slack every time someone lands on any of their websites.
Why? Because Booking is running over 1,000 A/B tests across its platform right now to part you from your dollars. It's one of the most A/B tested sites on the entire internet with countless variations of button colors, text spacing, discount offers and marketing copy.
Booking.com currently lists 31 million rooms in over 200 countries (homes, apartments, villas, hotels) and users book 1.5 million room nights every single day. That is just the accommodations. There are also numerous options for car rentals, flights and travel packages.
Between all those A/B tests and millions of users, there are hundreds of billions of different website permutations at any one time. Booking’s primary goal is to improve customer conversion by even a few basis points because that makes a meaningful difference when the entire group’s sales are at $24B a year (anything that includes the word “free Banh Mi” increases my conversion rate by 500%).
Thankfully, Booking Holdings has shared a lot of its methodology online and we’ll walk through the $170B+ company’s A/B testing machine including:
The Founding: Booking, Active Hotels and Priceline
Building The A/B Testing Machine
How Booking Does A/B Tests
AI and the Future of Booking
The Founding: Booking, Active Hotels and Priceline
The travel publication Skift has a fantastic hourlong read on how Booking Holdings came to be. It’s dubbed “The Oral History of Travel’s Greatest Acquisition” and let me summarize the main points.
In 1996, this initial seed of what would become Booking was founded in the technology epicentre known as University of Twente in Enschede, Netherlands. Geert-Jan Bruinsma had just graduated from the school and launched Bookings.nl, a site to help European travellers discover hotels and accommodations across the continent.
To get it off the ground, Bruinsma “borrowed” a lot of code from the website of hotel-giant Hilton (which was slowly dipping into the online game).
While I took a tasteless shot at the University of Twente moments earlier, Netherlands was actually an ideal place to launch an online travel agency (OTA) startup.
Why? First, the country has a long history as a global destination for trade. Second, the Netherlands is a smaller part of Europe — a population of 15m in a continent of 500m+ in the mid-1990s — and OTA founders had to prioritize international markets to make a business viable. This meant building the muscle to work with hospitality, airlines and other travel services that spoke in different languages and had different customs.
In parallel to Bruinsma’s entrepreneurial start, a trio of other Dutch founders had created a business called Bookings Online, with the URL Bookings.org. The two ventures merged in 2000 and smartly shook those low-value domain URLs (.nl and .org) for that juicy juicy Booking.com website (GoDaddy says I can pay a broker service fee of $141.99 to see if they can negotiate the domain for me; I’ll report back).
For real, though, that “booking” name was all the rage at the time. In 1998, Andy Phillips co-founded Activebookings.com, which would later become Active Hotels in 2001 following the Dotcom crash. That online platform aggregated hotel rooms across the United Kingdom.
The major opportunity for all these startups was that over 50% of the thousands of independent hotels across Europe didn’t operate on computers and were still using phone and fax. In fact, one of Active Hotels’ first products was an internet-enabled Alcatel phone that had access to a 36k modem that allowed hotel owners to update their rooms for rates and availability in real time.
Both Booking and Active Hotels created the booking engines as well as online management and payments systems for their hotel partners.
Interestingly, none of the Booking or Active Hotels founders were actually in the hospitality or travel industry. These were techy business folks spying a generational opportunity and it culminated with two of “travel’s greatest tech acquisitions” by US-based startup Priceline.
Founded in 1997, Priceline came to the OTA game with a pretty absurd business model. Dubbed the "Name Your Own Price System", a customer would say how much they wanted to spend on flight, hotel, car rental and sightseeing…and Priceline would come back with a non-refundable offer. It was an opaque mystery box but — as anyone that has seen the first episode of Family Guy knows — mystery boxes can be pretty enticing.
To be fair, this wasn’t even close to the dumbest idea during the Dotcom mania and Priceline ultimately raised ~$80m. With the help of pitchman extraordinaire William Shatner (shoutout to my fellow McGill alumnus) Priceline went public in 1999 and closed its first day up 4x with a market cap of ~$10B (despite losing $114m on sales of $35m in the year prior to its IPO, it had a higher valuation than United Airlines).
In September 2004, Priceline acquired ActiveHotels.com for $161m to expand its presence in Europe. Less than a year later in July 2005, it acquired Booking.com for $133m and merged it with ActiveHotels.
Both deals were done in cash and Priceline’s market cap had fallen to under $1B at the time of the acquisitions.
Active Hotels and Booking had briefly mulled a merger on their own because the synergies were clear, per Skift:
Active was very big in the UK and had some hotels in France, and a couple of other places on the continent, and was very oriented towards using affiliates to get demand. Booking.com, on the other hand, had a lot more hotels on the continent — the Netherlands, France, Spain — and they were very heavily oriented towards using [search engines].
The startups never pulled the trigger because neither party wanted to waste negotiating time while the online landgrab was happening. Also, there was (shocker) a clash of cultures between the Brits and the Dutch. This tension manifested itself once Priceline merged the teams, with many key execs from Active Hotels gone by 2006.
Either way, the results were undeniable: within 5 full years of the Booking.com acquisition, Priceline clocked a profit of $1.1B in 2011.
According to Skift, the deals “not only dislodged competitors like Expedia, Orbitz Worldwide, and Travelocity/Lastminute.com in the process, but Booking.com transformed the way travel companies market themselves, and consumers make hotel reservations.”
Perhaps unsurprisingly, the mastermind behind those deals — Priceline’s head of M&A Glenn Fogel — is now the CEO of Booking Holdings, which became the OTA’s official name in 2018 (note: every future reference in this piece will be about "Booking", sorry “Priceline” aka “Priceline Group”).
What led to the rapid growth? Booking kept offering more travel options at the exact time that the global appetite for tourism was going bonkers (international tourist arrivals went from ~500m in 1995 to ~1B by 2010).
Also, Booking had a much better business model than the Name Your Own Price System. To understand why, let’s walk through the most popular OTA business models:
The merchant model is when a marketplace buys up inventory for flights and hotels at a wholesale price and then re-sells them to customers at a markup. The OTA takes more risk with this model because it has to handle payments and customer service. But the benefit is that it can offer lower rates, mix-and-match products to appeal to different customers and have better control on inventory. Also, by collecting payment beforehand and only paying hotels after the guest stay, merchants get an interest-free loan which “enhances cash flow”. The American OTA leader Expedia long operated on this model.
The agency model is when the OTA platform connects travellers to the travel product (eg. hotel). The OTA doesn’t collect cash upfront and gets paid on commission for nights that guests actually stay. It is lower friction for the customer since they don’t have to pay until they arrive at the hotel. For hotels, there’s less “cash flow” but more conversions (Active Hotels COO Matt Witt: “We did debate it a lot but we never realistically had an opportunity to be a merchant because we were a startup at the moment. To be a merchant, you've got to win some big players and would have to establish a position, and our deal was [that this approach was] not realistic. If you've got a universe where clients can pay up-front or not pay up-front for effectively the same booking, it's almost a two to one difference.)
The agency model worked for Booking and Active Hotels. Meanwhile, American competitors such as Expedia and Hotels.com locked themselves into the higher cash flow (but fewer conversions) of the merchant model.
For customers, there are pros and cons to both. The major downside in the merchant models is that upfront payment. Trying to cancel or get a refund is — in industry parlance — a pain the ass. Conversely, the agency model processes payment when you arrive at the accommodation, which is much more flexible for planning.
While many believed that the merchant model would win out, Expedia tried to make it fetch in Europe and completely failed. Ultimately, Booking was able to create the strongest flywheel in the OTA game because the agency model converted more than the merchant model.
“Booking.com emerged as best in class in search engine marketing in travel,” explains Skift. “And it converted lookers to bookers so efficiently that it had an ever-larger stockpile of cash at its disposal to spend on Google, further widening its competitive advantage.”
The industry rapidly copied all of Booking’s conversion hacks — which we’ll discuss in later sections — but could never create the same flywheel and that’s why Booking is the largest accommodation provider in the broader $1.5T travel industry:
Booking Holdings: $172B (of which Booking accounts for 80% of the value or $137B)
Airbnb: $81B
Marriott: $71B
Hilton: $59B
Trip.com (formerly C-Trip from China): $37B
Expedia: $21B
Make My Trip: $10B
Tripadvisor: $2B
In sum: $137B in market value was unlocked by $294m in acquisitions. In terms of value creation, that is in the ballpark with Google acquiring YouTube for $1.7B in 2004. Or Facebook acquiring Instagram for $1B in 2012. Or Ray Croc realizing it was better for McDonald’s to own real estate and the only purpose of creating an assembly-line process for selling burgers was to make sure your tenant had cash flow to pay rent on said real estate.
Booking's savvy business deals were soon combined with a pioneering tech infrastructure: A/B testing at scale.
***
Building The A/B Testing Machine
Following the merger of Active Hotels and Booking.com in 2005, the newly-combined entity began instituting a rigorous A/B testing process.
Here’s how A/B testing fit into the Booking flywheel:
“agency model converts more bookers” —> “more profitable” —> “spend more money on Google search” —> “get even more traffic” —> “A/B test the F out of the buying funnel” —> “convert more bookers” —> “even more profitable” —> “spend even more on Google search” —> “get even more traffic” —> “A/B test the F out of the buying funnel” —> “Bob’s your uncle”
A/B testing is a way to turn all that bought traffic into insights. Well worth the effort, especially with Booking now spending ~1/3rd of its revenue on Google.
OTAs are one of the top verticals for spending on search ads. It’s up there with other high lifetime value (LTV) categories such as:
legal (personal injury, immigration, divorce, tax)
financials (credit cards, auto loans, mortgages, all types of insurance)
healthcare (dental, private clinics, rehab, pharmacies, cosmetic surgery)
home services (HVAC or plumbing because you didn’t listen to your wife when she said “please stop flushing tissues in the living room bathroom until we figure out what is going on”)
In 2024, the four largest OTAs (Airbnb, Booking Holdings, Expedia Group, Trip.com) spent a total of $18B on Google search, with Booking responsible for 41% of that spend ($7.3B).
Since A/B tests are such a net positive for Booking, the leaders built the entire organization around optimizing the A/B testing infrastructure (and they share a ton of details in a corporate blog).
Today, a new Booking employee can start running A/B tests for tens millions of people within weeks of joining the company.
In 2017, Stuart Frisby — who was Booking’s Director of Design from 2016 to 2019 — did a 27-minute presentation of Booking’s A/B testing process, which had led to a “conversion level 2-3x the industry average.”
The talk has a ton of nuggets and a key overarching insight is that an A/B testing infrastructure is as much about the culture as it is about the technology.
Here are 7 takeaways from Frisby’s presentation:
Create Hypotheses and Test Everything: Find someone that looks at you the way that Booking looks at A/B tests. They create hypotheses and literally run tests on every inch of the platform from the consumer portal to the internal tools to the backend for its accommodation partners. Mobile, Desktop. Tablet. All of it. Frisby says that “If it can be a test we test it. There aren't things that we don't test and things that we do test. We test absolutely everything. We test performance improvements. We test product features. We test everything that you can imagine you can test. If we can't test it, then we probably don’t [put it on the platform].”
No HIPPOs: So, HIPPO is either the most legit or most cringe corporate acronym ever (HIghest Paid Person’s Opinion). Since testing is the holy grail at Booking, a highest paid person’s opinion — usually a senior employee — is almost always secondary to the data. While C-suite execs can offer up new ideas, those ideas have to be proven out on the platform via A/B tests.
Org Chart Built For Testing: Back in 2017, Booking had 150 designers and they were embedded in small units that included engineers, copywriters, product developers and marketers all working together. These units come up with hypotheses and can test them across the platform. Teams don’t have to wait for senior execs to give a greenlight on a testing idea. Booking tries to remove bottlenecks and dependencies. To cross-pollinate ideas and maximize hypotheses creation, members of each unit are interchanged with other teams at least once a year.
100% Access to Data: For these individual teams to form the best hypotheses, Booking provides as much data as it is legally allowed to and employees can access them in a data warehouse (some privacy restrictions applies). The organizational make-up and widespread access to the data “democratizes” testing across the entire company.
Test Small: Frisby says 90% of tests fail. A major reason is that a lot of the low-hanging fruit has already been picked. Either way, Booking is happy to do thousands of small A/B tests instead of a major overhaul that changes multiple variables. Because of all the traffic that Booking gets — especially with the help of those paid Google searches — every small test can be exposed to millions of users and be statistically significant. Taking into account all of the atomic tests (>1000 at a time), there are more variants of Booking than every human that has every existed (>100B). While each individual test may not be huge, allowing experimentation and compounding all of the small improvements together makes a notable difference.
Guidelines, Not Rules: Teams have complete freedom in what they can test as long as a sound hypotheses and methodology is provided.
Research Informs Hypotheses: Booking spends a lot of resources to have employees on the ground and seeing how actual customers use the platform. They then form hypotheses for the product development team to test. Frisby notes that this is a reverse of how consumer usage is usually applied, “[At Booking], user research is an input to hypotheses testing, rather than the output of a product development process.”
A notable outcome of Booking’s approach is that the website’s design was literally created by the customers:
Your customers drive the product. So, Booking.com doesn't look the way that it does because I decided it should look that way. If I had done it, it wouldn't look that way I can promise you. It looks that way because — over 15 years — customers have told us that this is the product that they want in small measurable incremental steps.
Booking has built internal testing and data management tools to facilitate A/B testing at scale. But, also, the culture requires that everyone — from the most junior employee to the CEO — is willing to let data make the decisions.
“Are you OK being wrong?” asks Frisby. “Are you OK being proven wrong? Are you OK having your ideas tested like everyone else? Are you OK with changing your strategy based on an A/B test? Are you OK with relinquishing control of your product? Are you ready for unintended results?”
Those are very legit questions and A/B testing probably doesn’t work for every organization.
Let’s take a closer look at how Booking makes it work.
***
How Booking Does A/B Tests
An important point that Frisby makes about Booking in his A/B testing presentation is the difference between an idea and a hypotheses.
Ideas are truly a dime a dozen. I know this because all I do all all day is spout untestable, unverifiable and unfalsifiable ideas that sound awesome but are completely impractical (“I wonder what would happen if Chipotle made an AI-powered hardware button for your shirt to summon guacamole at anytime.")
Conversely, Booking is built around hypotheses. To paraphrase Matt Damon’s character in The Martian, Booking “sciences the shit out of its A/B testing.”
And Booking has to because: 1) the low-hanging fruit is gone and they need to keep finding new ways to drive conversions; and 2) consumer behaviour changes, so the reliability of an A/B testing result has the shelf life of a year or two and has to constantly be re-tested.
One of Booking’s largest flops in the mid-2010s was to launch an offering called “Villas”, which was a separate platform for booking villa-type homes. It was such a significant overhaul, that they couldn’t actually incrementally A/B test this product.
They launched and it flopped.
So, yeah, teachable moment: A/B testing is important.
How does Booking actually do these A/B tests?
In 2020, Booking product manager Saurav Roy gave an online presentation of the company’s internal A/B testing tools and it’s worth chewing on.
“We absolutely need a hypothesis,” Roy says. “Otherwise it’s like throwing spaghetti on the wall to see what sticks. Hypotheses protects us from our own biases.”
Roy provides an example statement that will look familiar to anyone that took Grade 8 biology or read a chapter out loud about that dude Francis Bacon:
“Based on [evidence], we believe that if we [implement a change] for [this customer segment], it will help them [impact]. We will know this is true if we see [your expected change] in [primary metric]. This is good for our business because an increase in [primary] metric is an increase in [business KPI.]”
One of the absolute north stars for Booking is…errr…bookings per customer site visit (aka do they convert).
Two caveats to note:
Game-ability: A platform can make a decision that drives short-term conversions but damages the customer relationship in the long-run (example: Booking creates a scarcity mindset by saying that there are only 2 rooms left on a listing; Booking actually got in trouble with this because while it was true that the platform’s own inventory only had “2 rooms left”, the actual hotel had many rooms).
If the customer find out, they would lose trust. So, Booking must make it clear that its available-room counter is specifically for the ones reserved on its platform. More broadly, Booking closely monitors if different A/B tests are conflicting with each other or creating poor short-term incentives.Multiple levels of impact: A related idea is that Booking has to closely follow how a change at the beginning of a consumer’s journey might impact a decision multiple pages and clicks later.
Below is Booking’s internal-looking A/B testing tool. Some notable variables are time frame (minimum 2 weeks; longer for more important strategic changes), number of visitors (need to hit a few million for statistical significance) and the entire test needs to meet a confidence interval threshold (>90%).
Here is a sample of a font and color change to a call-to-action button:
According to Roy, you can go on your laptop right now and see multiple versions of Booking by using a VPN, logging incognito, using a different device or checking at different times.
Again, there are thousands of micro A/B tests happening at one time and even small changes can have a material impact.
Let’s take a look at one example: in this comparison between two versions of the same listing, the bottom one converted better simply by reducing the spacing between the name of the property and and the description of the property (red box).
I clicked through the Booking blog and watched a bunch of other videos to collect other A/B testing winners:
Arabic writing (this one was pretty obvious but Booking originally had its Arabic sites read left-to-right when the text is read the opposite way; still, they tested it and made change after results proved better conversion)
“X number of rooms left” (classic scarcity play which gets me 76% of the time, damnit)
Showing WiFi availability didn’t matter (however, WiFi availability for specific activities — such as streaming — did move the needle)
“Free cancelation” (showing this text as early as possible in the transaction funnel helped conversions…probably a trust thing)
Showing “Toilet Paper Included” (self-explanatory)
# of Users Online (social proof and that urgency)
Reviews (highlighting different types of reviews for different properties could change conversion)
Most of these A/B test results are obvious now and used across the OTA industry…but that’s mostly thanks to Booking’s efforts.
Booking has a repository for results of these tests and employees can also sign up to receive e-mail alerts rounding up the results. It’s a lot to stay on top of but I’m guessing AI summaries are really useful now.
Speaking of AI, the rise of LLM-powered chatbots present interesting challenges and opportunities for Booking.
We’ll explore more in the next section.
***
AI and the Future of Booking
At some point in the past 4000 words, I’m sure you had a thought “hmm, Booking seems awfully dependent on Google search.”
Yes it does and that’s why the OTA industry was put on edge when Google launched Google Hotel in 2019 (this product aggregated the listings for hotel searches into a new module).
It wouldn't be the first time that a dedicated Google search module impacted other vertical search services. Yelp’s business was clapped hard when Google started defaulting to Google Reviews and Google Maps for local searches. Seeing the same threat for e-commerce, Bezos took action to build what is now Amazon’s $50B+ a year ad business.
Writing on the launch of Google Hotel, Ben Thompson noted that while it was a new toll booth that OTAs had to pay, it was also just a superior option for users than 10 blue links:
At this point the conclusion seems easy, no? Google being evil, yet again. In fact, while I understand the frustration of Expedia and TripAdvisor, I think it is a bit more complicated.
Start with the theoretical perspective: the stable structure of an Aggregator-dominated market is that the Aggregator controls demand and suppliers come onto the Aggregator on the Aggregator’s terms. In other words, there are three players in the value chain: suppliers—Aggregator—demand.
Notably, though, that has not been the case in travel, where Google has controlled demand but OTAs have controlled supply.
One way to achieve equilibrium would be for Google to become the one OTA to rule them all. Indeed, this would be difficult to compete with (and was a fear when Google acquired ITA in 2010). The truth, though, is that OTAs have put in significant effort to bring suppliers on board, and they deal with all of the pesky payment and customer support issues that Google loves to eschew. Instead Google has realized it can get OTAs to effectively pay Google to take care of the messy parts for them.
With the hotel module, Google captures demand more efficiently, which not only makes Google search more attractive to end users, but also transforms OTAs into suppliers, paying to provide the service that Google doesn’t want to.
Booking CEO Glen Fogel was quite sanguine on the Google Hotels threat saying that “In the end, what’s most important for us is to get customers to come to us directly…For us to have our own future is to create a service that is so wonderful, so good that people just naturally will come back to us directly. And we will not be as dependent on other sources of traffic.”
Fogel's view proved correct: over the past 5 years, Booking’s sales are up 60% from $15B in 2019 to $24B in 2024…meanwhile, Booking's stock is up +226% while Expedia is up +100% (TripAdvisor is down -29% over the same span; the company started out with reviews and sent traffic to other OTAs but Google ate its lunch and Tripadvisor’s attempt to build a direct booking platform was not a silver bullet and its current market cap of ~$2B is a far cry from a peak of $16B in the mid-2010s).
But now a new threat is rising with AI and chatbots.
An obvious travel use case with AI is “hey LLM, can you plan and book for me a perfect two-week trip to Europe and make sure I have an Aperol Spritz at least once a day…no mistakes please”.
There is a new protocol to facilitate such a request and its called Model Context Protocol (MCP), a standard to connect LLMs to external tools and data bases such as OTAs. All these AI agents we hear about will be using MCP to complete tasks.
What happens to Booking when AI agents proliferate? Will the OTA get completely disinter-mediated by the ChatGPTs, OpenAIs, Groks and Geminis of the world?
It doesn’t really matter how good your A/B tests are if no one is visiting the site.
The challenge and opportunity for Booking is in discovery. At present, the best way for users on Booking to fully express intent is through the dozens of filter options on the platform.
The conversational approach for LLMs is a much better way to reveal customer preference, per Adrienne Enggist (Booking’s Senior Director of Product Marketplace):
“When ChatGPT launched in 2022, I got this tingle. It reminded me of the early days of broadband access—this massive opportunity to change how people engage with travel. We knew this could help us finally crack the discovery challenge. We’ve always been really good at the last mile — getting people from search to booking. But discovery was different. We needed a way to meet customers earlier in the process, when they were still figuring out what they wanted.”
In the summer of 2023, Booking partnered with OpenAI to roll out such a discovery product called the AI Trip Planner.
That tool is primarily meant for use in Booking’s mobile app.
But what about within the major AI chatbots themselves? An important “moat” for Booking is that it does have all the proprietary data including those 200k+ rooms aggregated over decades across a fractured industry.
They’re not going to stupidly just give that away.
Another $170B+ marketplace in a similar boat is Uber. The company’s CEO Dara Khosrowshahi — who was CEO for Expedia from 2005 to 2017 — was recently on the Decoder podcast and asked about MCPs and whether they were a threat to his business:
Nilay Patel: [The AI agents will] all disinter-mediate your platform. You have to do the hard work of creating liquidity of supply and making sure the Toyota Camry appears [when it is ordered] and [you would let them] own the customer relationship? Why would you ever participate in [MCPs]?
Dara Khosrowshahi: It's a really good question. By the way, we talk about it all the time. And time will tell as to what the right decision is.
I believe in running open platforms. I don't believe in companies that try to fight the course that technology is taking. They always like get left behind.
I think we should go to where consumers want us to go. That's all nicey-nicey, right? But at the same time, one of the advantages that we have is we have unique inventory, right?
Essentially, we've got 8.5 million drivers and couriers. We got over 1.2 million merchants out there. And it's very difficult for a company who has an agent…to disinter-mediate [us] and go direct to the almost 10 million pieces of inventory, if you want to be quite impersonal.
These are people, these are businesses, etc. It's very difficult for them to draw that inventory. So if you have unique inventory, then while you may make your inventory available to these agents, you can charge a toll for that inventory.
I think the agents, if they bring us more demand and it means more orders for restaurant partners or it means more rides for our drivers, that's worth something for us as well. At the same time, because we're becoming more and more an everyday use case, the frequency now of our average user is six times a month and it continues to increase. We can create our own local agents as well, right? We know it's your commute time, why don't we get you a car?
Maybe we can actually make a dispatch to the car before you actually push the button to get the car because we can predict that you're gonna do it. So, I think as it relates to AI and these agents, etc., we want to work with these players.
I think we come from a place of strength because of the unique inventory and the fragmentation of these markets that we're organizing. And then we're gonna build our own agents as well. And I think that end state — as long as we're thinking about the consumer, we're thinking about our earners and our merchants — I think we'll be okay.
Using Uber’s framework, Booking seems to be in a decent position. Booking has the unique inventory of fragmented accommodations. It can receive new traffic and charge the major chatbots (ChatGPT, Gemini, Grok, Claude) a fee to access the listings.
Or, as Booking CEO Fogel noted, the company will have to keep making its own apps and digital properties the go-to source. The Booking AI Trip Planner is obviously a start.
At the top of this article, we talked about the film Glengarry Glen Ross. Well, the most famous line from Alec Baldwin’s speech is when he’s standing next to a blackboard with the letters “ABC” and drops the sales mantra to end all sales mantras: “ABC, Always Be Closing.”
As long as Booking keeps closing new room accommodations, it’ll have the unique inventory to stay relevant and thrive in the age of AI. In which case, it can ABAB: "Always Be A/B testing" to keep the Booking flywheel going.
This issue of SatPost is brought to you by Bearly AI
Are you looking for an encrypted and anonymized AI chat tool?
Then, try the Bearly AI research app, which provides an easy-to-use UX to access leading models from OpenAI, Anthropic, xAI, DeepSeek, Gemini and more.
While some AI providers keep chat logs indefinitely even after a user deletes them...all Bearly AI chats are encrypted and requests to LLM providers are completely anonymized.
If you're looking for AI chat privacy, get access to leading LLMs on our platform and try one month FREE of our Pro Plan using the code BEARLY1.
50th anniversary of Jaws
I just saw the trailer for the Jurassic World Rebirth, the 7th instalment of the franchise (ugh) starring Scarlett Johansson, Mahershala Al and so so so so so many dinosaurs (ugh).
There is a 0.00% chance I’m watching this summer blockbuster stop.
But the existence of this film is a useful way to understand the genius of Jaws: the original summer blockbuster which came out 50 years ago this weekend and remains a masterclass in suspense-building that so few recent summer blockbusters have tried to replicate.
In Steven Spielberg’s 1975 monster thriller classic, we don’t see the shark until 81st minute of the 124-minute film (65% of the way in).
This happened to be a stroke of storytelling luck.
The original script had a lot more shark in it but Spielberg’s shark animatronic kept breaking down. The electric wiring short-circuited in the ocean saltwater off of Martha’s Vineyard. So, Spielberg found other ways to “show” the shark (filming from POV of shark in water or people’s terrified reactions).
As the saying goes, “constraints breed creativity”.
In an old 60 Minutes interview, Spielberg explained how the failed animatronic made the film better:
“I decided to shoot without the shark and it increased, or torqued up the suspense of the movie because you really didn’t know where it would come from next.
So, rather than seeing the shark in every scene, I played a lot of the fear from the people in the water. From seeing their legs kicking. From the POV of the camera moving along the surface of the water.
That turned the movie into more of an exercise in suspense than just a horror film.
[With a working shark], I probably wouldn’t have had a movie as successful.”
The 1975 film made $470m on a $9m budget and became top-grossing film ever (before Star Wars knocked it off in 1977).
Spielberg did the exact same playbook with the T-Rex in 1993’s Jurassic Park: we don’t see the apex dinosaurs until the 63rd minute of the 127-minute film (50% way in). This also became the top-grossing film ever at the time (nearly $1B box office).
In total:
the shark in “Jaws” is on screen for ~4 of the 124 minutes (3%)
the T-rex in “Jurassic Park” gets ~8 of the 127 minutes (6%)
I hope our social-media attention spans can still handle those wait times, but just not sure anymore. We’re definitely not going to find out with Jurassic World Rebirth.
LA Lakers sold for $10B
The Buss family sold a majority stake of the LA Lakers to Guggenheim Partner CEO and private equity billionaire Mark Walter at a $10B valuation.
The most expensive dick-measuring contests franchise sales in US professional sports history:
LA Lakers (2025): $10B
Boston Celtics (2025): $6.1B
Washington Commanders (2023): $6.05B
Denver Broncos (2022): $4.7B
Phoenix Suns (2023): $4B
The late owner Jerry Buss bought the LA Lakers, LA Kings, The Forum stadium and 13,000 acre ranch for $67.5m in 1979.
Buss sold the LA Kings in 1988 and The Forum in 2000.
He died in 2013 and management of the Lakers fell to his 6 children, with daughter Jeanie ultimately taking the reins.
After years of in-fighting, the siblings signed off on this massive deal and Jeanie will stay on as Lakers governor for a few years (this Jeanie post was definitely the funniest Lakers sale-related post).
Because my brain has rotted from spending too much time on finance Twitter, the very first thought I had was “did this sale underperform the S&P 500?”.
Looked it up and, since 1979, a $67.5m investment in the broad US index would now be worth $13B.
I joked “So, the Buss family underperformed by $3B in exchange for 11 LA Lakers rings and infinite cultural clout. Worth it."
Many people yelled at me for:
Not adding LA Lakers operating income
Not taking into account S&P 500 dividends
Re-investment of funds for LA Kings, The Forum and the 13,000-acre ranch
Proper breakdown of LA Lakers ownership (Buss family closer to ~70% ownership while Mark Walter previously acquired a 27% minority stake)
Bro, it’s napkin math. So, let’s just call the dividends and operating income a wash. The LA Kings majority stake sales was ~$20m. That re-invested in the S&P 500 since 1988 would probably be $3B+. The Forum was sold for also ~$20m in 2000 and that invested in the S&P 500 is worth $1B+ now. If they still own that ranch, it’s probably worth $500m+.
By this (truly awful) napkin math, the Buss family would be at $12B to 15B. Basically on par with the S&P 500...AND they got to hang with Magic, Kareem, Shaq, Kobe, Pau, Lebron, Luka, Jack Nicholson, Flea from Red Hot Chili Peppers, Leo and Denzel.
Obvious no brainer.
The other pretty obvious no brainer to me is that owning a top-tier North American sports franchise is a bulletproof investment in 2025.
To wit, Steve Ballmer — who bought the LA Clippers for $2B in 2014 — talks about why he believes the asset value of LA Clippers is “more secure” than Microsoft:
“Unlike Microsoft, [the LA Clippers] cannot go to zero. The asset value...is far more secure than Microsoft. Why? They're not making more of them. As long as anybody in the world is getting richer, the buyer pool will only go up and people don't buy them for their earnings.
I wish we had more earnings but -- at the end of the day -- people are buying them more like a piece of art. I mean not everybody. Some people don't like negative cash flow but at the end of the day, the Clippers have the best market in the world. I mean, you don't want to own a basketball team in [any other city other] than maybe Miami.
The players want to play in LA. If you're a buyer, where do you want to go? If you don't live in LA, where do you want to go? Well, you want to go to LA or you want to go to Miami. You don't want to go to New York in the wintertime. If you're a foreign buyer, you want to go to LA.”
To be sure, the Clippers are a drop in the bucket of Ballmer’s ~$140B net worth, with Microsoft accounting for 90%+ of that wealth (he’s held ~4% of the software giant since leaving in 2014 and is the single-largest individual shareholder; Microsoft cuts him an insane $1B a year in dividends).
Ballmer actually outbid Mark Walter for the Clippers. But Walter has ended up with the LA Dodgers and LA Lakers. That’s a John Wick-level of “I’m thinking I’m back”.
After seeing Walter back up the Brinks truck for the Dodgers and winning the World Series last year with Ohtani, Lakers fans are creaming themselves rn. While there is certainly sadness that the NBA is losing one of its last true “mom ‘n pop” ownership groups, Walter & Co. means that the “The Lakers can finally be run like a real business” per a recent ESPN article.
Anyway, I posted this Ballmer clip from a must-watch Acquired podcast interview and many people had contrary feelings.
A common counter-take was that there is no guarantee that the NBA will continue to be preeminent as a league. The game is solved (everyone takes a million 3s) and sports can go out of fashion if the younger generations lose interest (look at the MLB).
Those are fair points but live sports is just the last thing really holding together synchronous media experiences. This was the case with the rise of social media and streaming and will only continue to be the case in a world of AI media abundance.
While the NBA game may have fewer viewers, the players still have major individual star power (perhaps more than any other league) and the off-court narrative is always interesting (often to the detriment of the actual game; as in, this Lakers news literally dropped before Game 6 of a now 7-game NBA Finals and totally overshadowed OKC and Indy).
Ballmer and Walter gonna be very Gucci with their investments.
Links and Memes
AI-Generated ASMR: A new trend on TikTok is people using Google’s Veo 3 video model to make clips of people cutting things that can’t be cut (concrete, lava rock) put against soothing ASMR sounds. The boomers are eating these videos, which are either the worst thing ever or the best thing ever.
***
The Titan submersible implosion…happened 2 years ago this week. Five people were killed on the OceanGate vessel that headed down to see the Titanic wreckage, including the company’s founder Stockton Rush. Two documentaries have come out on the accident. I watched the one on Netflix (Titan); it’s decent and follows the whistleblower who tried to prevent what was a completely avoidable tragedy. I’d also recommend checking out this 60 Minutes interview with James Cameron from 6 months ago. Cameron is obsessed with deep-sea diving and has very technical details on what went wrong for an industry that has an otherwise high-safety record (it’s pretty wild that filmmaking is only Cameron's 2nd true passion).
***
Nerds are the top-selling candy in America…with annual sales of $900m, it has knocked off Skittles, per Food Dive. This flippening is all because Nerds invented Nerds Clusters, which are the most addictive diabetes bombs if you ever tried them.
Nerd sales have jumped 1,700x from just $50m in 2018. That was the same year that Italian confectionary giant Ferrero Group — which built a $40B+ empire on Nutella — acquired Nestlé USA’s candy arm for $3B (it owned Nerd, Butterfinger and Baby Ruth).
Ferrero Group was the brainchild of Michel Ferrero, the real-life Willy Wonka who invented Ferrero Rocher, Tic Tac and Kinder Surprise. They took that inventive DNA and made the first real change to Nerds in its 35-year history: crunchy candy outside gummy middle. Now, these Clusters fly off the shelf at the theatre, airport bookstores and long-distance athletes looking for sugar boost.
These are addictive AF but it won’t be long before RFK Jr. is knocking on the factory doors.
***
AI-Generated Celebrity Math Tutors…as in, people are using AI to dub calculus lessons on top of Sydney Sweeney and Drake clips. Hate to say it, but the future of education is legit here.
PS. You may have seen a viral story on how MIT showed ChatGPT was making us “cognitively bankrupt”. BS Detector looked into the study methodology and it’s probably too soon to draw that specific conclusion.
***
WhatsApp Gets Ads: In 2014, Zuck acquired WhatsApp for $19B. We all knew those ads would eventually come, and yet it’s still somewhat surprising now that they announced ads will show up in the “Updates” tab. I’ve actually never ventured to that part of the app but apparently it's where the 1.5B WhatsApp users can put status updates.
While WhatsApp says nothing changes on the messaging side, I’m assigning a non-zero probability that Zuck will one day make me watch 10-second unskippable pre-roll ads before sending a message. I will then pay some subscription fee to get rid of those ads. Well played.
More broadly, the fastest-growing part of Facebook’s entire $160B+ ad business is the “WhatsApp Business Chat”, which did $10B in 2024 and directly links a customer to merchants. These ads sold on Facebook and Instagram and are very popular in South and Southeast Asia.
…and them wild posts:
A/B testing is evolutionary.