Q4 23 earnings Super Bowl
#239 What earnings reveal about e-commerce, SGE, and advertising
New from Wix: SEO resource center
This new addition to Wix’s SEO Learning Hub is packed with free resources created by top SEOs in the game.
The topics they cover range from backlink monitoring to creating SEO project proposals for potential clients and a whole lot more.
These downloadable checklists, templates, and toolkits let you develop smoother SEO processes and ramp up productivity—on any project.
Each resource can be customized for specific tasks and shared with teammates and clients for more effective collaboration.
While the SF 49ers and Kansas City Chiefs prepare for the Football Super Bowl, big tech companies had a Super Bowl of their own: annual earnings season.
The latest instance gives us a neat insight into the chatbot wars, the e-commerce landscape, and Google’s view on SGE.
Ad revenue: first down
Q4 was a touchdown for all big ad platforms:
Amazon ad revenue: +26.8% y/y
Google search ad revenue: +12.7% y/y
Meta's total revenue: +24.7%
Notice that Google's search revenue growth was only half of Amazon's. Investors also punished Alphabet's stock for coming in short of analyst expectations by about half a million.
But the biggest fumbled was Google Network revenue:
-5.6% CAGR over the last 2 years
As a reminder, Google's Network revenue consists of AdSense (display ads) and AdMob (ads in native apps):
AdSense allows sites to show display ads and other ad units that brands can bid on through Google’s ad marketplace (a constellation currently under review by the DOJ). Think Netflix running banner ads for The Witcher on imdb.com.
Over 2,000,000 sites use AdSense. Sites like HuffPost, Forbes, Buzzfeed, TechCrunch, Reddit, Glassdoor, Chegg, Twitch, IGN. What stands out? Most of them are publishers! Google's investment in News and developments in Discover are not just sticky push channels but a way to grow AdSense revenue.
Network revenue still came down to a sizeable $31 billion - more revenue than 3M, Nvidia, and Starbucks make in a year. A big chunk comes from publishers and they are starving from traffic malnutrition, which reflects negatively on Google. My argument was and still is that Alphabet cannot allow themselves to destroy a big portion of those $31 billion by launching SGE without a good way to keep at least some publishers alive.
Google's total Search revenue crossed $175 billion in 2023. Alphabet's Chief Business Officer Philipp Schindler was adamant in calling out that ads will play a role in SGE for investors:
As we shared last quarter, Ads will continue to play an important role in the new search experience, and we'll continue to experiment with new formats native to SGE. SGE is creating new opportunities for us to improve commercial journeys for people by showing relevant ads alongside Search results.
Bing is already charging ahead in experimenting with ads.
In 2023, Google paid a record sum of $50 billion (+8.2% q/q, +3.9% y/y) to be the default search engine on Apple devices, Mozilla, Samsung & Co., and build a moat around itself that makes it harder for old and new competitors to tackle it down.
From Elephant in the Room:
Add a $18b deal with Apple to be the default search on top, plus 36% revenue share for Safari revenue, plus a few billion to Mozilla and Samsung, and you realize Google built several traffic moats around its castle that give it a unique competitive advantage of user insights. Google knows what you want, how your profile is identical to thousands or millions of other users, and uses that intel to train predictive algorithms for the researchers coming after you.
When we take Alphabet's payment of $18 billion to Apple in 2021 as a baseline and multiply it with annual TAC (Traffic Acquisition Cost) growth on Alphabet's earnings sheet, we can assume that Alphabet paid Apple at least $20 billion in 2023. With over 50% of service revenue, Google is essentially financing Apple's RnD.
However, annual TAC growth slowed down from 7.4% in 2022 to 3.9% in 2023, while advertising revenue only slowed down from 7.1% in 2022 to 6% in 2023. The reason is mix shift: YouTube ads and subscriptions replaced Google Network's advertising revenue contribution. The lower the contribution of Search ads to total revenue, the less money Google pays to be the standard search engine and the higher its margins.
Amazon intercepts Google's shopping ad revenue
Alphabet course-corrected in 2023 but couldn't shake 3 key challenges that all come down to e-commerce:
Google Search's share of revenue across Alphabet, Meta and Amazon is at an all-time low with 42.9%.
Amazon's advertising revenue is at an all-time high of 13.1%. YouTube (~8%) and Meta (~36%) are flat, relatively spoken.
Amazon's advertising business has been the fastest of all players since Q2 2022 (except Q3 2021), and Meta is growing faster than YouTube or Search.
The quarterback in advertising revenue is shopping. While Google sees billions of queries every day, the majority of them are not monetizable. Shopping is one of the most profitable verticals, but over half of shopping journeys start on Amazon. That is a trend Alphabet cannot tolerate.1
The fact that all advertiser and e-commerce platforms grew significantly in Q4 2023 proves that the overall pie is growing. More people shop online compared to offline. Advertiser demand spiked as a result of more competition and recovering consumer demand.
From e-commerce shifts:
2023 was the first BFCM in which more people shopped online vs. offline and more on mobile (51.2%) vs. desktop. You’d expect more people to hit the stores as in-person events are hot goods after the pandemic years. But that was not the case. Google Trends shows a 4-point increase in searches for “black friday” in 2023 compared to 2022.
Before BFCM, Google added new shopping filters and exclusive deals to the SERPs. Sundar PichAI also called out AI-generated gift recommendations in SGE as a driver of shopping revenue:
Looking at our strong Search performance for the fourth quarter. Retail was a highlight. We continue to see a stronger start to the season up to and including Cyber Five. In Q3, we indicated that we were seeing early trends of consumers being very price-conscious, and we saw this play out in Q4. With promotional demand at an all-time high, deal seekers using Google had access to 2x the deals in the U.S. versus last season as well as a better shopping experience. Launches included a one-stop shop deals destination, new filters like GetItFast and AI-generated gifting recommendations in SGE. These new features drove incremental quarry growth during key shopping moments like Cyber Five.
China-based advertisers already make up 10% of revenue on Meta, which brings up the question of how much incremental revenue Alphabet made from Chinese 3rd Wave retailers Like Temu and Shein. Temu alone spends one billion USD on social media marketing and is expected to have made $16 billion in revenue but remained unprofitable in 2023.2 Like TikTok, Chinese platforms literally buy themselves into the market.
AI assistants enter the playoffs
All big platforms have either launched or worked on AI moats one year after Chat GPT’s pivotal breakthrough.
Amazon announced Rufus, a shopping assistant:
We launched Rufus, an expert shopping assistant trained on our product and customer data that represents a significant customer experience improvement for discovery. Rufus lets customers ask shopping journey questions, like what is the best golf ball to use for better spin control or which are the best cold weather rain jackets, and get thoughtful explanations for what matters and recommendations on products. You can carry on a conversation with Rufus on other related or unrelated questions and retains context coherently. You can sift through our rich product pages by asking Rufus questions on any product features and it will return answers quickly. We're at the start of what Rufus will do with further personalization and expansion coming, but we're excited about how it will make discovery even easier on Amazon.
Amazon already made it much easier for 3rd party merchants to upload products by turning photos into product titles and descriptions. Ebay is testing the same idea, and Shopify just launched it.3 I covered all of these examples in the recent State of Generative AI for SEO report.
Should Google be more afraid of ChatGPT or Rufus? Generative AI lowers the barrier to entry for marketplaces. But Rufus could play two leagues higher: the use cases sound, smell and look a lot like what users search on Google:
From broad research at the start of a shopping journey such as “what to consider when buying running shoes?” to comparisons such as “what are the differences between trail and road running shoes?” to more specific questions such as “are these durable?”, Rufus meaningfully improves how easy it is for customers to find and discover the best products to meet their needs, integrated seamlessly into the same Amazon shopping experience they use regularly.4
Google’s problem with generative AI is not just that it was OpenAI that brought it to the masses but also that it enables every platform to significantly improve the discovery and comparison part of the user journey. Amazon has just become a lot more dangerous to Alphabet, and so has Booking / Expedia, Tripadvisor, Pinterest, etc.
Google is fighting back by enhancing Maps with more ”LLM” to work more like natural search and hold context:
Let’s say you’re visiting San Francisco and want to plan a few hours of thrifting for unique vintage finds. Just ask Maps what you’re looking for, like “places with a vintage vibe in SF.” Our AI models will analyze Maps’ rich information about nearby businesses and places along with photos, ratings and reviews from the Maps community to give you trustworthy suggestions.5
Notice the importance of reviews and images for both Google Maps and Amazon. Both can use multimodal LLMs to extract more information from images and enrich them with information from reviews. It's likely that Google will replace the current SGE experience for local queries with this new, refined Maps product.
Meanwhile, Microsoft's AI assistant GitHub Copilot accelerates in revenue.
Satya Nadella on the Q4 2023 earnings call:
GitHub revenues accelerated to >40% YoY, driven by all our platform growth & adoption of GitHub Co-pilot...
We now have >1.3M paid GitHub copilot subscribers…
+30% QoQ & 50k+ organizations use GitHub copilot business to supercharge the productivity of the developers…6
Everyone eyes the new head coach on the field, SGE, who could fatten or starve Alphabet's cash cow. Alphabet's CEO revealed that the team made SGE 40% faster (in the US), which demonstrates that speed concerns are manageable.
Sundar PichAI also made vague comments about the utility of SGE, hinting at searchers using longer queries when interacting with SGE (also note the callout of sending traffic to publishers):
By applying generative AI to Search, we are able to serve a wider range of information needs and answer new types of questions, including those that benefit from multiple perspectives.
People are finding it particularly useful for more complex questions like comparisons or longer queries. It's also helpful in areas where people are looking for deeper understanding such as education or even gift ideas. We are improving satisfaction including answers for more conversational and intricate queries.
As I mentioned earlier, we are surfacing more links with SGE and linking to a wider range of sources on the results page and will continue to prioritize approaches that add value for our users and send valuable traffic to publishers.
But PichAI also dodged more specific questions.
Two, if I could. Sundar, a bigger-picture question, coming back to your comments earlier in the call on Search Generative Experience. When you think about the evolution of product over the next couple of years, how do you envision more traditional search and things like the Google Assistant continuing to evolve in a world of Search Generative Experience and Bard and what that might mean for elements of commercial and non-commercial search and how use cases might change in the years ahead?
Overall, one of the things I think people underestimate about Search is the breadth of Search, the amount of queries we see constantly on a new day, which we haven't seen before. And so the trick here is to deliver that high-quality experience across the breadth of what we see in Search.
If I had to interpret PichAI's answer, it sounds to me that not every query will be eligible for SGE. In this case, SGE would act more like a SERP Feature than replacing core search functionality and Google would develop a QDS factor (Query Deserves SGE) like it has for freshness (QDF = query deserves freshness).
An interesting snippet from Meta's earnings call brings back old rivalry vibes with Alphabet. Could Meta really have more training data than Alphabet?
Now the next key part of our playbook is learning from unique data and feedback loops in our products. When people think about data, they typically think about the corpus that you might use to train a model upfront. And on Facebook and Instagram, there are hundreds of billions of publicly shared images and tens of billions of public videos, which we estimate is greater than the common crawl data set. And people share large numbers of public text posts and comments across our services as well.
If Meta really has more, and maybe better, training data than Alphabet, it could use that data to train its many open-source models like LLaMA. If Meta succeeded in training the best models on the market, the open-source nature wouldn't just advance its own developer ecosystem but also destroy potentially trillions in market value for other LLM developers, including Alphabet.