A 40% bloodbath, is Applovin guilty of this?

$AppLovin(APP.US) has been shorted again! This time it’s a double whammy, with Culper Research and Fuzzy Panda Research both releasing bearish reports simultaneously. The shorts have been attacking continuously for half a month after the earnings release, and AppLovin, which once heavily punished the shorts, has finally been “beaten” badly.

Overall, most of the views from these two institutions are similar to those of independent analyst Lauren Balik and Bear Cave.

(1) Culper has added evidence of “silent installations” with backend code that bypasses user permission to install apps to elaborate on AppLovin's advertising fraud issues;

(2) Fuzzy Panda mentioned the “theft” of Meta data, which was also noted by Bear Cave. Fuzzy Panda also believes that AppLovin has violated regulations by displaying fraudulent gambling game ads, showing pornographic ads to child users, and establishing multiple subsidiaries only to deregister them for tax evasion, as well as issues related to the reduction of shares by technical executives.

Among the “charges” listed by the shorts, Dolphin Research believes that the main issues worth paying attention to are “unauthorized automatic installation and download” and “Meta data utilization.” The potential risks here depend on how Google and Meta view the situation; if the two giants take corresponding avoidance measures, it may affect how the

Charge First: Deceptive Advertising Effects - Inflated Conversion Rates

Since 2023, Applovin has experienced a two-year surge, with both the company's descriptions and customer experiences primarily stemming from a significant increase in its advertising system's ROAS (return on ad spend). The company further explains this as the effect of the AXON model, but the market has been unclear about the true power of the AXON model, leading to it being jokingly referred to as "black box technology."

In the gaming sector, one of the metrics for measuring advertising ROAS is the click-through rate or installation rate, as well as the underlying costs incurred by customers. However, Applovin's existence has misled users or directly sent false click-through rates, and collaborated with OEM manufacturers for pre-installed behaviors, resulting in significantly inflated conversion rate metrics. For example, Applovin claims an advertising click-through rate of 30-40%, while industry experts believe it is generally around 3-4%, with the inflated portion being ineffective clicks.

1. Misleading Users to Click Download Links.

For instance, the advertisement sets up a link to redirect to Google Play or the App Store on the full-page ad, but the "x" button to close the ad has a very small touch area, making it difficult for users to close it or easily misclick and be redirected to the app store.

For example, in a screenshot of the Fuzzy demo, during the advertisement playback of the mobile game "Endless Winter" international version, when users clicked the close ad button "x," they were instead taken to the App Store download page, and it took seven attempts to finally close it.

2. Sending False User Click Data

Even when users do not click, the Applovin SDK sends interaction data of user clicks on the download link to the backend. Additionally, some ads come with a feature that awakens the app store link without requiring user clicks or other interactions.

3. Binding OEMs to Obtain Silent Installation Permissions

In addition to the inflated "click-through rate," Culper mainly pointed out the tricks behind the "installation rate."

Applovin collaborates with mobile OEM manufacturers (Samsung, T-Mobile/Sprint, OPPO/RealMe, Xiaomi, TCL, etc.) to pre-install Array (a mobile application management suite that displays similarly to an app store) and AppHub (a centralized repository for Android applications) on user terminal devices, enabling the display of advertisements on the device while bypassing user consent to achieve silent installation (as long as the user clicks the ad) Applovin wants to distribute applications.

Generally speaking, pre-installed software has higher permissions on devices. As shown in the image below, after pre-installing the Array suite, app recommendation ads will appear on the main interface. This advantage of traffic exposure on the device is akin to a splash screen ad within an app. According to Applovin's official website, Array has currently reached 1.4 billion mobile users, meaning that 1.4 billion devices have pre-installed Array.

AppHub is also software pre-installed on phones that can provide app recommendations and distribution, as well as download functions. It has been granted the "INSTALL_PACKAGES" permission by phone manufacturers—allowing the app to directly install other application packages (APKs) on the device.

Having this permission allows for downloading applications on Android without going through Google Play, and it also does not require separate user download authorization.

Therefore, the process is:

When a user clicks on an ad displayed by Array, the embedded Applovin SDK triggers AppHub. AppHub uses the "INSTALL_PACKAGES" permission to silently download the app recommended by the ad, without prompting the user for permission like Google Play does.

Dolphin Research Interpretation: "Deceptive advertising effects" is an issue mentioned by almost every short-selling institution. However, Dolphin Research believes this may be one of the lighter issues among several charges. While Applovin may have some compliance issues and false conversion data, Dolphin Research repeatedly emphasizes that, based on research information, advertisers are generally aware of this, and this phenomenon is not unique to Applovin in the industry.

Short sellers claim that users have complained multiple times on social platforms about this issue, but do users only complain about Applovin? The data shows otherwise. For example, in the image below, several users on Reddit have also complained about the difficulty in closing Unity ads.

On the other hand, regarding the high load rate of ad bombardment, there is no disagreement from either the game developers releasing ads or those placing ads. Especially for casual games that primarily rely on low spending/advertising monetization, continuous bombardment of ads is necessary to achieve an increase in DAU For game developers (i.e., advertisers), what they truly care about is the cost of investment and the resulting real DAU and revenue, which is the ROI of input/output. As long as they can achieve the highest ROI on the Applovin platform, the compliance of the process and any slight increase in false click rates are not that important in the eyes of developers.

It is worth mentioning that these casual game developers are likely to be both advertisers and publishers (initially acquiring customers through Applovin and later profiting from ads through Applovin). Therefore, even if there are some false conversion data on Applovin, developers are still pleased when they become ad publishers.

This is also what Dolphin Research mentioned earlier; in the gaming sector, for some game developers using the IAA in-app advertising business model, "investing in ads to acquire customers and accepting ads to profit" may be an internal cycle game. Developers may earn a little in and out, but Applovin, which profits from bidding price differences, is always a steady winner.

The so-called ad fraud is not uncommon in the in-app advertising market for games and has existed for a long time. If this ROI containing false conversions does not satisfy clients, it would have been abandoned in the highly competitive market long ago. Existence is reasonable; the so-called "low-quality ads" in the eyes of short sellers do have value in the eyes of clients, especially when the CPM for ad bidding on Meta and Google continues to rise, and smaller platforms are constrained by Apple's IDFA, resulting in poor conversion rates and a limited user base. Applovin, which can provide services throughout the entire product lifecycle (from customer acquisition to monetization), has instead become a preferred choice.

Charge Second: Stealing Meta's "Results"—Using Meta's user matching data and reporting more favorable bids

This issue was mainly raised by Fuzzy Panda, who confirmed the existence of the problem through their own terminal user experience, multiple expert interviews, and whistleblower information. They determined that Applovin is leveraging Meta's user matching data to enhance its ad recommendation accuracy and is using its ad matching platform Max to report more advantageous CPM based on known Meta bidding.

Fuzzy listed a statistical analysis provided by a "secret whistleblower," showing that Applovin's ad accuracy in e-commerce advertising is highly correlated with Meta.

1. Terminal Experience: Applovin Tracks User Behavior in Meta

Fuzzy Panda believes that Applovin tracks the content users search for in the Meta search bar. For example, if a user searches for hot tubs on Facebook, they will soon see ads for hot tubs in games published by Applovin. Additionally, they tested searching for Robinhood on Facebook, and shortly after, ads for Robinhood appeared in Applovin games.

2. Data Statistics: Applovin Reuses Former Meta Employees

Fuzzy Panda found through LinkedIn data queries that only 20 Applovin employees mentioned AI and machine learning in their resumes, of which 5 out of the 11 identifiable members of the research science team came from Meta.

3. Expert Interviews: Applovin Plays the Information Asymmetry Card

Fuzzy Panda learned through interviews with advertising technology experts that Applovin uses its bidding platform to gain detailed bidding information and the highest value customers matched by Meta ads. They then combine Meta's bidding information, device information, and data purchased from third-party data vendors (including a large amount of personal information) or encourage clients to use Adjust to restore the data on Meta's ad conversions and user matching.

They then use this information to enhance their user tags while making more advantageous bids to targeted customers.

Therefore, the evidence chain from <1-3> leads Fuzzy Panda to conclude that Applovin is stealing Meta's "data resources" and "advertising results," and if Meta were to become aware of this issue, it would respond with immediate action.

Dolphin Research Interpretation: The issue of "data theft" should mainly arise during Applovin's expansion into the non-gaming advertising market. In the gaming sector, although Applovin exhibits the aforementioned non-compliance and inflated data phenomena, at least Applovin has a relative advantage in gaming user data and does not need to steal data from Meta.

Thus, it is only in unfamiliar areas such as e-commerce, which are not gaming-related, that there would be a significant motivation to utilize Meta's user matching data to enhance their conversion rates. This is also a concern that Dolphin Research has consistently mentioned; the unified developers in the gaming sector can cycle as both advertisers and publishers, which Applovin cannot fully replicate in e-commerce.

After all, the ultimate goal of e-commerce advertisers is not DAU, but GMV (Gross Merchandise Volume). Merchants need to sell the products they have, rather than utilizing DAU like casual game users to monetize through ads. Therefore, providing a high DAU without precise users that cannot deliver satisfactory GMV to merchants would only be a burden with no value and operational costs for the merchants. Under this logic, Applovin needs to complete the first step of "user data accumulation" just like it did when entering the game advertising field. This user data accumulation needs to include shopping behavior tags behind the users (income and expenditure levels, interests, etc.), which are significantly more factors than those contained in game user data and are also more complex to analyze. Therefore, doing this while standing on the shoulders of giants will undoubtedly yield twice the result with half the effort.

Isn't this very similar to what Dolphin Research previously analyzed in depth regarding how Applovin establishes its data advantage in the gaming field? A comparison table would be more intuitive.

On the other hand, the action of Applovin leveraging information asymmetry to gain bidding advantages has been mentioned by Dolphin Research before. For specific reference, see the previous article "The Big Reveal of Applovin: A Winning Strategy Five Years in the Making." We believe that due to not having a significant advantage in the e-commerce field, Applovin may further expand its use of this information asymmetry advantage to gain favor with e-commerce users.

The "theoretical" risk here is that Meta may cut off Applovin's data acquisition channel and no longer cooperate with Applovin to access the MAX bidding network.

However, Dolphin Research believes that the probability of this risk occurring in the short to medium term may not be high:

a. Who determines data authorization?

The true ownership of the data lies with the advertisers, who have the right to know and obtain information about the display situation and user interaction after advertising on Meta (Meta generally provides advertisers with basic user tag data, device data, and full-link user behavior data from display to conversion, etc.). At the same time, advertisers will have a more comprehensive understanding of the data tags of users who are genuinely interested in the products or have made transactions. Therefore, whether data can be provided to Applovin does not depend on Meta, but on the advertisers themselves.

b. Is the relationship between Meta and Applovin solely competitive?

According to the short-selling report, Applovin requires advertisers to invest at least $600,000 per month on Meta, and in CEO Adam's response to the shorts, he mentioned that Applovin had already acquired 600 e-commerce clients last December, with an expected annual advertising budget of $1 billion. If each of these 600 e-commerce clients needs to spend $600,000 on Meta, that means Applovin has only added $1 billion in e-commerce advertising revenue, but this operation guarantees at least $4.3 billion in advertising budget for Meta from these advertisers in a year. At least in the short term, Meta has not suffered significant losses In summary, Dolphin Research believes that although Fuzzy Panda adopts the viewpoint of a certain Meta executive who said "Meta would shut it down" to argue that Meta will cut off Applovin's access to data, in practice, this is not easy (1. It cannot stop customers from voluntarily authorizing Applovin; 2. Meta currently does not feel seriously harmed).

Of course, it is hard to say from a long-term perspective. If Applovin's AXON 2.0 significantly improves in advertising tracking and accuracy due to conversion data obtained from Meta, and achieves good real purchase conversions, growing larger (for example, over $5 billion), then Meta is likely to take certain measures out of competitive threat.

Charge Third: Illegal Distribution of Ads

In this issue, short sellers mainly focus on the distribution of gambling-related games and the display of pornographic ads to children. They believe that such illegal ad distribution could lead to Applovin's SDK being kicked out of the app stores by Apple and Google, as well as facing fines from regulatory agencies.

Dolphin Research Interpretation: The issue of displaying pornographic ads to children was raised solely by Fuzzy Panda, and the evidence is relatively thin, seemingly based on a few cases they discovered. Moreover, due to Applovin's MoPub having a history of being sued for violating children's data collection in 2017, they extrapolate that Applovin still has this issue after acquiring MoPub.

We believe the impact of this issue is debatable. Currently, there are no large-scale complaints or grievances from parents, which indicates that it is essentially still a minority of cases. In ad distribution, there may be some oversights, and if there were indeed improper operations in the past, then Applovin should adjust promptly. In previous years, Google and YouTube have also faced fines for violating children's privacy.

The issues with gambling games and cash incentive ads have been mentioned in reports by Lauren Balik, Bear Cave, and Fuzzy Panda. It can only be said that there is indeed a risk; peers, especially leading platforms like Meta and Google, have stronger compliance in this regard.

Applovin has indeed distributed some gambling games, which may involve fraud, such as users being able to directly participate in PVP competitions with cash, but it is also possible that users are competing against a "bot account" with pre-set probabilities. Since this involves "real cash + fraud," it needs to be strictly regulated or directly deemed illegal. If this is pursued, such games indeed face the risk of being banned by Apple or Google. Currently, there are games with gambling litigation stains that can still be downloaded on Apple and Google, such as the games from AviaGames studio, which was sued in 2023 for conducting illegal gambling operations. However, considering that different countries have different legal constraints, and even the requirements between different state governments in the United States vary, it can only be said that such potential risks do indeed exist. Applovin has also consistently highlighted the risks that such gambling applications may encounter in its financial reports.

As for the gambling game "Cash Tornado Slots" operated by Applovin's subsidiary Zerro mentioned by Lauren, the last time Applovin referred to it was in the Q1 2024 report listed by Lauren. Starting from Q2, this gambling-like game has not appeared in the reports. From the remaining list of 10 studios, it can be seen that Applovin has quietly sold Zerro. (Applovin has been dealing with game studios acquired in earlier years over the past two years, and the proportion of revenue from its apps has been continuously shrinking).

Additionally, during the Q4 earnings release, the company announced that it would sell the remaining 10 game studios (Machine Zone, Magic Tavern, Lion Studios, PeopleFun, Belka, Athena, Clipwire, Leyi, Zenlife, and Zero Gravity) for $900 million. Dolphin Research believes that while Applovin is reducing costs and increasing efficiency, it is also shedding some potential risks.

In conclusion:

Looking back, several short-selling reports are indeed based on factual evidence, listing charges against Applovin. Some of these were already known to Dolphin Research, while others were learned for the first time this time. **We do not deny that Applovin does have elements of "wrongdoing," which from a moral and public order perspective, Applovin cannot refute **

But from a business perspective, as long as Applovin's actions are not illegal, they can be dismissed as unfair competition if they are significant, but trivialized with a simple "this is just business warfare" if they are minor. Of course, this depends on how competitors view it, especially those with direct interests like Meta.

As analyzed earlier, at least for now, from a commercial interest standpoint, clients, Meta, Apple, and Google have little motivation to impose constraints or resistance against Applovin's long-standing "crimes." This means that currently, the entire ecosystem still exists with shared interests, and Applovin's exposed behaviors may just be an unspoken rule in the advertising and app industry. However, Applovin's actions are too obvious, and with its market value increasing tenfold in a year and continuously reaching new highs, it has attracted more short sellers to dig into the details. If peers, who have lost their sense of justice, accuse Applovin, does that mean they cannot implement unspoken rules for profit in the future? Will such behavior be resisted by the industry?

Since last year's first half when it aggressively entered the e-commerce advertising market, and this quarter announced the spin-off of its game studios, Applovin seems to want to further cleanse its dark history and step into the ranks of legitimate advertising. Dolphin Research believes that the advertising fraud and other operations that Applovin has previously engaged in within the gaming market, even if Applovin makes corrections now, will not significantly affect its competitive position, because the massive accumulation of gaming user data in recent years is already sufficient for Axon to truly take off.

However, if we reflect this to e-commerce, it is easy to identify Applovin's problem—there is not enough e-commerce user behavior data, conversion data, etc., to feed into Axon 2.0. In other words, it is unclear how to achieve a high conversion rate for e-commerce advertising based on its past knowledge accumulation. Therefore, it needs to first "steal" Meta's data, just like it initially acquired a dozen game studios, to obtain different user tag data attracted by different products, as well as the success rate of facilitating users from exposure to click to order.

Thus, Dolphin Research predicts that before reaching a certain scale in the early stages, Meta will not easily intervene. As analyzed above, Meta is currently profiting rather than losing. However, if Applovin's e-commerce advertising scale is not at the billion level, but rather at the level of 5 billion annually, then Meta is likely to take action.

In the gaming advertising sector, with the expectation of slow growth in downstream gaming consumption, Applovin's management still believes it can achieve a 20% growth in gaming advertising, which means Applovin needs to further compete for market share.

Regardless of Unity's plans to launch a new model (due to unknown effects, there is no need to inject too much optimism for now), it is the issue that Dolphin Research has previously mentioned: under limited inventory, connecting more demand will raise CPM and lower ROI. Will this bring certain growth pressure to Applovin? Therefore, Dolphin Research still insists that there is no short-term worry for Q1-Q2, but there are doubts about the sustainability and scale expansion of e-commerce customer demand in the medium to long term, requiring a wait-and-see approach. Especially if Q1 continues to beat expectations, caution is needed against overly linear market expectations. It's similar to when Applovin was doing game advertising; the advantages of ROAS did not become apparent so quickly before the scale and model optimization of related data reached a certain level.

As mentioned by Fuzzy Panda in the report, some marketing technicians have begun to believe that Applovin's e-commerce conversion data is not as good as initially imagined (during the first tests), especially as the net new additions after the first batch began to decline rapidly.

Currently, Applovin has fallen to 105 billion, and according to market expectations and the confirmed 1 billion e-commerce advertising budget revenue mentioned in this CEO response, the current market value implies performance expectations for 2025 and 2026 (excluding its own APP business), with EV/Non-GAAP EBITDA at 31x and 23x respectively, gradually returning to normal valuation (but considering that U.S. stocks are adjusting, the valuation is still slightly higher than peers, but the premium rate has significantly decreased).

If one is relatively optimistic about Applovin's prospects in the e-commerce market, this short attack can be seen as a phase opportunity. Of course, this also depends on each investor's moral benchmark and subjective preferences, whether they can accept such a morally tainted but potentially promising dark horse company.

Dolphin Research "Applovin" Historical Study

Financial Report

February 13, 2025, Conference Call Minutes: Applovin (Summary): The Best Case of Technology Meeting Cross-Market Demand

February 13, 2025, Financial Report Commentary: Surging 30%! How Did Applovin Crush the Shorts?

In-depth

January 10, 2025, Initial Coverage Part Two: Following the Model, Can Unity Replicate Applovin's "Cash Ability"? On January 3, 2025, the first coverage of the article "The Ultimate Revelation of Applovin: A Winning Game Planned for Five Years"

The risk disclosure and statement of this article: Dolphin Investment Research Disclaimer and General Disclosure