An Ecommerce Marketer's Perspective on the Muddy Waters AppLovin Short Report

Muddy Waters, Fuzzy Panda, Culper Research: ALL Short $APP. Do they have a case?

🔎 Examining the Details of the Muddy Waters AppLovin Allegations/Claims

Illustrating the rise of AppLovin & the attempt by short-sellers to drag it down. Created with ChatGPT.

Introduction

Last week (March 27, 2025), Muddy Waters Research published a scathing short report on AppLovin; stock ticker $APP ( â–˛ 7.23% ) .

They titled it: “Deep Data Analysis Shows APP is Just Another Scammy AdTech Company.”

In the past month, that makes it 3 short reports:

  1. Fuzzy Panda: “APP drives fraudulent conversions”

  2. Culper Research: “APP steals data from Meta to drive sales”

  3. Muddy Waters: “Performance is SO GOOD it's impossible”

There were a lot of claims in there, some I can speak to and provide more color on than others. I will review each point below with my perspective.

Disclaimer: I am NOT a paid promoter of AppLovin. I am not paid by them or anyone with an interest in them to speak on their behalf. I own an immaterial amount of their stock (< 1% of my portfolio) which was purchased by me due to my belief in their ad platform & upside potential. The views expressed in this blog are my own and do not reflect the views of my employer or any company with which I’m affiliated. The information provided below does not constitute financial advice and is for informational purposes only.

The Structure of Muddy Water’s Review

Muddy Waters actually brings data to the table, which the other reports lacked. In the report, they reviewed:

  • 5 Shopify Brands that advertised on AppLovin

    • apparel, beauty, hobby, healthcare, & other verticals

  • 7 week span: 12/30/24 to 2/16/25

    • these 5 brands did $300M in revenue over this time frame

    • If we extrapolate this, we know the average size of the brands comes out to ~$446M in annual GMV. These are large brands.

  • 37 million unique users in this time frame reviewed

    • If we extrapolate this, $300M/37M = $8.11 revenue per user

Let’s start here - - - The brands tracked in this study are LARGE brands (averaging 0.5 BILLION in annual revenue). Given that the average revenue per user is $8+ (standard deviations above the industry norm), this already indicates that these brands rely heavily on returning customers. This sets the stage for some of the claims made.

CLAIM I.

âťť

Only ~3.4% of users had any clicks from APP ads as indicated by URL level tracking by these advertisers

My Take: In order to have any merit on a claim based on % of users on site from a particular source/medium, you would need to know how much the brand spent on ads during that time frame, what % of their budget it was, and what % of their site traffic came from paid channels. Without that, a 3.4% figure has no context. Also, for brands of this size (avg of $446M annual GMV each), you would never expect a high % of users to come from a new paid channel. Especially when the brands average $8+ revenue per user, indicating a large percent of traffic being returning customers or subscription re-bills.

CLAIM II.

âťť

For ~52% of the users who has a purchase event on the last click attribution, APP was not the first click to the website

My Take: Regarding the retargeting claim, MW states that 52% of users who had APP as the last click purchase were not also the first click to the website. First, we all know by know that APP’s biggest limitation is their lack of audience segments and thus their inability to exclude previous website visitors or existing customers from the campaigns. This is an issue with user identity on their platform at the moment (which I believe can be resolved over time as more PII is stored from ecom ads).

HOWEVER- given the size of these brands that were referenced, they most likely already have tens of millions of customers lifetime and likely are not solely focused on new customer acquisition. They have likely saturated a significant part of their TAM already. Now add in the fact that you can not exclude previous website visitors or existing customers from your APP campaigns. In reality, the fact that 52% of APP last click conversions had APP as the first click in the user journey AND the last click in the user journey is actually pretty damn good.

Also- let’s clear the air here. You can retarget any website visitor on any ad platform (even if that visitor never came to the site from that ad platform). So, there is nothing wrong with retargeting Meta-driven users on AppLovin’s placements. Brands do this across all platforms already. The only downside is that there’s no way to control targeting on AppLovin, so if new customer acquisition did decline on AppLovin, there would be no way to control for it, and thus would limit spend potential on the platform.

And- keep in mind these are CLICK-through conversions. This is much different than view-through conversions which have far less value (arguably closer to 0 than 100% impact). Brands of this size are more concerned with driving incremental conversions and less concerned with which audience segments the conversions came from.

CLAIM III.

âťť

Sales Are Only ~25-35% Incremental

My Take: So, lets discuss the “incrementality” reference as it’s more important. Muddy Waters claims that AppLovin only drove 25-35% incremental conversions. But their definition and calculation methodology are objectively wrong. They basically assumed that incremental conversions could only come from conversions where APP was the first click (WRONG- you can have incremental conversions come from existing customers, retargeted users, etc), so they took 48% (100-52%) that had APP as the 1st click and they then excluded 29% of these users who also had other platforms in the click journey of the user, leaving them with 34% “incrementality.” Again, you could have an incremental conversion that had other clicks in the user journey. They then calculated the low bound of incrementality by referencing Shopify’s benchmark of 28% repeat customer rate and excluded an industry avg repeat customer rate from the leftover “incremental” conversions, which gave them the 25% low bound. Once again, you could drive incremental conversions for repeat customers through advertising.

Also, keep in mind AppLovin only attributes CLICK-THROUGH conversions. This is far different than view-through conversions where retargeting can absolutely fluff numbers. If someone clicks through an ad and purchases a product from that brand’s website within a tight window, such as 1 or 7 days, there is a better than average chance that ad was impactful in driving the conversion, aka it was likely incremental.

The definition of incrementality in performance marketing is this: did the ad cause that person to purchase from the brand? OR were they going to purchase in that time frame anyways? The only way to quantify incrementality is through structured testing like a holdout study.

CLAIM IV.

âťť

~22.6% of APP’s E-commerce Customers Appear to Have Churned in Q1 2025

My Take: On January 3, Muddy Waters scanned websites and found 776 websites with the AppLovin Axon pixel installed. In March, 171 no longer contained the pixel. There are likely also cases in which advertisers paused spend and did not remove the pixel, so in reality there could be more.

However, something to keep in mind is that during this window of time, AppLovin actually took down MANY lead gen advertisers who had low quality funnels. AppLovin is mindful of user experience and does not want user experience to take a hit from their mediated ads. Some of this churn can easily be attributed to this issue.

There are definitely instances where the platform did not work for some brands and those brands churned. That’s normal & expected. Who’s to say those brands won’t be back? WAY too early to track churn & draw conclusions from it.

The real data to look at would be what % of advertisers started spending their own ad dollars after the ad credit ran out? I don’t have this data, but something that would be noteworthy to look into.

I have a few anecdotes from advertisers who paused spend there because they simply didn’t have the internal bandwidth to devote to another channel at the time and give it the attention it needs to create alpha and true impact on the business. Keep in mind, in order for something to become a fixture in any brand’s media mix, it must have tangible impact on the business. This is done through scale. Brands require scale to drive impact (especially if they’re doing at least $10m in annual GMV- which this closed beta requires). If a brand doesn’t have the internal resources to unlock scale, it gets cut.

CLAIM V.

âťť

APP Systemically Violates TOS by Creating Persistent Identity Graphs (PIGs)

My Take: This is actually the most serious allegation there is, and the most difficult to prove without forensic investigation. Let me preface this by saying that I am by no means a conversion tracking/data expert. There is an entire industry dedicated to this within ecommerce.

However, I have a solid enough understanding to be clear that ecommerce brands all have different data practices with their own first party data. There are an endless amount of Shopify plugins that will append data and send them to platforms, per the brand’s direction. This is their own first party data. They have the right to leverage their data to sharpen platform algorithms in their optimization practices. This has nothing to do with ATT. Server sided tracking is actually what was built as the solution to ATT & that firestorm that ensued.

If a brand has a plugin like Elevar, and Elevar sends data to AppLovin, it’s actually going to the other platforms as well, as the brand intended. Brands choose whom to send data to and which plugins to use. AppLovin’s CTO highlighted this example in his recent blog article that examined data practices for ecommerce brands.

If a brand does not have such a plugin that appends additional information (as in the Crocs example the CTO highlighted), the Facebook ID was NOT sent to AppLovin. There is visual evidence/proof of this in the blog article linked above.

AppLovin’s CEO also put out a blog article in response to Muddy Water’s piece, which clearly stated that when information is sent to App’s servers that it does not ask for, it’s purged when it hits their servers.

I spoke with CEO of Source Medium (ecommerce data infrastructure company), Fei Wang, about this situation and he summed it up well:

For marketers, the key question isn't just whether data is collected, but whether it's actually stored and used. Independent verification through cross-client comparisons and controlled testing environments would be necessary to conclusively settle this technical dispute.

- Fei Wang, CEO of Source Medium

As such, AppLovin is likely only using data it is rightfully in possession of, it still has not been proven otherwise. Innocent until proven guilty, right?

CLAIM VI.

âťť

APP’s PIGs Violate Apple’s TOS and Designed to Avoid Detection

My Take: In 2024, Google prevented 1.3 million apps from getting excessive, unnecessary access to sensitive user data. In 2022 and 2023, Apple terminated over 500,000 developer accounts. If AppLovin was egregiously violating Google & Apple App Stores TOS as well as Meta’s Audience Network TOS, don’t you think it would have been detected by now?

Droves of dev accounts have been taken down over the past few years. Wouldn’t a company with $100B market cap be the first to get detected and eradicated? Especially on Meta, who some perceive to now be a competitor. This company didn’t come out of nowhere, they’re on an $11B annual run rate on mobile game and ecommerce ad revenue now.

Common sense in me tells me that they have not flown under the radar, especially now in light of all the attention they’ve gotten.

With regard to Muddy’s claims that App embarked on a token-changing journey across environments, Fei had the following thoughts:

Muddy Waters’s description of token changing across environments raises legitimate questions, but proving intentional circumvention versus standard technical implementation requires specialized forensic analysis. When evaluating these claims, we must distinguish between standard device signals used within privacy frameworks and true fingerprinting that creates persistent cross-app identifiers.

- Fei Wang, CEO of Source Medium

This topic is out of my purview for the most part, but common sense has me thinking that AppLovin is not engaged in any malicious cover-up scheme which would be seriously shocking and disheartening.

CLAIM VII.

âťť

APP’s Unauthorized Use of PIG Data Gives It A Black Edge in Ad Auctions

My Take: There’s a reason why AppLovin has 4x’ed ad revenue in the last 2 years since launching Axon 2 (their new algo). Because it actually works. Their AI/ML systems have outperformed long before ecommerce ads were even a thing for them.

At this point, I’m under the assumption that AppLovin is operating with integrity & in accordance with privacy policies. They aren’t maliciously extracting, storing, and using data that it does not have legal rights to.

Conclusion.

Anytime a short report comes out that can move a market, there’s typically inherent bias. These reports absolutely can move a market and these short sellers are often out of the trade shortly after. So these reports are intended to be the catalyst to make their self-fulfilling prophecy come true: to drive the stock down and make quick money.

That isn’t to say there isn’t merit to any of these claims potentially turning out to be true. Can some of their PIG/fingerprinting claims end up being true? Sure. But, the jury is still out and time will tell. Now that the company has gotten so much attention, I assure you Apple, Google, and Meta are all investigating very closely. If anything malicious turns up, it will be swiftly identified and made publicly known.

From an ecommerce marketer’s perspective, many of these claims are grasping at straws and are very difficult for a layperson to analyze if they do not possess inherent knowledge about digital marketing and conversion tracking.

It’s pretty clear to me that from an ecommerce perspective, the retargeting & incrementality claims lack the merit and context required for Muddy Waters to prove their point.

To this very point, my POV is that AppLovin’s success is mostly attributed to its non-skippable ad format, its strong Axon 2 algorithm, its scale potential, and the novelty factor it has with its audience. We still haven’t scaled a brand yet to anywhere near the level of Meta, but the amount of spend we’ve been able to achieve there this quickly makes it highly intriguing.

I welcome all feedback. Good, bad, everything in between.

Hit reply, and let’s hear it! 👂

📧 Share your thoughts or what you want me to cover next!

Yours truly,

Jonathan Snow

If you’d like to see more content of mine, be sure to give me a follow: 

Twitter • LinkedIn • Instagram • TikTok • Medium

📆 Schedule 1-on-1 time w/ me here.

Reply

or to participate.