📈 Cardlytics $CDLX - The Ultimate Ad Platform
Reimagining the investment case for Cardlytics
My dear fellow Hermits 👋
Welcome back to 🧙♂️ The Hermit 🧙♂️
ICYMI:
🧗 Our Investing Journey: Chai with Mohnish Pabrai
📝 Company Updates: Cardlytics Dilution
If you haven’t yet, subscribe to get access to this post, and every new post
About Investment Thesis
Investment Thesis is a series of posts dedicated to providing a foundational understanding of a listed Microcap. Additional corollary posts will add to the knowledge base of each investment case. The series's core intention is to achieve a timeless understanding of how a business works rather than reflecting on specific items related to current events (there is a separate series about these). Every month, content on an investment case will be posted in either post or podcast form, or both.
This content is intended for informational purposes only and should not be taken as investment advice. The author does not represent any third-party interest, and he may be a shareholder in the companies described in this series. Please do your own research or consult with a professional advisor before making any financial decision. You will find a full disclaimer at the end of the post.
This series is particularly time and resource-intensive, please consider…
Business Model Overview
Cardlytics has developed an advertiser platform integrated into the digital channels (web and app) of traditional banks and neobanks. This platform allows advertisers to directly address bank users with hyper-targeted cash-back offers. Cardlytics’s bank partners process +55% of all transactions in the US which feed into a real-time database. Thus, enabling action on quasi-perfect information on a substantial portion of the population.
The company serves most major US banks (Bank of America, J.P. Morgan, Wells Fargo, US Bancorp, American Express, etc.), several UK institutions (Lloyds, Monzo, etc.), and a few neobank platforms (mainly Venmo). In the past, these organizations financed cash-back offers with profits from credit card charges, but these have limitations (more on this later). This alternative presents a set of targeted offers financed by advertisers, personalized for each financial institution, and powered by Cardlytics. Offers look something like this:
Origin Story
Cardlytics was founded in 2008 by Scott Grimes and Lynne Laube, Capital One bank employees, who thought it would be interesting to implement campaigns within the banks. Their approach differed from competitors in a simple yet powerful core feature, they built the infrastructure so the data never left the banks’ servers.
Their winning innovations played out like this:
→ Advertiser information was collected
→ The campaigns and offers were created and managed outside of the firewall
→ The specs for each campaign were sent to the banks
→ Anonymized target information would be gathered within the firewall by Cardlytics
→ A summary would come out of the back end to adjust the campaigns
Cardlytics got its big break when it won a contract with BofA in 2011. At the time, everyone thought the company would be the next Google, but it turned out it was all about reach and engagement.
Reach relates to the number of Monthly Active Users (MAUs) and engagement is the activation of offers per visualization. Advertisers and users play out a never-ending chicken and egg problem. Scale and reach are required for advertisers to be interested, and users are only interested in sites that provide large quantities of high-quality offers. Advertisers also demand proof of concept without the need for a technical explanation. In other words, they want to see significant results in their numbers (more on this later).
From 2011 to 2019, the company continued to struggle as advertisers were not interested in the small user base (50-70M MAUs at the time). Cardlytics nearly went broke multiple times, and their IPO almost failed. However, in 2018 they signed JPM and Wells in quick succesion leading to an explosion in reach and, consequently, advertiser interest. The company still faces issues today, but size is not one of them.
Added Value and Parties Involved
Cardlytics proposes the following 4-sided win scenario:
For Advertisers. The tool provides best-in-class targeting and measurement. It can infinitely segment customers with great accuracy and thus address the specific needs of advertisers. Common segmentation types are recurring buyers, area of purchase (what they buy), wallet share (what competitor do they buy from), loyalty programs (who are they engaged with), spending deciles (how much do they spend and when), and offer sensitivity (how much does an offer affect their behavior).
All this information is then contrasted and tested against a holdout group that can provide evidence for their case of incremental spending. The robust methodology is a much better option than the digitally extended multi-touch attribution (more on this later).
Cardlytics offers lead to a return of 4-5 USD per dollar spent by advertisers. The comparison with other advertiser platforms is quite difficult as Return on Ad Spent (ROAS) is substantially different from incremental ROAS. Meta’s ROAS is about 2 USD, but Meta cannot directly associate sales with spending so the number will inevitably be an estimation. In contrast, if Cardlytics says it provides a $3.23 net return, it means it measured to the penny $4.23 of increment per $1 spent by the advertiser (more on this later).
For Banks. Financial institutions used to fund offers with interbank credit card fees, which limited the offers in both total offer cash-back terms and advertiser budget reach. With Cardlytics, banks receive $0.34 of every dollar spent by advertisers as profit while increasing the reach and engagement of each offer. Even though getting advertising money for free is nice, banks are more interested in customer churn. In bank terms, Cardlytics translates to modified behavior that increases customer loyalty and credit card spending. Ultimately, leading to higher client retention and increased credit card fees.
Let’s look at an example using JPM’s data:
→ The bank has about 80M MAUs. Credit card spending for the entire base was about $426 bn for Q3 2023, about $5,300 spent per customer per quarter ($21,300 per year).
→ JPM charges 2.6% to 3.5% plus a fixed fee ranging from $0.1 to $0.25 for every transaction. We’ll assume JPM closed 7 bn transactions and that these are equally distributed (average ticket $60.86).
→ If the bank loses 1% of its user base (800k users), that would be $17 bn in transactions and $450M to $610M in fees ($442M to $595M in variable fees plus $7M to $14M in fixed fees).
For context, JPM currently makes about $70M from Cardlytics.
For Consumers. Cardlytics offers a best-in-class product for consumers who have historically received coupons, points, or 1-2% cashback for their purchases. Instead, the company provides a high cash limit 10-11% average cashback offer to modestly change spending habits. Cardlytics ultimately promotes and gamifies savings aka encourages very healthy financial behavior.
From the $1 provided by advertisers, customers typically get $0.32 plus whatever banks want to add, which can be significant given the previously explained point. I’ve seen extraordinary offers for Chase Sapphire users (JPM premium) such as $500 cash back for $2,000 spent in a luxury hotel chain.
For Cardlytics. Cardlytics gets the remaining $0.35 in exchange for optimizing offers and bearing the costs related to campaigns, technology, advertiser management, etc.
In addition to the previous points, Cardlytics provides customer insights without the need to extract personal data. With its 170-180 MAUs and growing engagement, Cardlytics can provide advertisers with relevant data-driven information that should benefit all parties as it increases supply, which in turn increases demand, which in turn increases supply.... Once again, chicken and egg that can be summarized with the words of a major investor:
Cardlytics is a broken slot machine - Cliff Sosin (CAS Investment Partners)
Cardlytic’s Profitability Model Explained
Cardlytics has 3 key revenue items that every investor must understand:
i) Billings: Total amount spent by the advertisers
ii) Revenue: Total amount earned by Cardlytics and Bank Partners. OR. Total amount spent by advertisers minus consumer incentives (redeemed offers)
iii) Adjusted Contribution: Total amount earned by Cardlytics including share-based compensation (for sales and distribution)
Source: Cardlytics SEC 2023 10-K filing
We can observe the trend of Cardlytics increasing its share of the pie as presented in the retained percentage.
Consumers - 31.80% (2023); 32.53% (2022); 32.21% (2021)
Bank Partners - 33.22%; 35.14%; 34.90%
Cardlytics - 34.98%; 32.33%; 32.89%
Cardlytics heavily relies on scale and operational leverage to run optimally. In other words, the company has high fixed costs related to software development and customization for each bank partner. This item is both the main barrier of entry for competitors and the reason behind the extreme valuations associated with the company. The increase in advertiser count, advertiser budgets, user count, and user activity that drive Cardlytic’s revenue do not lead to substantial variable costs. Simply put, as the company scales its unit economics become extraordinarily good.
M&A and Subsidiaries
Previous management turned to M&A in 2021 and 2022 as it felt the market’s pressure for growth at all costs. During this period the company acquired three competitors intending to purchase tech stack and reach. In retrospect, Cardlytics paid too much for these. I’ll describe them in order of relevance.
Bridg was acquired in May 2021 for $350M in cash and stock. The company is a PoS data analytics software that provides information at an SKU level (Stock Keeping Unit). The company integrates data from multiple sources to aggregate consumer profiles into a single anonymous identity that can later be targeted with hyper-personalized campaigns. Bridg mainly serves retailers and is partnered with Snowflake, theTradeDesk, LiveRamp, and ThoughtSpot.
Source: Cardlytics SEC 2023 10-K filing
Bridg’s contribution to Cardlytics numbers may not appear substantial, but it creates a close loop with the end user creating robust offer targeting. Additionally, it enables SKU-level campaigns (more on this later).
In 2022 and 2023, investors went through a rollercoaster of emotions due to the popular belief that the earnout payment for this company would bankrupt Cardlytics. For a more detailed breakdown of what happened please check Bridg earnout resolution.
Dosh was purchased in March 2021 for $275M in cash and stock. Dosh operates identically to Cardlytics (in practice) but focuses on partnering with neobanks instead of traditional banks. Dosh operates with a more modern tech stack that enables local offers and an easier API integration with modern banking alternatives. Dosh’s key account is Paypal’s Venmo.
Entertainment was acquired in March 2022 for $15M in cash and stock. The company focuses on local SME restaurant discounts through vouchers and coupons. The purchase came as a way to push for geography-based discounts and the development of a self-service platform. With new management returning to the basics, Entertainment was deemed a distraction and consequently sold in December 2023 for $6M in cash and an additional $1M in earnout.
Lastly, Cardlytics developed a data and media network under the brand Rippl that has been steadily growing since its launch in August 2023. The company hosts 70M MAUs comprised mainly of regional retailer users and has insight into $40 bn in spending from +21,000 locations. The initiative aims to create a 360º understanding of customers’ profiles and help retailers monetize users through predictive analytics.
Multi-touch attribution (MTA) vs Randomized Controlled Trials (RCT)
In general, people don’t trust measurement in digital media. In particular, competitors have an inferior quality of measurement commonly known as MTA. Any company using this measurement methodology will result in better-stated returns and lower actual returns. With Cardlytics you get the precisely stated returns. MTA commonly uses one of the following models:
Full path model. Examines all participants’ contributions and rewards accordingly
Linear attribution. Rewards all participants equally
Time decay based. Rewards all participants weighing contributions based on timestamps
W-shaped and U-shaped models. Rewards all participants with the first and last contributors getting a larger share
Custom models. A combination of all of the above
Source: newbreedrevenue “The Pros & Cons of 4 Single & Multi-Touch Attribution Models”
The most common model, the Full Path methodology, will split the earnings proportionately but does not do so based on causation. Instead, the model rewards elements that may have no connection to the product's sale.
Brand and word of mouth are two chief examples that come to mind. How much of a sale can or should be attributed to the repeated exercise of watching ads, observing promotional campaigns, and talking about the product with other users throughout the years? Should we attribute a portion of today’s iPhone sales to the iconic 1984 Apple commercial? Should we pay the team (surely retired) Coke team responsible for turning Santa from green to red? I don’t personally think we should. However, the MTA model, which in theory should attribute a portion of the merit, is, in practice, not paying these people and thus not acknowledging their contribution.
Source: Pete Judo’s Youtube video
Ultimately, MTA does not pay the parties for causation, but instead pays them for action. For this reason, measuring the yield of a campaign is quite hard since you don’t know why the person bought the product or service.
Nevertheless, MTA models are considered the industry standard because of their ability to showcase a message to millions of people in an instant. Their ease of use makes targeting a “non-essential feature”.
Contrary to popular digital marketing skeptics, Cardlytics does deliver on its promise. And it does so by implementing campaigns based on targeting a segmented portion of the credit card-using population.
Cardlytics conducts measurements of purchase behavior on two identical (or similar) groups, one is targeted with the ads and the other is not. By measuring the difference between the amount redeemed by the first group and the second, Cardlytics gets the incremental figures. These represent a direct behavioral change of consumers caused by the ad campaign aka attributable returns.
This testing cannot be shown directly to advertisers as data is confidential, but it is retested and published by Nielsen, a third-party independent auditor (more on this later).