$CDLX: How It's Made

Also a bit of an update

Intro:

Little bit later on this than I was aiming for. Currently working on something that will probably start seeing some eyes in Q1 with a full release by Q2/Q3 of next year. It’s along the lines of alternative data/finance education type of stuff, so hopefully it’s interesting to a variety of followers. Will try to get out some more articles that are more “general” in nature over the coming weeks as I’m not a huge holidays guy so get some free time.

Perhaps controversial but in general I’m not super excited about any unowned stocks right now. There are things to kick around but just generally it’s going to be a lot harder to find deals this year compared to a year ago. I’m not saying I won’t come across something interesting, but write ups in the near term will likely not be specific ideas for that reason.

CDLX: Brief Update

It’s been awhile so I figured I’d toss in an update on CDLX. The majority of this article will be more “how it works” not “how it’s doing”, so a brief catch up might be helpful.

As usual visibility into CDLX’s current quarter revenue trajectory is basically impossible. Nobody can say for certain if they’re above or below guide. Below guide would likely cause some problems given expectations and the need for growth, but on the flip side above guide might be a nice surprise. The stock will likely be quite volatile as we get into early 2024 with the investor day and Q4 earnings. Do with that what you will.

In terms of actual business results, the clear message from Q3 was that growth was not happening as quickly as one would wish given the rollout of Chase on the new UI. Recent expert calls and personal sources agree that while the bank tech and bank product has improved considerably, the advertiser facing tech and product has not changed much. This of course has two implications:

1. Growth will likely take longer than expected, as the advertiser facing product likely needs some updates for less sophisticated advertisers

2. If many advertisers are unaware of new product improvements, that is a potential future growth vector that is not yet tapped.

So one positive, one negative. The overarching story however is of course the debt due in late 2025. If growth takes longer than expected. liquidity may start to be a constraint.

By no means is this a new development, this is the same concern that has plagued CDLX for all of 2023. The hope was that Q3 would show outsized growth from the new ad server and put this concern to rest, hence the significant stock decline when that didn’t happen.

Below I’ll get into some more details about why I think CDLX can grow, however it is by no means guaranteed. Do with that what you will.

The Mechanics of CDLX:

The way CDLX fundamentally works can be a bit confusing. I’ll go over this both from an advertiser perspective, a bank perspective, a user perspective, and a CDLX perspective.

Advertisers:

For advertisers the CDLX offering is a touch unique. Traditionally if advertisers use an ad platform, they can either use a dashboard for that platform, or pipe in metrics via an API to their own models. These “models” are basically just a bunch of math and data trying to match purchase data to ad spending. Say you do $1m of Facebook ads and get 100,000 clicks, how many of those converted? Who were they? Did they see other ads of yours? etc. The goal is to try and attribute advertising spend to consumer purchases so you know how effective your ads are. The systems are complicated but the goal is pretty simple.

CDLX is a bit different. Most advertising platforms are using their own data or public data. Meta for example is simply using their own user data. CDLX uses bank data, which makes using data for the purpose of advertiser verification a bit difficult. CDLX does not have a clean way to integrate into a standard advertiser model, as for the most part they are not able to share data with advertisers, instead relying on solely their own dashboard. This dashboard is also below industry standard, as it is delayed 1-2 weeks as CDLX verifies transaction data. Meta or Google dashboards are practically updated in real time. The consequence of this is actually quite significant.

As an example say you want to spend $1m via Facebook ads. You set your bids and all and you can immediately get feed back on impressions and click through rates. If you’d like to spend more and get more traffic you can adjust your campaign. If you were too aggressive, you can scale back targeting immediately. With CDLX the way it works is you essentially submit a campaign request with parameters to CDLX. CDLX will then *data science magic* some targeting parameters and create the campaign. This is a rather manual process compared to many other platforms where advertisers can simply do everything themselves, hence the CDLX focus on self-serve in 2020/2021 as a product focus.

Those following for awhile obviously will know that self-serve platform was essentially scrapped, which would be confusing given it lowers friction and would make CDLX more industry standard. The logic however makes quite a bit of sense, as you can think of building self-serve as building the steering wheel before you build the car engine. The real failing with the CDLX advertiser process was with their dashboard. Historically this took 1-2+ weeks for advertisers to receive any sort of metrics on their campaign. If your campaign was being run too aggressively, you wouldn’t know until your budget was almost gone, which creates an awkward conversation. Similarly, if the campaign was going slowly, CDLX would need to have another conversation with the advertisers and create a new targeting profile. This profile creation historically was very manual as well and could take >1 week whereas a Meta or Google campaign could be created almost immediately.

As a result CDLX has focused on bettering that platform for advertisers, now getting campaign launches to a few days and automating targeting algorithm creation. This sort of work isn’t visible to consumers at all, but plays a big role in how usable the platform is for advertisers. CDLX will always sort of suffer a structural disadvantage given they are working with non-primary data, but some of those issues can be engineered around.

An astute researcher would also notice that CDLX has largely focused on bank facing tech this year as opposed to advertiser facing tech. This could be puzzling given recent Tegus calls highlighting how some advertisers don’t trust attribution from CDLX. In reality, what CDLX is doing is actually quite rational. Banks have discretion over their data sharing, but generally operate on a sort of tit-for-tat ideology. If CDLX is able to significantly improve the bank facing technology, maybe banks will be more willing to share data with advertisers to help ease attribution concerns. This has actually happened at Chase, who will share transaction data with advertisers in select cases. I don’t believe WFC or BofA are doing this currently, which supports the idea that a closer relationship with banks via product launches enables better advertiser service (or Chase is unique, which would be unfortunate).

This data sharing to advertisers can go even further to outright consumer intelligence reports beyond the scope of the ad campaign, which is quite valuable. Say McDonald’s runs a campaign they could be told wallet share in users redeeming ads for months after the campaign to gauge stickiness and other such use cases. I believe this angle is why CDLX has prioritized working on the bank side of things first, as they need to set up the base to build upon a more compelling advertiser product.

That isn’t to say of course that the advertiser facing product is unusable, simply that it requires more effort than other channels to get comfortable with. Given that CDLX is a rather niche platform it’s hard to “feel” the uplift from your CDLX ads, so when CDLX simply sends you back a report “Congrats on a 5x ROAS! Come again!” it can be met with suspicion. I doubt this is actually an issue given sophisticated users such as Walmart and Panera have tested the product extensively and significantly ramped spend, but that level of effort and sophistication is not available at all advertisers. Like in theory if CDLX says “50% of users with an offer converted, 30% without an offer converted” that’s a pretty easy way to say ok we went from 30→50% conversion. In reality because it doesn’t cleanly plug into a model, some advertisers will not believe the results or not put the effort into understanding the results given the non-standard reporting model.

All in all, CDLX certainly faces some headwinds in regards to landing advertisers given the niche product and non-standard reporting, however the product clearly seems like a benefit to the advertisers that put significant effort into working with it. The limiter seems to be educational and technological, not structural. If there was no liquidity constraint, it would be easy to give them time to execute. Unfortunately they’re on a bit of a clock, but progress has been made over the past few years and once WFC and BofA are on the new UI, progress should accelerate with a consolidated GTM from the side of CDLX.

All of this is in theory of course. We’ll have to see how reality develops with the CDLX push into a better advertiser offering. In theory many advertisers can likely spend multiples of what they are now at good returns, it’s just hard to get them there without good data and trust building. A game of inches, not sudden leaps.

Bank Perspective:

CDLX and banks is probably the most opaque relationship given bank desire to not have it public that they’re monetizing users transaction data! At the same time, banks are leaning more and more into using consumer data, and will likely continue to do so for the foreseeable future as standard bank activities such as mortgage origination become increasingly less attractive (2023 wasn’t their greatest year). CDLX ends up being a great way for banks to monetize user data while also providing even more valuable card benefits. If you ever get a credit card ad for example it’s likely along the lines of APR, % cash back, random perks, etc. CDLX is another facet of that offering and attractive offers like % cash back at Walmart or Starbucks can be decent drivers of consumer spending while building loyalty on that card.

There are of course a couple of caveats.

The first - to what extent does uniqueness matter to banks? This a rather common pushback I see on why banks would not be excited about CDLX. It does make sense right, if you have cardholder benefits why give out your data so that other banks can have good cardholder benefits? This should be sort of self evidently not an issue given numerous banks participate already, but we can also reason to their fundamental logic. As always with something like this, a key factor to consider is what happens if parties defect? If WFC/BofA/Chase were to leave CDLX, they would lose some of the scale that makes the platform compelling to advertisers. They would also need to rebuild these capabilities in house including the tech and sales teams, which is not an insignificant effort. I’d bet everything I own and then some that conversations along the lines of “lets try what AmEx does” have happened dozens of times throughout the years in bank meeting rooms.

Perhaps an easy to see reason why that fails is how hard it is to get banks to do simple tech updates with CDLX such as turning on a new UI being a multi-year ordeal. Additonally CDLX actually can enable some form of bank unique value. Banks can create their own offers and target specific card holders. Say you have a client looking to buy a house, maybe send them offers for Wayfair so they think about you as they idealize their dream house. CDLX can facilitate the offer supply both from advertisers AND banks, which Chase has used off and on historiclly. This is also why CDLX does not brand their advertising platform, and with the new UI will enable advertisers to have a touch more control over presentation so it feels more integrated into the platform. This can be as simple as using Chase colors in Chase, WFC colors in WFC, but also extend to offer category pushing - say Chase wants to provide more travel benefits and grow Chase booking, you can have a travel category and point users to it as part of a marketing campaign. The uniqueness concern doesn’t seem to hold water as a real limiter to growth.

Another concern raised with banks is how much the platform actually matters to them. This can be a bit of a tricky question given we don’t have good insight into what’s happening behind closed doors, but it’s a good question. At Chase we can see what appears to be a rather concentrated push into consumer intelligence and a more data enabled business model. Dimon has mentioned on numerous occasions investing more into consumer offers and shopping support, of which CDLX is a rather key part. CDLX for Chase will soon be doing around $100m of earnings, which likely correlates to $1-2b of consumer spending. While that isn’t a vast amount compared to the volume Chase sees, it is not insignificant either and ideally grows to multiples of the current value. At that scale CDLX is by no means a key pillar of Chase, but it would be a quite profitable side project spitting off hundreds of millions or billions in earnings while providing cardholder benefits to millions of cardholders. The requisite effort from their side ends up being quite low after the recent tech transition, so I see very little reason why they would not continue this program. This seems evident given Chase giving a larger partner share to CDLX recently.

On that tech note, it begs the question why WFC and BofA are not yet on the new ad UI. To briefly touch on how it works, CDLX effectively receives transaction data from banks, runs an algorithm against it to determine who gets ads and who used ads, then sends that information back to the bank. Naturally the AWS transition which has occurred at most banks by this point makes that whole sequence much easier to execute. Prior to AWS the whole sequence of turning transaction data into something usable and sending back targeting was much harder to update. CDLX would basically have to kindly ask the bank to let them push an algorithm update under the old system. Now that we’ve largely gotten to full AWS deployment, we started seeing CDLX announcing updates to their algorithms and the like pushing redemption rates higher. Part of these updates however will rely on the banks.

To my best understanding, the way CDLX works with banks on a base level is that the bank sends CDLX a user ID, CDLX sends back a pool of offers for that id, and the bank then displays said offers. Because CDLX is not actually displaying the offers, anything to do with offer displays will rely on the bank to implement, whereas better decision making around what offers to give is in the hands of CDLX. For something like the updated ADE that CDLX has been talking about, it’s as simple as (I believe) working with the bank to change the way in which offers are sorted upon presentation. This will require the bank to essentially plug in a new API call to their app and website, which while not a ton of work, can be rather time consuming given all of the conservatism around bank software development. As such we see Chase and WFC on the new ADE, but not yet BofA. The new UI is even more complex and requires the bank to set up new API calls for images and the like. CDLX has already largely done all the work to build what the banks need to connect to, but now has to convince the banks to enable a bunch of new core features.

Chase naturally is done with the transition. WFC I believe is on the new ADE, implying they are on the new ad server. BofA I believe is not on the new ad server with all systems in AWS, thus they are not on the ADE and not on the new UI. The hardest part is likely the transition to systems in AWS. Thus I’d expect WFC on the new UI sometime relatively soon, but BofA may see more delays on the mid 2024 target if we don’t get an announcement of an AWS migration in the near future, as they need to both migrate to AWS an enable all of the new API functionality. In theory this could be done rather quickly, but Chase did not operate that way and WFC seems to be taking multiple months as well.

Again it seems clear that WFC and Chase take CDLX relatively seriously given they have committed engineering resources to bettering the platform over the previous 6-12 months. BofA is very questionable given the lack of updates, which is odd considering it’s one of CDLX’s longest standing bank partners. For some evidence you can take a look at historical disclosures of partner share concentration.

During the six months ended June 30, 2021, our top two FI partners combined to account for over 75% of the total Partner Share we paid to all partners, with each representing over 30%. During the six months ended June 30, 2022, our top three FI partners combined to account for over 75% of the total Partner Share we paid to all partners, with the top two FI partners each representing over 20% and third largest FI partner representing over 10% of Partner Share. No other partner accounted for over 10% of Partner Share during these periods.

Q2 2022 10-Q

During the nine months ended September 30, 2022 and the nine months ended September 30, 2023, respectively, our top three FI partners combined to account for over 85% of the total Partner Share we paid to all partners, with the top FI partner representing over 50% and the second and third largest FI partners representing over 10% of Partner Share. No other partner accounted for over 10% of Partner Share during these periods.

Q3 2022 10-Q

In this case the #1 bank is Chase, #2 is BofA, #3 is WFC.

Trajectory for each as follows:

Chase: >30% in 2021, >20% in 2022, >50% in 2023

BofA: >30% in 2021, >20% in 2022, >10% in 2023

WFC: <10% in 2021, >10% in 2022, >10% in 2023

Clearly Chase and WFC are growing. If we call it ~40% Chase in 2021 and ~8% WFC in 2021 vs ~55% Chase in 2023 and ~12% WFC in 2023, that implies the following:

Chase: ~$37m first nine months 2021 vs ~$60m now, 61% growth over two years

WFC: ~$7.5m first nine months 2021 vs ~$13m now, 75% growth over two years

Now assume BofA went from ~35% to ~23%

BofA: ~$32.5m first nine months 2021 vs ~$25m now, 23% shrinkage over two years

Total Partner Share: ~$93m first nine months 2021 vs ~$109m now, 17% growth over two years.

Obviously these numbers are not exact, however I’d imagine the directionality is quite accurate. The clear implication is that BofA is drastically lagging in partner share and also not investing as much into the CDLX tech. If BofA was entirely removed partner share would have grown ~40% over two years. If we adjusted for the loss of Santander it’d be like ~50% over two years. It’s rather interesting right, as we have BofA and Santander clearly not prioritizing the program, then WFC and Chase prioritizing the program and growing 20-30% per year through poor macro conditions as they do so.

Unfortunately it’s very hard to say WHY Santander and BofA are not prioritizing CDLX. Santander replaced CDLX with a new solution that just seems sort of strictly worse, so it’s not like they didn’t like the idea. BofA was trying out Figg for awhile in 2021, so again it seems like they weren’t against the idea. What seems likely is that CDLX bank tech solutions and partner service was severely underprioritized by previous management which led to platform disinterest from a couple major partners. Santander has of course exited, but BofA is supposedly going to be on the new ad server and UI during H1 2024. If this timeline is true, then in theory that could be a significant boon to CDLX. If BofA went from shrinking 10% per year to growing 20-30% per year that would be a >10% swing in the CDLX topline growth rate just from BofA combined with WFC getting on the new UI and allowing CDLX to have a more consolidated GTM strategy. Of course BofA could also simply continue to not invest in the platform and deteriorate further, hard to say with any certainty.

Beyond growth rate implications BofA creates an awkward situation for CDLX. As an example of how the offering works, when adverrtisers give a campaign to CDLX they set a budget that basically determines how many redemptions are allowed. At the same time, CDLX does not want offers to disappear from users accounts mid campaign as that would create a poor user experience. This creates a somewhat complicated game where CDLX needs to say show 10m people an offer to get 500k to redeem it (numbers irrelevant). The issue arises with the fact that CDLX also needs to spread these offers around the different banks.

This creates a bit of a conundrum. If BofA is not investing in the platform, what share of ads do you show them? If ROAS is higher on Chase due to product improvements than on BofA, what % of ads do you show to BofA? In terms of near term financial results that number could actually be 0, however you still need to try and be fair and appease BofA. I’d imagine CDLX monetization is somewhat artificially depressed by having to serve offers to BofA at worse monetization rates. While I don’t have direct evidence of this, it seems intuitively true.

As an example, say a user opens their bank app to redeem an offer. If we believe that the new images increase conversion and we believe that better offer sorting (to show more relevant offers first) improves conversion, then we would see higher activation rates per user at Chase. Like say a user at Chase converts on 10 offers per year and a BofA user converts on 7. Then CDLX is being forced to throw offer volume at BofA that is less likely to convert to maintain both offer density and the BofA relationship. Thus CDLX could have better monetization and a much better advertiser product if BofA simply invested in the platform with higher possible ROAS in addition to the consolidated GTM strategy and tech stack.

In sum, BofA getting on the new ad server and UI is a very big milestone in H1 2024 with not a whole lot of reliable signal. This is probably the #1 bear case as in theory it should happen and has been said to happen, but BofA has sort of inexplicably lagged since 2021. If the explanation is simply they were unhappy with previous management execution of bank tech, so we get new management with a focus on bank tech that is supposedly going well, then in theory BofA should resolve positively. If that is not the case, then we may face significant growth headwinds which could spell investment failure given 2025 debt maturities.

Not pretty, but I think it works out.

User Perspective:

These last two sections will be brief, but hopefully provide some helpful context.

For users CDLX is relatively simple, you go into a bank app and redeem offers. I do think opinions on the user value proposition are a bit distorted by the fact that asset managers typically aren’t a discount focused audience (perhaps ironically?). Compelling offers are stuff like Walmart and Turbotax and lower end Marriott hotel rooms in off season. Not typical NYC 6-9 figure comp consumption areas.

What I do think is interesting is sort of how does CDLX drive user engagement? Like at the core the way CDLX works is somewhat non-standard. They will have a monetizable user base of engagement that I presume is steadily increasing over time, however relative monetization of that userbase can vary. As a sort of simple visual example:

Like say we get a Q4 2022 (or Q4 2023 apparently) where advertisers are hesitant on spend, revenue may not progress meaningfully in that quarter, however the key metric for long term growth is user engagement. This is actually a far more important long term KPI than revenue as the opposite can occur as well. If in 2021 user monetization is obscenely high, following year growth will look weaker despite potential solid business execution.

Unfortunately CDLX does not currently share any relevant metrics on longer term user engagement trends. My theory is that the new ad server as well as bank effort is a rather significant boon which seems to be evidenced by the growth rates at Chase and WFC. An old Tegus call mentioned Lloyd’s sending push notifications for offers causing redemption rates to grow >100% MoM (I believe Lloyd’s is the highest ARPU bank currently?) My own calls have indicated that users are actually rather sticky once they are aware of the platform, however many users simply aren’t aware. Part of what makes users sticky is the plethora of QSR and common day shopping offers which have been sort of the core of CDLX such as Panera, Starbucks, Dunkin, Walmart, etc. Once users reach offer density they will intuitively be more liekly to continue to engage on the less core offers.

This concept is quite clear if you create a fresh bank account. The offers to start will be very general larger name offers as they typically will have the widest targeting net and higher budgets. It’s harder to entice users naturally with some smaller budget niche advertiser if they aren’t familiar with the platform.

Ideally we end up getting an investor day in Q1 that starts to highlight some of the core user engagement metrics. They clearly exist as we get drip fed some each quarter, but aren’t shared en masse due to concerns about “cherry picking data” from the new ad server as well as banks being a bit stingy with sharing metrics. I would strongly encourage anyone currently involved to ask for better disclosures on user analytics as it will get increasingly relevant once we are on the new ad platform to track underlying business progress.

The CDLX Perspective:

As mentioned earlier, CDLX is walking a bit of a fine line of sometimes conflicting incentives. What may monetize the best may not be the best for the business, and significant effort may be required to bring underperforming banks up to par (why they “fired” Citi). Given I largely invest in “turnarounds” generally when evaluating management and the “investability” of a stock I try to determine if management is acting rationally and doing the best they can be doing. Part of the reason I have a decent degree of confidence in CDLX’s ability to execute is that unless I’m misunderstanding something, it seems like bank relationships had deteriorated under previous management and banks that lean into the new bank tech see significant growth. Growing the engaged user monetizable base would naturally create a much better product to turn around and sell to advertisers.

In theory it makes a lot of sense. It’s not “innovative” or anything like that but it doesn’t really need to be. If the team didn’t have debt looming over their heads I think the stock would be a blatant steal, but sadly that’s not the reality we live in. If they are able to get some combination of updated BofA/higher partner share at remaining banks/new bank launches/ABNB back/a couple large signings/etc then it’s entirely feasible to see >30-50% growth by the end of 2025 which creates >$50m+ of AEBITDA, at which point debt can hopefully be refinanced cleanly without significant equity dilution. This isn’t a guarantee by any means and a key part to the puzzle is why BofA is shrinking. If my hypothesis about previous management poor focus leading to a bad bank product which is now fixed is not true, we may see significant equity dilution and disastrous returns.

My Perspective?

For what it’s worth my underlying idea with CDLX as an investment has not changed. The expected value I would guess is drastically higher than the current stock price given the value of a company doing ~$50-100m of earnings growing topline by 15-30% with 50% GM and a relatively fixed cost base is likely worth >$3-4b. CDLX’s path there seems somewhat achievable albeit not certain.

I did however make a rather significant purchase at $6 recently that I may trim a portion of back to the original allocation. A touch of tax optimization, a touch of getting concentration down to a slightly less unreasonable level.

I likely won’t do another specific stock update for a few weeks as we get more info on Q4 and into Q1 as I work on some other projects and do some more general journaling/writing.

If you’d like to subscribe and read more occasional thoughts you can do so below.