What Makes A Mistake?

Intro:

I’d like to spart spitting out some more general investing blogs over the next couple weeks/months as I’ve mentioned on Twitter. Call this blog the start of that. For name specific stuff you can read my

Clarity Markets

blog and Twitter where I’ll start posting updates on free CVNA 0.00%↑ KMX 0.00%↑ CDLX 0.00%↑ stuff. This article will naturally touch on those, but they won’t necessarily be the focus.

I also will definitely touch more on AI in the future. I can’t stress enough how impressive Gemini is (or any model really) compared to where we were a few years ago. I also chucked some money at GOOGL 0.00%↑ for posterity’s sake.

Anyways, feel free to subscribe and such. The format here is a bit all over the place, but perhaps the reading experience is helpful in jogging your own thought process. Cheers.

Defining Mistakes:

My background is primarily Long Only equity investing. Long term, embrace volatility, blah blah, I’ll stop short of “private equity in public markets”.

I say this to provide context on what I mean by “mistake”. I’m not talking about “I entered $509.1 instead of $50.91” or “My Newtonian-Lagrangian asymptote pi kappa phi was off by 7 Dusseldorf’s within 5 picoseconds on that trade and we lost .00001 bps”.

In this context, by mistake we’re talking about mistakes predicting differentiated future business trajectory 5 years out from an outside perspective. Really simple stuff in all honesty.

In this area, even defining mistakes can be a challenge. To be realistic, there are a bunch of constraints on a fundamental equity analyst/PM. You only have so many hours you can feasibly dedicate to research, so many dollars you can spend on research, and basically unlimited knowledge to collect. Many public companies will have dozens of internal teams working to create simple financial projections that can be totally wrong even with internal data. Our industry hires 20 somethings to outperform their work with way less info!

Personally, I am very lenient when it comes to what constitutes a mistake and dislike the results bias from allocators. For example, it would have been rather ridiculous for an asset manager to be ridiculed on 9/17/2001 for not foreseeing the recent market drop. It of course hinders their results in that moment, and the job is to make money, but some level of detaching outcome and mistake is clearly warranted.

Yet at the same time if you for whatever reason happened to both manage money and have a good relationship with Al-Qaeda, you could have made generational returns (ignoring the future indictment of course). The conditional probability of two Boeing 767’s hitting the twin towers that day would have very different answers depending on who you kept as friends.

Now of course, 9/11 was an extremely tragic one-off terror attack that permanently changed aviation safety. The reaction was not ridiculing asset managers nor focused on asset managers at all, it was largely a nation coming together to offer condolences and help where necessary and focusing on not letting it happen again.

Equity research is clearly much lower stakes, but I use the extreme to highlight that even very very bad outcomes, arguably one of the worst events in the history of our country, can occur with little justifiable ability to sling around blame even if theoretically the outcome was knowable.

So what would I call a mistake? Let’s do a non-exhaustive list:

1. Missing table stakes information.

   1. Say your model has 10% growth for CVNA units in Q1 heading into earnings even if that is practically impossible based on alt-data. You can’t pull the “unknowables” card if clearly everybody else knew.

2. Being overly emotional.

   1. If the market makes you feel some type of way and you blow out at the bottom or over indulge at the top, that can be extremely value destructive.

3. Optimizing for things other than returns.

   1. You’re a fund and you buy some avant-garde thesis vs a higher underwritten return in say MSFT just so you look differentiated for LP’s. Or you’re an individual and you buy GME so you can chat with your buddies about giving it to the “hedgies”.

4. Limiting your opportunity set.

   1. If you only did “predictable future cashflows” stocks, you probably underperformed the indices over the past couple decades as we witnessed almost all returns come from like 10 tech stocks.

5. On the flip side, straying from your lane.

   1. There are also plenty of people and firms that really had no idea what they were doing, but hopped on hot new trends. Sometimes it works, like AI/crypto. Sometimes it’s hilariously value destructive like 3D printing. It’s good to know where you can realistically provide good analysis, but it’s a balancing act with not getting stale.

6. Trusting the wrong people.

   1. This is the most nebulous of the above. You can’t get inside someone’s head, but it’s good to occasionally take a look back and see if your judgement of people has been directionally correct. Was a CEO truly a visionary or just a good liar?

7. General judgement errors and logical consistency.

   1. If there truly were a bunch of unknowable’s, what’s the track record of your judgement with those? Is there a way to pin down what went wrong? Why? This is sort of the catch-all for bad outcomes that flow through the above without a clear error.

I like to have this list as a way to mentally benchmark everything I own and the outcomes, it even extends far beyond equity investing. Each mistake is different, the ways you handle mistakes differ, but having some general buckets makes it easier to parse when you don’t always know what the mistake even was. “Why did my investment underperform” can be entirely unrelated to “Why did this business underperform”. Say you loaded up on physical retailers in November 2019. You got totally and completely fucked. The business underperformed because of Covid, but your investment likely underperformed (assuming you held) because you got something wrong with modeling the risk there in a new environment. The business couldn’t have avoided covid, but you personally have much more flexibility in public markets than say Party City did to entirely change their business model while heavily levered.

Running your own outcomes through the list above hopefully creates some more intellectual honesty about where process improvements can happen. We don’t want to place blame for unknowable’s, but we also don’t want them to be a scapegoat.

Why Opine On Mistakes:

My background includes a bit of work on the allocator side of things not just the direct investing side, so I’m likely a bit more indexed towards the “theory” here. Like anything that’s a highly competitive multi-player field, I believe that to be good at investing requires constant iteration and learning. Analyzing mistakes happens to be quite good for that! You can also obsessively try to not make mistakes and still get bad outcomes, which is probably the best time to take a step back and analyze the situation for potential lessons.

With myself for example, I have had one position that has caused meaningful drag on performance in the last few years. I easily liquidate it, continue on with my life, quote a high hit rate, say things were unknowable, and move on. That is certainly better for marketing, and perhaps even a good approach overall, but instead here we are, trying to do better.

The Flow Chart Of Mistakes and CDLX:

Let’s do a quick case study using my mistake list and CDLX. To be clear ahead of time, I still own CDLX and don’t hate the stock, but I think at this point it may warrant a “mistake” flag even if I’m not yet 100% sure. I do this to be illustrative of the process, condense my own thoughts, and perhaps provide a learning opportunity. So let’s take a self-deprecating journey together using my list from above:

0. Preview:

   I first purchased CDLX in I believe late 2021 or 2022. I wrote extensively about the stock here and on Twitter. It has varied from a small starter position, to at peak ~50% of my capital, to currently 5-10% given on the day. I’ve added, I’ve trimmed, I’ve seen 3 CEO’s, I’ve had it be up 3-4x, I’ve had it be down >75%. Really the been around the block a few times in a shockingly short timeframe. This will be a very condensed analysis of how things have gone and my perspective.

1. Missing table stakes information.

   The first mistake I always consider is if I just missed something. Especially with small caps, you can have days when stocks just arbitrarily go up and down 5-10%. Combined with the rather common semi-neurotic boss, it’s easy to stress out that the market knows something and you fucked up. What I can say with CDLX, is that I don’t believe I was ever behind the curve on information. Due to my own research, my network, calls with management, the work of

   Austin Swanson
   , expert calls, 1P data collection, etc, I’d like to imagine I was rather informed on the actual established “facts” of the business as well as one could from the outside.

   That’s not to say I knew everything I wanted to know, as CDLX is perhaps the most opaque stock I’ve had the pleasure of analyzing thus far. The secrecy of banks combined with general small cap inefficient IR leads to a company that can’t even name their own clients. We have no disclosures on activation rates, MAU’s has been redefined twice and none of the three versions are particularly good, economics per client are directional guesses at best, ROAS is drip-fed by agency LinkedIn posts, etc. There’s a ton of missing info, but none of it is table stakes.

   So I’d tentatively say that my actual fact finding with CDLX was done decently well. Of course maybe it becomes obvious in the future I did something clearly wrong, but as of now I’d say the judgement leans in my favor.

2. Being overly emotional.

   If you ask my ex-girlfriends, my problem here is typically the opposite! I do think it’s natural to get an emotional link with any position, especially if it’s volatile. It does suck for something to erase value, it is great when something creates value. The trick is to minimize the extent to which that has an effect on investing returns.

   I am quite transaction light in the account I manage, so I wouldn’t say emotions really have much impact on the returns there. My PA is more transaction heavy, more volatile, but has ended up in a pretty similar spot return wise. Doing a little PA trading and arguing online can be an outlet for when you feel some type of way, but leave the legitimate transactions to times when the headspace isn’t clouded. I’d like to think I did a decent job of that.

3. Optimizing for things other than returns.

   Thankfully I’m just some guy. I was managing internal capital and beholden to a boss, not numerous external LP’s and fundraising needs. I definitely made him quite unhappy at times for a variety of reasons, but a little bit of an adversarial relationship when managing funds for someone can be healthy to keep focus. This does become a very hard problem if you do need to fundraise for survival, so it can be tough to call it a mistake, but when our lens is simply “are you doing what’s best for returns”, we have to be a bit ruthless. If I end up in a fundraising position at some point, I’d hope to not fold on this principle, but at the end of the day you need money to manage money.

   From a personal perspective however, there’s certainly some conflicts when you start publicly writing about stocks. In a world where CDLX was say $40 currently, not only would I have a lot more money, but I’d also likely have a lot more followers, fewer upset DM’s, etc. When publicly giving opinions on the future of anything, those opinions will come back to help or hurt you. Your reputation becomes leveraged to the opinion. This has the opportunity to produce some very unideal doubling down, which can not only be destructive of your reputation, but more importantly your returns.

   I’d like to again say I handled this somewhat well. I don’t think my writing has had a material impact on my portfolio management, if anything it helps with thought crystallization which can make you think harder about your assumptions. Of course I can’t say definitively as I’m the guy who wrote about the stock. What would I have done if I never wrote about it? Impossible to say for sure. I have added, I have trimmed, generally because I think the stock is cheaper/more expensive relative to other opportunities. My Twitter/Substack following was never the prime consideration at least consciously. So we’ll tentatively call it a pass.

4. Limiting your opportunity set.

   This mistake isn’t really applicable for a stock that you actually invested in, more of an error of omission type deal. That said, there is an opportunity cost to every second of your time. For example, I’d say on average, each minute of research on a small cap is more likely to produce marginal insights vs each minute on a megacap. CDLX was definitely one where at the time incremental work felt productive as things like SBUX temporarily returning was visible via app tracking and saw the stock up something like 15% over the course of the day as the news spread. Little nuggets of differentiated information were obtainable, but was it worth it? For small cap differentiated stuff you have to be checking niche channels like singular apps, personal LinkedIn’s, etc. It isn’t generally cross productive with your other holdings. It’s also very hard to formalize by nature.

   From the lens of reviewing my research process, I definitely could have been more time efficient. Like in hindsight if I threw a dart at my watchlist from 2022/2023 everything is up astronomically. Many things I briefly owned heavily outperformed CDLX. Would more time spent on those have been value accretive? Very possibly. This realization is part of why I’m trying to build something that makes it easier to aggregate niche research, but WIP.

   Lesson learned? Going forward try to be much more tech enabled with aggregating weird info flows. Not always possible, but good to keep an eye out especially with how fast AI changes.

5. On the flip side, straying from your lane.

   To be completely honest, I don’t think CDLX was much of a stray. Much of my investing has been in things that are some mix of finance/tech/consumer. My largest position is an online car retailer that packages up loans. The portfolio I covered was almost entirely SaaS or consumer facing tech, with revenue largely coming from advertising for the latter. I really liked META, hated SNAP, thought TTD was too expensive, etc. CDLX was not pulled out of a hat one day, it was a somewhat natural fit at least in terms of understanding what could be understood about the business. I never felt like anything was too complicated to understand.

   That being said, I do think I leaned a bit on the fact other people I believe to be quite smart also owned the stock. Austin Swanson, Cliff Sosin, numerous anon twitter accounts, etc. Not to a degree that I would at all say I didn’t “own” the position, but I’m definitely younger, less experienced, and VASTLY poorer than some of those guys. I’m not exactly the most respectful of authority, but it was definitely a comforting factor that numerous smart guys believed in the story. I think this can even be helpful in terms of figuring out if the market knows something you don’t. If everyone has a pretty similar idea of how things work, then you likely aren’t fucking up Mistake #1, which is a pretty good spot to be. I’d tentatively say the networking here was additive even if I may have sold without it. On net I’d say it made me more aware of the known truths, and what I do from there is up to me.

6. Trusting the wrong people.

   This is where I think we start to get interesting and have to diagnose a bit. The difference between the world now and when I first entered CDLX is something like.

   1. CLO’s are much more common, with Chase launching their own and AmEx adding CDLX in to their network

   2. The advertising world has generally ramped significantly post-covid with META and GOOGL materially larger than a few years ago. The economy has grown and with it digital advertising

   3. Chase saw numerous years of double digit growth and showed the power for the new ad UX combined with bank push to create large growth for CDLX.

   4. CDLX integrations with partners have lead to better proving out of ROAS

   5. Despite the above, BofA is pseudo dropping the contract and explored different vendors. Chase pulled some volume in house. CDLX is smaller than it was a few years ago.

   So why has CDLX underperformed the sort of general market of both digital advertising as well as CLO’s? Why have new entrants taken up some of their space?

   The exact reasons here are still not 100% clear, nor may they ever be, but I have some guesses. Let’s highlight each

   1. Tech development when you don’t control distribution and have 20 different versions in highly regulated spaces with unique security audits is probably really hard. CDLX will naturally take longer, with higher OpEx, to do rollouts relative to those that control distribution. They are also beholden to tech dinosaurs that operate businesses on the opposite end of the spectrum from “move fast and break things”

   2. Given a lack of distribution control and not owning the data, sharing results naturally has some “trust me bro”. Numerous expert calls have highlighted disbelief of CDLX ROAS given the platform is niche and thus not “feelable” in results like a META or GOOGL. Even if CDLX is say providing $1m in value, is that meaningful enough to survive general corporate politics and turnover within advertisers?

   3. Numerous initiatives have been a total waste of time at best, or obscenely value destructive at worst. Dosh was acquired for ~3x the current market cap and no longer even exists! Bridg was even more destructive. The push for self-serve was dropped after a year of wasted effort. Product level offers were promised ~5 years before a single one went live. Local offers was a total bust. Fintech partnerships amounted to nothing, etc.

   4. As a result of the above, we’ve seen management musical chairs and some partner loss for banks that inexplicably make their offering worse instead of modernizing and dealing with CDLX.

   That’s a very watered down version of events, but it’s my basic view of how things developed with CDLX. The general market has left them in the dust unfortunately. The question for our purposes here becomes, how much of that is on the people?

   I truly think the biggest problems in hindsight really were the lack of distribution control and lack of technological control. This is an n of 1 company, there isn’t a plethora of deeply integrated transaction data advertisers. Going in I knew it was a struggle and probably would continue to be, but would have the people really changed that? I think it’s possible, but hard to prove of course. I wouldn’t say CDLX has been dealing with A players, but one would hope the businesses they invest in don’t require A players, as getting A players in the small/mid cap space can be quite difficult, especially if they have to work with banks.

   Where I think CDLX has really struggled on the people side is the insane value destruction and opacity. I definitely will have a higher quality bar when evaluating M&A activity going forward. I definitely won’t be playing lawsuit chicken where we say everything’s fine then arbitrarily settle with no justification. I will absolutely have a higher bar for transparency when decisions are made that are contrary to what I believe should be happening.

   For example, when CVNA was tanking in 2021/2022 my initial models had them cutting units to reduce OpEx and weather through a demand crunch. They did not do this, and instead opted towards growing their way out of the hole to some extent. This didn’t match my thesis going in, but I could see in the alt data and the models that such a strategy could make sense, that they were making headcount reductions without imploding the business model, etc. With CDLX the opacity mixed with the messiness created some tough problems, and it’s hard to have faith the tough problems are being solved when nothing is communicated.

   I’m certainly sympathetic to their plight given the space they work in, and I do believe current management is a better fit, but I think it’s clear in hindsight that Lynne didn’t really know what was going on under the hood, that Karim didn’t have a good idea what he was signing up for, and that the communication through it all was lacking.

   The biggest lesson here is probably that if a CEO doesn’t know how their own business works and I can’t externally prove it out, then it will also be hard to maintain and grow partners/customers. That said, it’s a tricky game because stocks are cheapest when everyone thinks the management is dumb as bricks or criminals. Look at META, look at CVNA, look at TSLA, etc. My bar will simply a bit higher in terms of balancing transparency and opacity.

7. General judgement errors and logical consistency.

   Thus in all my hubris I’ve determined I did not largely make mistakes above, I simply had bad judgement!

   With CDLX, what it really comes down to performance wise is a couple things. First off is that despite being directionally correct about CLO’s scaling over time and Chase likely growing >20% per year by my guess, the business has not seen that output. This is largely due to losing BofA volume, very slow ramps at partners other than Chase, and a touch due to Chase in-sourcing.

   Like to be clear, management was initially telling us the new ad UI/UX would be rolling out starting in 2022. We are now in 2025 and not a single old partner besides Chase is there. US Bank has had the new ad server since 2020 or 2021 and never updated to the new UI. BofA simply never updated to the new ad server, let alone any UI changes. Simply adding some pictures and marginally better ad delivery is an endeavor that has taken ~5 years to get live for one partner. CDLX could have alternatively never acquired Bridg for product level, never acquired Dosh for UI/UX acquihire, fired everyone not just working in product maintenance, and it would likely have 10x-20x the market cap just from eliminating the value destruction.

   The basic idea was like “you got these banks to sign up and share transaction data, surely getting them to add some pictures won’t be that bad” and here we are. I DRASTICALLY underestimated how much of a pain the bank tech issue would be, and I was highlighting it in my initial pitches! Hell AmEx was announced over a year ago and we only started getting some offers a year later, still waiting on a full rollout! I’ve heard that CDLX even offered to loan BofA engineers to make the tech transition happen and instead BofA just degraded their consumer experience because tech is too hard. The real lesson here with the banks is that they all absolutely suck besides Chase and AmEx, which if you had invested in them instead you’d be far better off!

   Beyond my error with the bank tech, I also came into this expecting advertisers to be a bit more rational. It’s a bit of an alien view for me to have CDLX quote you a ROAS, have Nielsen quote you a ROAS, then you turn around and say “I don’t believe you”. I can understand the limitations of the CDLX model that can create that inclination where others don’t have it, but the idea that banks are providing transaction data to a company defrauding you just never jived. The idea was that AmEx and other partners would add volume to make results “feelable”, but the issues with BofA and other partners never scaling into new tech made that idea moot. The idea was that if CDLX was growing >30% a year, surely there was momentum there and belief in the product. Instead it seems like the growth was just because it was a shiny new thing, and once times got a bit tighter, the shiny new toy wasn’t so shiny and everyone retreated to META and GOOGL. My error here was assuming advertisers were playing the game I had defined in my head of turning advertising budget into revenue at the best rates possible. Instead, they were turning advertising budget into job security and herding. One the new toy lost it’s luster, why come back even if it was now a better product than when you were spending on it before?

   As a result of the above, I got pump faked not once, but twice on CDLX. Two times I was up hundreds of percent, and currently it’s down ~60-70%. Drastic differences in IRR! I was indeed correct that the opacity could create very large mispricing, but likely should have had a better model to deal with that. It’s also entirely possible to stock does extremely well from here, because again opacity creates very large mispricing, but it’s also possible it doesn’t and I locked up capital in something where had it been invested elsewhere it would be worth 3-5x or more. I’d tentatively say that a large difference between CVNA and CDLX was that with CVNA I could tell that the business was executing and momentum was moving in their favor, even after big headlines like debt restructuring and such passed. With CDLX we’d get some nice snazzy headlines, tie ourselves in knots trying to figure them out while management said “we can’t comment on that”, then get rug pulled on the disappointing results.

   It’s hard to take definitive lessons from bespoke situations, but I’d guess it’s likely beneficial in the future to more seriously consider the sustainability of quarter to quarter narratives. This is somewhat antithetical to the “long term volatility embracing” approach, but just because something sounds great long term doesn’t mean the narrative power will hold before it’s realized. CVNA had the benefit of beating sell-side expects for nearly 3 years straight. CDLX was a black box of results with product timelines measured in years. Even BLND for example we could see the rate of home purchases and just plug in BLND taking price. If I’m taking positions in things based on over-done narratives to the downside, I need to be prepared to exit positions based on over-done narratives to the upside. Obviously harder said than done, but it’s a process.

Fin:

The goal here was to illuminate my thought process around making investing mistakes using an example from things I’ve owned. If you’re unfamiliar with the name it will likely fail to resonate, but hopefully the general ideas can be worth the skim.

Part of the purpose of this blog is to condense my thoughts and just write stuff, and the theory of investing is a big share of the mental work whereas much of the writing is the stock specific output. I think mistakes are just generally under covered and discussed, and reading through other people’s learnings can be quite helpful to me, so maybe this was helpful to you.

Feel free to subscribe or share as you wish. Cheers.