$XLF: Thread on the intersection of technology and finance | Dec 2019


I suppose I’ll get in on the fun:

1 like = 1 opinionated thought about the intersection of technology and finance, up to a cap of 100. 

1/ The financial industry is actually both an industrial cluster and a pervasive infrastructure layer, much like “manufacturing” or “tech” is, and almost all statements about it are necessarily overbroad and incomplete.

So, sorry in advance; it’s still useful point in mindspace. 

2/ Geeks underestimate technical sophistication of the financial industry by a scandalous amount, largely because most of it happens below the waterline and not in retail-facing experiences.

Finance probably has invested more in software, and earlier, than almost any industry. 

3/ *Within living memory* every bank statement, every stock trade, every invoice, every check, every mortgage, and every purchase made at a drugstore and backed by credit required labor of multiple trained artisans to accomplish. 

4/ “We” thought that the primary use of technology in finance was going to be adding up numbers really quickly without making mistakes. This was only the first order effect.

The second order of effect was enabling cheap transfer of risks over abundantly connected networks. 

5/ The finance industry has known for decades, and the tech industry has discovered recently, that the increasing complexity of the world has detractors and that lacking a vocabulary to describe complexity detractors will find a widely distributed scapegoat to point at for it. 

6/ Financial firms have natural economies of scale both for operational reasons and because scale historically functions as both a force-multiplier for distribution and because it offers first-order and Nth-order buffers against risks and the business cycle. 

7/ One of the bets for startups offering financial services is that they can innovate faster on customer acquisition to attract sufficient desirable customers from incumbents, while keeping the risks relatively constant, to overcome their lack of ability to rely on scale. 

8/ This is a bet most startups offering financial services will, bluntly, lose horribly, but that’s startups for you. 

9/ Geeks both underestimate how #%()#)&ed the engineering practices of large financial firms are and underestimate how impressive they are.

They make radically different tradeoffs than a startup would, including routinely spackling over technical debt with thousands of people. 

10/ One of the biggest problems in financial services qua tech is that the entire technical universe changes over every national border (sometimes within it, too) and that there is a weird goldilocks zone in terms of scalability on approaches for going international. 

11/ The goldilocks zone goes something like: you can’t go international without truly massive resources, which you can never justify allocating to you in a small market. 

12/ The upper end of the zone: If you have truly massive resources, it is tempting to use the difficulties of international as a barrier to protect margins of some of your business lines (you have a lot, almost by definition), causing you to avoid international cooperation. 

13/ For this reason, historically the really interesting international businesses in finance happened when they were young and hungry.

(“Who?” You can trivially name four global businesses in finance. It’s surprisingly hard to count to eight. More are under the hood though.) 

14/ Finance is a much, much more regulated industry than tech is, and a surprising amount of the competitive landscape is downstream of the preferences of your polity-of-choice and their elected representatives, which sometimes look irrational. They generally are not. 

15/ For instance, the United States does not have a strong opinion on the right number of tech companies in it, but it *does* have a strong opinion of the right number of banks, and it sets that number about 100X more than Japan does or 1,000X what the United Kingdom does. 

16/ Strong opinions on the polity for the number of banks are downstream of opinions as to what degree the financial industry should be an executional instrument for policy preferences.

Saying that outloud in as many words is uncommon, but the industry is that, too. 

17/ (We appear to be in the early stages of a societal debate, conducted largely via misunderstood proxy questions, as to what degree the tech industry should be an executional instrument for policy preferences.) 

18/ There is a tension between bundling and unbundling in financial services.

These are generally thought of on a services level, but you can also think of them in terms of customer segments and subsegments, which is probably more illuminating. 

19/ Because the most important thing for large financial services firms is distribution to customers and the cheapest customer to reach is one you already have, firms who might have once had a core product for a core customer now offer many, many non-core products to them, too. 

20/ Startups said “Aha, here is an opportunity. If there is a 100×100 matrix of broadly consumed financial products and potential customer segments for them, we can compete on product for one cell of that matrix, win by utter domination, and then start expanding from there.” 

21/ Startups which did this found, much like the finance industry did, that it is rare one can build a freestanding business (and extremely rare that one can build one on the venture growth trajectory) by chipping off a few percent of one of those cells. 

22/ This is further complicated by incumbents being able to both dynamically alter their allocation over that 100×100 matrix and also being able to dynamically alter their pricing, because they *also* read Innovator’s Dilemma, thank you, and didn’t get this far by not mathing. 

23/ Which is not to say there are no opportunities and no obvious opportunities, but I think I feel comfortable with “there are almost no obvious, popular, consensus opportunities.” 

24/ One “obvious, popular, consensus opportunity” is “Gee, wouldn’t it be swell if we just had a no-fee bank account made for folks in economic precarity. There’s lots of them; clearly this is a business worth being in.”

If only banks had good people who are good at math, right. 

25/ (Though, on that exact subject: Square’s Cash App is one of the most impressive product innovations I’ve ever seen, because it managed to basically deliver a thing that the industry thought rounded to impossible by making it a side effect of a bigger, better opportunity.) 

26/ Let’s talk discount brokerages, which I’ve talked about at length before.

kalzumeus.com/2019/6/26/how-…

This is a case where startups had interesting product innovations for 3~5 years and then incumbents said “Cool ideas. We will yeet these into being, as you cool kids say.” 

29/ The opportunity that most startup brokerages are really going for, and it usually isn’t messaged as this (probably for talent management reasons), is:

“We’re building a book of business to a point where it is attractive enough to warrant integration and digestion costs.” 

30/ A lot of the interesting bits in brokerages are actually a layer beneath, in e.g. custody, routing, KYC, etc etc, sometimes caricatured as being “arms merchants” or in the “selling shovels in a gold rush” business.

This lets more firms experiment with marketing/UX cheaply. 

31/ It also opens the intriguing possibility where someone might successfully launder better internal X systems into Schwab (or similar) by making them indispensable to a particular distribution engine and, hopefully, winning argument on which system survives acquisition. 

32/ Deposit pricing is, like taxes, one of the largest line-item bills on a household or business, and this is underappreciated because of how the bill is delivered to you and the relative level of sophistication required to understand that there was actually a bill. 

33/ Geeks who grok this the first time often think of deposit pricing as being, if not outright theft, theft-adjacent, but partly it’s tradition for how financial services are priced and partly it is a progressive cross-subsidization of some customers by others. 

34/ One of the issues in finance to have an opinion on is whether increasing velocity of money, and ease of allocating it via a swipe (or, in B2B, by a computer doing it for you) will cause most deposits to migrate to cheap-to-the-customer places or not.

Not obvious, either way. 

35/ Maybe zooming one level back for comprehensibility purposes:

There’s a risk-free rate for deposits. Your bank or brokerage doesn’t give you that. They give you a much smaller number.

The spread is deposit pricing. 

36/ “What are you paying for with deposit pricing?”

I don’t know, what are you paying for with software? The engineers or the person who will answer your email or the beautiful pixels on the website or the person awake at 3 AM to deal with security issues? 

37/ But different firms can allocate the bill differently, via pricing, and so it matters *quite a bit* if you think that customers and businesses discover price sensitivity in deposits in the next 10 years, because that implies that the price of non-deposit services is going up. 

38/ I used to be freaking mystified why it seemed like all the banks went out of small business lending at traditional rates after the 2008 financial crisis, and would only offer credit on credit cards to that segment. 

39/ This is *essentially* a pricing decision, with some regulatory background.

Non-intuitively, a bank would far prefer to give you access to a credit line at 9% APR if you access via a credit card than a 9% APR unsecured line of credit. 

40/ The reason for this is that the credit card would generate interchange revenue, which effectively acts as a kicker to the APR (potentially a 10%+ kicker depending on your behavior!).

I never understood that when I was on the customer side of credit cards / loans / etc. 

41/ Here’s that complexity again.

This is a bank relying on a four-party agreement between them, a credit card brand, a supplier, and a small business to facilitate what a lot of people think is the #2 reason banks exist, to provide loans to businesses for cash flow needs. 

42/ Effectively the conceptual negotiation, which critically never takes expensive human thought because all parties have pre-committed to their position, is:

Bank: “I can’t take all this risk to build restauraunt.”
Town hardware store: “They’re my customer. I’m good for a bit.” 

43/ If you *can’t* have the town hardware store, or somebody else, agree to subsidize your risk (via interchange revenue), your traditional options at banks either dried up or got much more expensive, or you now go to OnDeck/etc wrapping private capital providers for much more. 

44/ One of the big opportunities in financial services are places where fees are theoretically supposed to pay for unconflicted advising by expensive people but where the quality of the advice is, ahem, highly variable. 

45/ You can often find these opportunities where it is literally illegal to read the industry’s uncontroversial best practices to the customer accompanied by the words “You should” unless you are a licensed professional. 

46/ For example, under some regulatory regimes, I can probably say that whole life insurance has fat margins, that it embeds extremely suboptimal investment decisions, that the most valuable part of it is the actual life insurance, and that term life is a better vehicle for it. 

47/ But I’m probably legally prohibited from saying “You should buy term life insurance” or causing a computer program to do so if I’m anywhere near a transaction, at least under some regulatory regimes.

I haven’t passed the test. 

48/ (I also haven’t posted a reputational bond which would allow a regulator to threaten disproportionately severe punishment if I were to exceed their tolerances for salesmanship, which is an intended component of regulatory regimes that tech has insufficient appreciation for.) 

49/ Regulatory compliance is a fun area for financial companies, largely because the regulations aren’t the regulations and the regulations that aren’t the regulations aren’t static. 

50/ On the extremely suboptimal side of the house, successfully grokking the *real* regulations is treated as a proof-of-work, such that if you’re capable of talking the talk and walking the walk you’ve earned a bit of a buffer around your activities. Few would admit to that. 

51/ On the better side of the house, regulators are not uniformly clock-punching value-destroying intermediaries (like geeks often assume) but are frequently informed professionals with lots of institutional memory and some level of appetite for listening to proposals. 

52/ Firms largely choose one of a couple engagement strategies with regulators. The dominant one is Enthusiastic Compliance (TM).

“Break the rules until you can make the rules” is extremely, extremely not the norm in financial services, except in the cryptocurrency economy. 

53/ There is, as of 2019, more paper, and more *required* paper, than you would think in financial services. This is decades after frontpage stories about Paperless Offices.

But there has been substantial improvement, even in the last few years. 

54/ The same goes of phone calls, faxes, API-via-secure-FTP, harddrive-over-FedEx/couriers as an encrypted API transport, etc. 

55/ Costs of customer acquisition in financial services are routinely in the hundreds to thousands of dollars for *retail-facing* accounts, because they anticipate customer lifetimes measured in decades and growth in the value of those accounts over times. 

56/ When e.g. Chase paid $1,000 apiece for signups to their new top-tier credit card, there was widespread chortling by competitors, not because they thought that was unfair or unseemly but because they thought it was *unwisely calibrated*. 

57/ This is unfortunate for competitors to financial services, which want to grow explosively but want to do it on extremely light marketing budgets, relying on e.g. product quality, social amplification, and better understanding of App Store SEO to drive adoption. 

58/ Startups don’t want to spend hundreds of dollars on customer acquisition with razor thin margins because then sustaining their growth ramp will cause them to lose more money every year, costing them more dilution every year. 

59/ This is as true of e.g. mattresses as it is of financial services, but startups in finance have a problem that other startups do not: in finance you can pay the cost of customer acquisition *directly to the customer* without this looking like accounting shenanigans. 

60/ Ever wondered why “Silicon Valley preferentially makes products for rich people in the bubble” is widely believed and yet Chase and Amex probably have 90% wallet share of SV participants? This is basically why. 

61/ When AppAmaGooBookSoft think of financial products they think of ones which can be distributed virtually universally over their userbase (e.g. Apple Card).

When startups think of them, they often are forced to look at customer groups that Chase et al don’t want. 

62/ The toothbrush test is approximately as applicable in financial services as it is for marketplaces. (The toothbrush test: no marketplace you don’t use twice a day will be a massively viable business. Uber for X is a lot worse than Uber for Uber.) 

63/ The exception to the toothbrush test is originating extremely, extremely valuable set-it-and-forget-it transactions, ideally ones which are hugely backweighted. Unfortunately these are limited in number by nature and switching costs are extremely high. 

64/ I think you will find many more interesting startups in financial services either at the toothbrush test or at the “giant, compounding transaction with multi-decade time horizon” than you will in the No Man’s Land of e.g. credit card refinancing. 

65/ In the No Man’s Land you have to pay essentially your entire customer acquisition cost over again to reacquire customers you already won, while not having them increase in value in interactions with you, and also not materially decreasing in risk or operational costs. Taking a break for family dinner; 35 more coming later. 

65/ Wealth management historically sells a small amount of very expensive labor (portfolio allocation math, done by expensive professionals) with a large amount of medium expensive labor (counseling delivered by a salesman with greater facility with numbers than most clients). 

66/ Computers will entirely replace the first bit. The big question for the industry is to what degree the second bit ends up being provided by an app, by brand positioning, etc, and to what extent it gets provided by an actual human. 

67/ Many startups bet that there has been a fundamental shift in buying behavior around the thermocline created by “Did you grow up with the Internet?” and that people on one side of that thermocline have preferences which are *unrecognizable* on the other side re: sales. 

68/ The incumbents believe something closer to “Well, our cost base assumes that we put modestly expensive professionals in expensive real estate whether we like to or not, plus we’ve met young people before and heard Oh God They’re Nothing Like Past Generations before, so.” 

69/ Good sales people in retail-facing financial services are paid tiddliwinks by comparison to good sales people in tech, and almost everyone capable of selling mutual funds at a wirehouse should be selling B2B SaaS instead, ironically often to the same demographic. 

70/ “That won’t happen, they don’t understand software” say some commentators, who probably overestimate how competent in distributed protocols one needs to be to sell workflow SaaS and who underestimate the intelligence of people who have passed the Series 7. 

71/ Much like in other B2B services, at certain tiers of sales in finance, people stop calling you a sales representative. Now you’re an investment banker / consultant / etc.

Their productivity likely improving due to both CRMs and business processes which augment capabilities. 

72/ A large theme in investing is bringing capabilities which used to require bespoke, artisanal work and making them available to the mass affluent. Tax loss harvesting is one example.

One question is whether, if you build it, they will come. 

73/ A trickier question is whether all products of the financial services industry can be offered to substantially everyone, in a way described publicly, without attracting extreme adversarial attention or causing excessive amounts of brand risk for too little revenue. 

74/ In financial services as well as in adtech, the accusation “sells your data” is made extremely frequently and accurately descriptive of a transaction extremely rarely.

Imagine the meeting. “Interesting idea, but could we just take their money instead? We’re set up for that.” 

75/ Scaling computer systems is scaling computer systems, but an interesting limiting factor in scaling of computer systems in finance is “Does this system have rate limits controlled by human interaction?”, in which case scaling from a technical perspective rounds to trivial. 

76/ That didn’t used to be the case. Visa volumes used to be big numbers requiring some of the smartest engineers on the planet to use the biggest iron available.

But Moore’s Law compounds faster than the global birthrate does. By a lot. 

77/ There are still extremely non-trivial scaling challenges in financial systems where computers talk to computers, such as e.g. adtech (which isn’t a parallel system of financial exchanges but isn’t not a parallel system of financial exchanges), stock exchanges, etc. 

78/ A prediction for something you’ll see in the future and think is really weird, but which is virtually inevitable given the economics: Internet celebrities are going to start offering Durbin-exempt debit cards to their tribe, enabled by an underlying platform and bank(s). 

79/ The reason this doesn’t happen in the status quo is because it requires too much bizdev by too many people but fundamentally the financial industry wants to facilitate this transaction if you can move 100k of anyone or 10k of the right people to use plastic with you on it. 

80/ Settlement times, much like interest rates, have a natural lower bound of zero… and we will quickly discover that that lower bound wasn’t real.

In the case of settlement times, it is because you can use T-X settlement as a marketing differentiator w/o all that much risk. 

81/ We’re starting to see that from some challenger banks (e.g. Monzo), and the natural response from incumbents is to spend 3 years not noticing it in surveys, 2 weeks doing SQL queries, and 18 months doing a product cycle, then deploying it to 100 million people simultaneously. 

82/ There are substantial opportunities available both within traditional finance and within startups to become a domain expert at the other industry and explain/advocate/plan/etc within one’s own industry.

(Some folks will end up doing white elephant projects, but not all.) 

83/ Every startup offering financial services has both an integration with their partners and then named people on speed dial who they can call when the integration breaks. Sometimes this by necessity means a lot of people in a lot of places. 

84/ This goes in the other direction, too, and it’s both terrifying and sort of beautiful. If a tiny bank in, I don’t know, Ireland has a backhoe hit their Internet connection, it is moderately likely that the 1st person in outside world who notices is a little fish in big pond. 

85/ (The general contents of the first phone call would be “Hey, this is your account manager at $COUNTERPARTY. I was just checking and it doesn’t look like we got the upload for today… any color for me? Resubmit window closes in…”) 

86/ Because large financial firms have competing fiefdoms and organizational subcultures in much the same way that large tech companies do, it makes a lot of difference to strategy and day-to-day operations which fiefdom the CEO/etc came up in, just like in tech companies. 

87/ @cperciva pointed out that I missed two items and duplicated another one, which brings up an important point: despite being the canonical use case for ACID databases, almost all financial firms aspire for (and achieve) “eventual consistency, almost all of the time.” 

88/ Financial firms are likely to slightly, but not much more than slightly, lag enterprise SaaS with respect to adoption of sales and account management via webinar and 1:1 videoconferencing.

This can 5X utilization of advisor while having them be in a cheaper city than client.  

89/ The most interesting thing which is widely deployed in Japanese banks that I haven’t seen in US banks, but which will inevitably be deployed in US banks, is phonebooths with videoconferencing software and ID-reading hardware connected to a call center. 

90/ This will be generally good for customers who want more human touchpoints, and will bring down the OPEX of account opening and servicing substantially.

It will, less fortunately, further cause deskilling of the branch-based bank employee, which has been ongoing for decades. 

91/ (The availability of backoffice doesn’t by itself make branch bankers less capable; it means a business process which historically trained them to answer the range of everyday to sophisticated financial problems encountered by anyone in their area doesn’t need to exist.) 

92/ There are many problems in startup financial services which are technical problems, but there are virtually none which are uniquely solved by adoption of an architecture, language, stack, etc.

Most of the technical problems are “How do people/organizations work together?” 

93/ You might sensibly say that “‘How people work together’ is more a management problem than a technical problem” but in this case the controlling abstraction is sometimes *very literally* where you draw the API boundaries. 

94/ People from finance backgrounds and people from tech backgrounds have extremely different ways of looking at the same artifacts, often coming to the conclusion that “This makes no sense and it is impossible you got this far without having some competent adult supervision.” 

95/ Although you would think that both finance and tech would naturally cater to the wealthiest members of society and so they would consistently have access to the best services, a combination of Innovator’s Dilemma and Worse-Is-Better means that future is present on fringes. 

96/ Examples of this: M-Pesa was one of the world’s most advanced (and best distributed in customer set) payment networks within years, and relatively not-too-rich folks using the US->Mexico remittance corridor have much better FX pricing than rich Americans. 

97/ In my experience, more people in the US (and, broadly, Western) financial industries should be paying attention to customer-facing UXes from Asia.

QR code payments are probably underestimated, in the same way that QR codes themselves were underestimated. 

98/ Virtually every conceivable subfield on the intersection of these two industries could justify its own hundred tweets blurbing a shelf of books covering a variety of costly learnings. Fraud. Identity verification and management. FX risk. Behavior during financial crises. etc 

99/ All of the professionals in tech and finance who make decisions which affect the systems that your live depends on work on the same Internet you do. Many would be happy to get coffee. Relatively few feel like enough people care about their line of work. 

100/ It occurs to me that in 99 tweets I haven’t even mentioned cryptocurrency, which feels appropriate given its demonstrated level of impact. </rimshot>

Alright, that was fun. Don’t love this form factor, but let me know if subtopics here are worth an essay sometime. 

(In the spirit of mathematical rigor: it’s closer to 35X, not 100X, between the US and Japan.) 

Leave a Reply

Your email address will not be published.