I Posit: Tokenization Does Not Require Tokens

Discover why McKinsey's latest tokenization insights raise concerns about their feasibility, and explore the complexities and potential pitfalls of tokenized financial assets.

McKinsey published a new tokenization article titled “From ripples to waves,” so thank you for volunteering for featuring in "Rolling the Dice on Tokenization.” Initially I was worried that there isn’t much to say. McKinsey, you know : top notch CEO level strategic insight.

I do thank the person who made me aware of it, ensuring a timely review. Keep it coming.

False Tokenization Déjà vu

McKinsey starts by making an interesting observation that when it comes to tokenization, “there have been many false starts and challenges thus far.” It would be very interesting to get their perspective on this. False how? A failed product development isn’t necessarily false, I suppose, even if commercially unsuccessful. And even so, it would suggest there is a ‘true’ or correct approach, and they have relevant criteria that allow them to make correct predictions about the false and true innovation presumably meaning the commercial success of a new tokenization model before it is launched. Since I doubt this is the case, perhaps technology innovation and product experiments that fail to succeed should not be thought of as false. But if the can do this with high accuracy then I want to sign up but they probably wouldn’t want me.

The paper argues we do see a tipping point with ‘the first at-scale applications transacting trillions of dollars of assets on-chain per month.’ The evidence they give is a McKinsey article from 2023 which discussed whether tokenization could “be at an inflection point.” It is a bit ironic, then, to use evidence they had a year ago to argue now is the time. Funnily enough, the 2023 article was prophetically titled: “Tokenization: A digital-asset déjà vu.”

They refer now and in 2023 to a solution by Broadridge, which McKinsey, I would say, ‘falsely’ describes as “a fintech infrastructure company” facilitating “over $1 trillion worth of tokenized repurchase agreements” per month. They provide technology and back-office services to capital markets; so do IBM or Microsoft. And engage in digital and blockchain. Does this make them a FinTech? Hmm? Broadridge sometimes puts that sticker on its presentations, that’s true.

Broadridge’s ‘Distributed Ledger Repo’ solution is a shared workflow solution, and there are a handful of others such as Euroclear Taskize dealing with other activities in securities markets not necessarily using blockchain. “So far, Broadridge is hosting all the client nodes,” is what they said in 2021. I have not checked, but I doubt that there was a material shift since them. So the fact it uses DAML blockchain doesn’t mean you have a an environment that resembles grand ideas we have about decentralised peer-to-peer solution with fractional ownership etc. It doesn’t make it a bad solution, quite the opposite, but it is not as transformational as blockchain technology is often described to be or lacks the condition. Its process automation and guess what, banks have been doing this before the invention of blockchain. And thus it raises an interesting question about true and false models: given the constraints imposed on the technology, what does it tell us about feasible business models and of course predicted market volumes.

And then McKinsey explains tokenization in a way that sounds superficially okay but is very misleading and gives the wrong impression. “Tokenization is the process of creating a unique digital representation of an asset on a blockchain network.” Saying this is, in a way, a ‘false’ definition.

What makes this statement false: “Tokenization is the process of creating a unique digital representation”

The phrase "unique digital representation of an asset on a blockchain" implies a unique artefact, physical or logical separation of a data record that does not exist in practice. The token is not a separate entity but rather a construct within the blockchain’s decentralised data structure, where consensus among nodes validates their existence and transactions. It's crucial to emphasise that the token exists because the network of nodes agrees on its state and ownership. Assume there was a blockchain with two nodes and a token issued through a smart contract; both nodes construct the state, but neither node has ‘the’ token. What confirms the token is the fact that the nodes agree there should be one based on their independent calculation, not the fact that their ledger copy has that information. Again what makes a token in blockchain different from other databases is exactly, the inability to locate a place that is the authoritative data source: it is being constantly verified and updated what truth was and how to make the next state change into an updated truthful ledger state. And the block in blockchain only has the transactions for the tokens. I wrote about this before (and it features one of my best art creations I think, this portrait of Francis Bacon being interrupted writing his manuscript of the ‘Novum Organum').

It is vital, in my mind, for the industry to adopt terminology that is informative and to prioritise these technical concepts in how we define the terminology. This would allow us to think about the situation with more precision and at the needed level of detail. I think instead of DTCC, Euroclear, and Clearstream ‘wasting’ investing more time in offering co-parental rights over their orphaned control child, spending some time to work this out would help the market to make progress. Call me? I might tell you how for a little fee, but I won’t accept MMF tokens as a form of payment, just to be clear.

Digital Asset Controls to Mitigate Risk. No! Control Mitigation? No! Controlled Risk? Better.

The challenge how this can incorporated into a legal and regulatory environment that does not recognise such a concept is what makes it so challenging, yet it is driving the efficiency of blockchain.

“For example, tokenized money market funds could be used for payments, enhancing capital efficiency for holders. Such products are offered by incumbents BlackRock, WisdomTree, and Franklin Templeton, as well as Web3 natives Ondo Finance, Superstate, and Maple Finance.”

Some are open-end mutual funds registered with the SEC under the 1940 Act, for example, WisdomTree’s Government Money Market Digital Fund (23 October 2023, revised 20 November 2023). And the prospectus clarifies:

“Purchases and sales of Fund shares will only be processed on days and during hours that the Fund is open for business and you must complete your order prior to 1:00 pm (Eastern time) to receive that day’s NAV.”

And they say something important in a model where all buys and sells are undertaken against the issuer since MMFs typically do not have a secondary market:

“The Fund does not monitor for market timers or prohibit short-term trading activity. [..but such activity could require] the Fund to hold more cash, disrupt the Fund’s management, increase its expenses, transaction costs, administrative costs or taxes, and/or dilute the value of Fund shares held by other shareholders.”

In short, MMFs are liquid as long as the inflow and outflow are kept in balance, not with huge swings in AUM size. Fund would need to be more restrictive on the assumption that the use as a payment instrument could imply receivers who don’t want to stay invested would redeem. And look at consumer spending patters, they are cyclic: we tend spend more during Christmas. This would all need to be baked into the design to make this robust.

And this, of course:

“WisdomTree Transfers prevents transactions between unknown persons or unknown blockchain wallets.”

So, such a fund could be used for payments if the person you want to pay is a client of WisdomTree, has downloaded an app called “WisdomTree Digital” via the Apple App Store or Google Play, successfully completed the onboarding process, and once WisdomTree changes the fund structure allowing for transfers as opposed to a primary shares’ issuance and redemption process today.

Some of the other funds are not regulated money market funds but unregulated liquidity funds using Reg D exemptions with sometimes minimum investments of $5 million USD, and strict restrictions with respect to eligible investors and holding restrictions, and no ability to transfer these instruments to anyone not onboarded into this setup.

I would say that it would be a ‘false’ conclusion, at least under the current conditions, to think these could be payment instruments.

The Risk of Extinct Asset Managers

A final comment and another example of why this is unlikely to work is as follows. Different MMFs are geared towards retail or institutional investors or have different share classes that come with different fees. If you send me your tokenised MMF shares as payment, let’s say the Aberdeen Standard Liquidity Fund (Lux) – Sterling Fund, the ‘Archaeopteryx’ MMF token, but not on blockchain, but on Hedera, let’s not dwell on it:

Investing this way would buy you an interest in this:

• Share class L-1 (Income GBP)

If it has an L-1 share class, what are the odds it has different classes? This fund also has these (abrdn Annual Report and Accounts for the year ended 30 June 2023):

• A-2 GBP

• I-2 GBP

• I-3 GBP

• J-3 GBP

• K-1 GBP

• K-3 GBP

• L-1 GBP - ARCHAX version

• L-3 GBP

• X-1 GBP

• X-2 GBP

• Y-2 GBP

• Z-1 GBP

• Z-2 GBP

• Z-3 GBP

What does that mean? If you invest in the L-1 sold via Archax, the abrdn management fee will be 0.10%. This would suggest there could be different management fees in different share classes. Intriguing idea. And the fact that L-1 seems to be in the middle, perhaps this says something about the relative price level.

A-2 would be 0.42% and Z-3 would be 0.01%. It's shocking, I know, but there is a big injustice here, almost as ‘frivolous’ as airlines charging more the closer the date of travel or Starbucks handing out a free coffee once in a while to loyal customers who bring a lot of business.

Let’s assume we tokenize, and I was a “Z-3,” and you were only an “L-1” and you need to pay me, then your L-1 MMF token is no good for me. Instead, I could offer you a trade. I could sell you half of my ‘Z-3s,” but you need to pay me in real money; this way, we could arbitrage the abrdn fund rules and get you into a discount that you are not eligible for otherwise. Such a scenario would either bankrupt abrdn, or payment would need to be restricted to transactions between holders of the same share class (that’s trial and error to find out because that info is not on blockchain) or my identity and status of share class holder needs to be exposed to blockchain, or the business and distribution model needs to change, and asset managers need to ask themselves how to differentiate once price discrimination has been taken off the table as one way to differentiate.

Have they thought about this? I hope so.

Why are only FMI to gain financially?

“[..] 24/7 instant settlement and composability provide better user experience and new revenue sources for FMIs.”

FMI would be financial market infrastructure such as DTCC or Euroclear, but I am not sure if this implies new revenue opportunities only for FMI and, if so, why. They also write this a few paragraphs before:

“Tokenizing creates new possibilities [..] for financial institutions engaging with the new financial market infrastructure (FMI).”

This would be FMI other than Euroclear and DTCC. And banks could hope for ‘possibilities’ of making new FMI friends, and FMI’s write a bill. There is no explanation here why and how they are specific about FMI, and considering that Broadridge or MMFs are not examples of FMI activities yet are used as evidence to say tokenization is taking off, I must say I find it poorly argued and confusing.

“While the overall demand for tokenized money market funds partially depends on the interest rate environment, it now acts as the early green shoots of traction for other funds.” They cite a blog article from the US government's Office of Financial Research.

This needs a few comments. McKinsey's summary seems to have a formal error, as it combines two points that are not directly related to the evidemnce. The first part of the statement correctly addresses the impact of the interest rate environment on the demand for tokenized money market funds (MMFs). However, the second part, suggesting that tokenized MMFs are early indicators of traction for other funds, does not seem to be supported by the specific evidence provided in the detailed analysis. This is not acceptable.

Breakdown of MMF Inflows:

Total Increase in 2023 of $1.2 trillion

Inflows Due to Bank Stability Concerns: $480 billion (40% of Total)

• Retail: $96 billion (20% of $480 billion)

• Wholesale/Institutional: $384 billion (80% of $480 billion)

Inflows Due to Interest Rates: $720 billion (60% of Total)

• Retail: $504 billion ($600 billion - $96 billion)

• Wholesale/Institutional: $216 billion ($600 billion - $384 billion from stability concerns)

The data reveals a slightly different story. While retail investors were significantly influenced by higher interest rates, institutional investors' inflows were more driven by concerns over bank stability. Institutional investors allocated $384 billion due to stability concerns, representing 64% of their total inflows, while $216 billion (36%) was driven by interest rates. The article cited by McKinsey also confirms exceptional circumstances driving peak: “The pace of MMF asset growth, however, decelerated in subsequent months after regulators intervened and risk levels subsided.”

The statement made by McKinsey is not so problematic per se, although it is a bit simplistic, but the evidence given and the citation error make it unacceptable. The detailed evidence shows that retail and institutional investors had different motivations, which were not fully captured in the summary. Such misrepresentation diminishes the reliability and quality of the analysis.

Just for your information, here is why it is simplistic to say rising interest rates are good for MMFs without qualification, i.e., assuming, for instance, bank deposit rates are not increasing in line with rising market rates and thus creating a relevant delta between interest on bank accounts and MMFs. If not, rising interest rates should not impact the preference between holding cash or investing in MMFs. Here is how WisdomTree’s Digital MMF assessed this risk:

“There is a risk that interest rates will continue to rise, which will likely drive down the prices of bonds and other fixed-income securities. Changing interest rates may have unpredictable effects on markets, may result in heightened market volatility, and may detract from the Fund’s ability to achieve its investment objective.”

McKinsey, please let me know if you want me to mark you down for an error in market sizing for 2024 or 2030.

We know from before that McKinsey says there is $1 billion in total value in Q1 2024 for tokenized money market funds. Even though not all are MMFs, I know what they mean.

They now say:

“Total tokenized market capitalization could reach around $2 trillion by 2030 (excluding cryptocurrencies and stablecoins), driven by adoption in mutual funds, bonds and exchange-traded notes (ETN), loans and securitization, and alternative funds.”

Last year’s paper said this, referencing research from Citi:

“Analysts have forecast that $4 trillion to $5 trillion of tokenized digital securities could be issued by 2030.”

They consider $4 trillion now in a bullish scenario, and their own number is more conservative than the $16 trillion number promised by the FMI risk cartel and its adoption service. They mention some stuff about cold start problems and other blah blah including random things like this:

“We believe tokenization feasibility is highest for asset classes with lower technical complexities and regulatory considerations.”

Is mutual fund processing and ETF or securitization of lower technical complexity or with low regulatory implications or lower than equities? They come with their unique requirements. What do they really mean by complexity? Complexity does not mean it is unstructured. Scale makes a difference if a process needs to be automated. Some don’t until one has finished more pressing matters, as you expect. And in defence of slow moving banks, they thrive in complexity but not in decision criteria that create contradictory output and thus an inability to manage risk or say anything about expected returns.

“And finally, high-profile successes or failures could propel or restrict further adoption.”

So we all have a responsibility here: not cheering every new venture but really thinking about which critical aspects in market structure, definitions as mentioned before, etc., need addressing before there can be any hope for meaningful growth here. And don’t think I am negative at all: I am good at such things that require sussing out the impact of a variable change in complex systems. It’s like designing a new system, which I have done a few times in my life.

A user presses a button, and now what:

Hi Sup? 

Did you mean to push this button? Should I care if not? Nah!

Huh, who says you can tell me to do this? I wish I could use my protocol “remove user role entitlement” more often. Huh!

Huh, what now? How would I know the answer when the back office is not sending data? I don’t know when I will know. Yes, I can see in my transaction history I always get my info after 8 am when the humans come back. So? This might be just consistent coincidence? 

Ok, cheers. Your request to instruct is hereby noted. Whatever. Don’t humans understand what a closed back office means?

Fine, here is the answer to your query: “Instruction Status: Pending.” Happy? Why is this human still logged in?

Let’s show this user a delightful video clip of how tokens fly through cyberspace, and I am sure they’ll think that’s theirs and make them go away.

Excellent, it worked. My workflow is on hold. Time for a break!

Tokenization is the same thing but in a virtual blockchain environment. In my head, I often think looking at tokenization research: Invalid or even Failed Instruction causing an error in my brain’s ‘smart contract execution’.

Coming back to the $4 trillion estimated market size for tokenized assets:

“Our estimate is exclusive of stablecoins, including tokenized deposits, wholesale stablecoins, and central bank digital currencies (CBDCs) to avoid double counting.”

I think it means the included tokenized deposits are excluded, although the reason given makes no sense as I thought the plan was to estimate market value and not transaction size. But since they are excluded, there is no need to remind McKinsey that bank deposits generally cannot be transferred. Cash deposits are non-transferable tokes or ‘NTT’. But my invention needs a different name cause I already start hearing the jingle from TV ads for [warning this video can cause brain freeze] NTT DoCoMo.

Then comes a graph with the 2030 market size and says 1.9 trillion. Adding up the numbers only gets you to 1.8 trillion, which is, of course, less than the stated total value. Probably due to rounding, but I hate it when this happens. There is an interesting methodological inconsistency in this whole thing that makes me always wonder if consultants understand their fancy models.

They say they are afraid of double counting, hence excluding cash, which makes no sense. But earlier on, they claim that “tokenized money market funds [..] surpassed $1 billion in total value in Q1 2024.” I am doubtful that the existing products are attractive enough to grow to $400 billion, as McKinsey suggests. But let's assume it is, then $1 billion is ‘wrong’. They give a source, tokenized treasuries rwa.xy, as of 20 May 2024.

Amongst others, it includes (as of 23 June 2024):

• BlackRock Digital Liquidity Fund (BUIDL): $481,422,165 AuM - 14 (!) investors (addresses)

• Ondo USDY LLC: $279,095,19 AuM with 4,677 investors (addresses)

At least 45% of BUIDL is held by Ondo who simply moved their existing assets they previously held at BlackRock into the new fund (this sequence is only an assumption based on the fact they used BlackRock before). This would mean a form of double counting!

And you can’t be picky about such a problem unless you are constantly picky.

Then comes more rather unconvincing stuff about how tokenization benefits different assets but none of it bears any tangible relation to how or what is required to make that happen. And sentences like this:

“Having data on a shared ledger reduces errors associated with manual reconciliation.”

Maybe, if the data is correct and can be correctly interpreted. How tokenization helps when, for instance, the existing fund tokenization models increase reconciliation needs by adding more ledgers but not making relevant data available is unclear to me. Plus, automation should always help to reduce errors in manual processing. Don’t imagine that when a banker says, “we do something manually,” this means pen and paper. It means that when the computer finds errors, they need to be investigated, or certain reports are very specific and require data from domains that are not exchanging information. Things like that. And that’s a big part of the reconciliation issues, and tokenization does not automatically resolve that.

They are citing “embedded compliance (for example, transferability rules encoded at a token level)” as great. Personally, I think it is highly undesirable and unworkable, but I will write about this another time.

They do love Broadridge’s repo solution, and in their write-up, they confirm my notion that this has nothing to do with tokenization:

“Unlike some tokenization use cases, repos do not require value-chain-wide tokenization to realize material benefits.”

This is funny in a way, creating a novel strategic insight instead of saying: a workflow manager using blockchain is not tokenization. Again, I am not saying it’s bad, but the fact Broadridge uses a tech stack they call blockchain doesn’t change much compared to a more conventional technology choice they could have made. That’s all. And what’s value chain-wide? And later, they say in relation to private funds:

“To fully realize the benefits of tokenization, however, underlying assets must also be tokenized.”

So why does Repo get away without it, but alts don’t? All very unclear.

McKinsey’s Proposed Solution

Earlier on, they set out this objective for themselves:

“We examine the ‘cold start’ problem and offer practical steps for how it may be overcome.”

By cold start problem, they refer to a need that successful innovation requires providers and users to grow—there needs to be demand for supply! Ok. Is there more?

“While there are billions of dollars of tokenized bonds outstanding today, benefits over traditional issuance are marginal, and secondary trading remains scarce.”

Two questions: how does McKinsey know what is outstanding or what is traded? They don’t, or else we would have seen the numbers. This already hints at one of many ‘false’ design choices.

The other is: what are the marginal benefits? Right now, there would be material disadvantages in investing in such a construct.

But here comes the solution:

“We posit [!!! ok] that a minimum viable value chain (MVVC) (by asset class) is required to enable the scaling of tokenized solutions and overcome some of its challenges. To fully realize the benefits laid out in this article, financial and partner institutions must cooperate on common or interoperable blockchain networks.”

What do they mean by this?

“Examples of MVVCs are the blockchain-based repo ecosystems operated by Broadridge, and J.P. Morgan’s Onyx with Goldman Sachs and BNY Mellon.”

That’s kinda it. J.P. Morgan has a nice marketing budget, and Broadridge doesn’t have much to do with tokenization. We overcome ‘cold start’ problems by a blockchain that doesn’t have them and call it MVVC?

Is that it? Seems so.