In my previous article series - When Data becomes a Financial Asset Class - I discussed the problem of data valuation. Here's what I wrote in this regard:
In a world where Data Access is a commodity, an ecosystem dedicated to the continuous evaluation and pricing of that data will emerge.
Even if a small fraction of daily generated data (~400+ Exabytes as of 2022) finds its way out, we will have a deluge of Data Assets coming on to the market. As a consequence, formal Data Asset quality estimation and valuation techniques will be developed.
Data Asset valuation will be based on 3U's: Uniqueness, Utility and Usability. Expect these qualitative metrics to be quantified in the coming years.
In the days that followed the publication of that article, I became aware of Craig Danton's work on the same topic. Here is how Craig defines his framework that I'll refer to as fINTUC for the sake of convenience.
Value ∝ F(I, N, T, U) — C
I — amount information (or insight) about an entity (event, person, place, company, etc),
N — the number of applicable entities this information pertains to,
T — the length of the predictive time horizon,
U — the uniqueness of this data source,
C — the cost of the data to acquire as it reduces the total value.
There's a decent amount of overlap between the 3U framework and fINTUC. Both feature Uniqueness as an attribute. Utility can roughly be mapped on to the I, N and T in fINTUC.
Here's a collection of my thoughts on the problem of Data Asset valuation in general and the two frameworks above in particular
- It might be useful to distinguish between two types of valuation frameworks one might apply to the problem of Data Asset valuation.
The first is the type of valuation (let's call it Indirect Valuation) that goes into conventional equity valuation frameworks such as Discounted Cash Flow (DCF) analysis. Here, the valuation of an equity is a function of the free cash flow produced by the underlying entity (the company whose equity is being analyzed). Note that the thing being valued (the equity) is different from the thing that is producing the cash flow (the company).
The second type of valuation (which we can call Direct Valuation) is the subject of mine and Craig's 3U and fINTUC frameworks. This is the type of valuation that seeks an objective dollar valuation for a Data Asset. - The dollar value that Direct Valuation seeks to establish seems analogous to the concept of a Book Value of company assets in finance. The book value is the difference between the total assets and the total liabilities of an entity. If we mapped as follows
Total assets <--> Value of Data if Base Rights/IP were sold
Total liabilities <--> Cost of acquiring Data
the analogy (especially with fINTUC, that has a -C to account for cost of Data) holds up well. - We still have the problem of establishing the monetary value of the Base Rights/IP objectively. In the Book Value analogy, you start with a fixed price asset and depreciate it every quarter at a given rate. The depreciation rate differs for different assets. Land typically has a negative depreciation rate while furniture has a positive rate.
The fixed price of the asset at time t=0 is necessarily a function of how much the market is willing to pay for it at the time. For instance, the appraised value of land is not determined arbitrarily, but has a strong correlation to the market value. - We are in a bind, because we started off wanting a pricing mechanism that gives us an objective value of the Base Rights/IP of a Data Asset, but instead, landed up in a place where this is dependent on market forces. Let's bookmark this insight for now.
- In Indirect Valuation, the price (P) of equities transacted in the stock market is related (or more precisely, should be related) to the price of goods and services sold in their respective markets (let's assume free markets :)). When the latter results in a net profit or earning (E) the P/E ratio (current and future) becomes a valuation metric.
Assuming that we aren't talking about Base Rights/IP, but ongoing monetization, what is a good measure of E for Data Assets? - It's tempting to conclude from the above that "You need a market to solve for the market". In other words, there is no fundamental value except that which is determined by the market. But we know this shouldn't be true. If there isn't a market bid for a Data Asset, does it mean it's book value is $0?
- Perhaps two markets instead of one is a solution? This might seem obvious, but the markets that price the goods and services that are sold by a company are different from the equity markets that decide the stock market valuation of the company.
By the same logic, having one market for Base Rights/IP and the other for access/use of the Data Asset seems like a place where we can stand on with respect to valuation. Let's examine this in more detail. - To put it more concretely - Imagine you had two Datatokens. An NFT representing IP (Let's call this T_a), and a ERC20 (or equivalent) Datatoken representing transaction (Let's call this T_b) and usage rights. Let's examine two price scenarios:
A. The total value of all T_b exceeds T_a. This is analogous to the market capitalization of a company exceeding its book value. Implicitly, the ERC20 token market is assuming that the IP will be used to generate "earnings" going forward. In other words, the IP will be put to productive use and that this productive use is more valuable than the Base Rights.
B. The total value of all T_b is less than T_a. In other words, the ERC20 token market is pricing the IP higher than future earnings. The market does not expect this IP to be of significant future use, but the base IP itself is priced at a utility value determined by the NFT Market. - By measuring their relative valuation versus IP, we have at least one way to measure over/undervaluation of the equity portion (usage/trading datatokens) of the Data Access Market. The valuation of the IP itself remains unsolved. From the discussion above, though, it seems like a dollar value determined by a fundamental framework still needs some sort of market pricing mechanism.
- I'm tempted to extend the analogy in point 7 above further. Take the example of land used in Book Value calculations, for instance. The market that prices land is different from the market that prices goods and services. Both of these are different from the market that prices equity.
The end-consumer of the land will price it differently based on their usage scenarios. A farmer will quote a different price for the same piece of land that a retail mall developer is prospecting. Both might bid for this land, which might have a base-price and/or a minimum quoted. The land eventually gets sold for the highest bid, which is usually higher than the ask.
The bids are both a function of market expectation, ask price and the buyers need. The buyers need in turn is a function of the availability of land and its suitability for the use case. This also allows us to make a roundabout connection to the 3U/fINTUC frameworks. Availability seems analogous to Uniqueness/scarcity and Suitability to Utility and Usability.
We'll need to dig in to map this out in more detail, but I suspect this is already firmer ground than where I started from. - The really nifty part about NFT marketplaces is that the auction process is the default transaction mechanism. The problem of course is that a massive speculative premium or discount will still be baked into the auction price. Even so, if a Data Asset is truely useful, there should be a bid that reflects the subjective utility for the bidder. In this respect, the "fundamental" valuation problem is still unsolved. Ultimately, it may be that there is no price independent of the users judgement.
- What might be uncontestable, though is the cost of creating/acquiring the Data Asset. Amortized electricity and storage costs, labor costs (Data engineers, analysts, etc) should give a lower bound for the cost of Data Asset acquisition. In this sense, we might have a Under/Over valuation basis for Base Rights/IP.
- Let's recap.
I outlined two Data Asset valuation frameworks: 3U and fINTUC. Based on analogy with mainstream finance, I outlined Direct and Indirect Valuation methods. Direct Valuation is analogous to Book Valuation, while Indirect Valuation is analogous to DCF. Book Valuation may be valued fundamentally, but is ultimately a function of market forces.
Book Value, Earnings and Market Capitalization-derived Value are all determined in different markets. For Usage/Transaction tokens (equivalent to equities), a potentially useful Under/Over valuation framework can be established by comparing it to the value of the Base Rights/IP, represented as an NFT.
Auction-based NFT markets indirectly incorporate user preferences, needs and willingness to pay quoted/bid prices in exchange for desired Utility, Usability and Uniqueness. Data acquisition/creation costs are a useful lower bound for establishing a Under/Over valuation framework for Base Rights/IP.
Acknowledgments: Many Thanks to Craig Danton for reading drafts of this article and providing feedback!