This is the continuation of the part 1 that can be found here.
3/ The phoney claim of efficiency
Some evangelists and some serious companies pretend that blockchain will save businesses “billions of dollars”, even though, by design, distributed systems are more complex and more resource-intensive than centralized systems. So, let’s deconstruct the bogus claims.
Gains are overestimated
The potential gains are usually over estimated: up to 50% and even 70% savings based on the full cost base . Even in the blockchain utopia of universal bookkeeping, one can never get rid of all controls, investigations & reconciliations. Organizations will always have to deal with garbage and humans. Besides, some reconciliation costs are unavoidable, especially internally where the front, back and accounting visions are different by design. It is a feature and not a bug. More generally, entire functions are segregated to avoid conflicts of interest: front functions are separated from middle & back offices, the buy side of the bank from the sell side, etc. Externally, it is true that banks have to reconcile transactions “1 to many”, but the purpose of central third parties is precisely to reduce this complexity to “1 to a few”.
More generally, it would not be possible to use a single general-purpose ledger because financial products are too different. Banks would need to run several ledgers and to reconcile them with the global ledger of the organization. You would also have to exclude a very significant part of the business which is about Lending and Financing, because it does not square at all with the original use-case for a blockchain: the controlled transfer of ownership of digital assets under a controlled supply law (preventing double spending)! The concept of money multiplier is a pillar of the economy — now square that with blockchain!
To conclude, I can tell from experience that 20% savings is rarely achieved even with proven process re-engineering methodologies (because the manual processes have already been outsourced for years). Now imagine with a more complex technology such as blockchain!
Finally, if we sum up the 20–50% gains promised by the “experts” for each new technology such as blockchain, robotics process automation, internet of things, cloud technology and artificial intelligence, then banks can save between 100% to 250% of their costs.
Costs are omitted
The claims often focus on savings without assessing the costs. The running costs of a decentralized system such as bitcoin are high. In a business context, the running costs of a pseudo-blockchain would be lower, however the transformation costs would be gigantic. Imagine having to re-architecture how the entire financial markets trade & settle dozens of financial products! 
Let’s do a cost analysis on a simple case, a decentralized version of Airbnb compared to the original, everything else equal (functionalities, user experience, etc.):
- Development costs would be higher due to the expertise required and the need for deeper auditing and testing
- Sales & Marketing costs would be identical (same number of customers, …)
- General & administrative costs would be slightly lower if the team is decentralized
- Help desk & customer support would be identical (same service level agreements)
- Infrastructure costs would be much higher, from x2 to x100 including the costs of external validators, depending on the efficiency of the consensus protocol
This simplistic model only illustrates the fact that a centralized system is generally more efficient than a decentralized one. 
The claim that “a blockchain is desirable when the existing cost of trust is excessive” sounds reasonable but once again it is unfalsifiable. Because decentralization is generally less efficient than centralization, it is on the people who make such claim to demonstrate their claim with a sound analysis of what is the exact “cost of trust” they are addressing and to compare it with the cost of implementing a (pseudo)blockchain.
There are always some trade-offs
A recent research paper  demonstrate in a scholarly fashion what we know from common sense:
- A system cannot be at the same time correct, cost efficient and decentralized
- Compared to centralized ledgers, POW blockchains reduce monopolistic rents but consume more resources & introduce miscoordination inefficiencies
- Both incentivize honesty, but of course in different ways (centralized ledgers want to keep their rent, public blockchain use POW algorithms)
- Pseudo-blockchains don’t eliminate rents
- Pseudo-blockchains are on the same side of the triangle as centralized ledgers
Pseudo-blockchains don’t eliminate rents
The Darwinist argument
Public key infrastructures (theory from 1976–1978, RFC 1422 in 1993), hash trees (Merkle tree in 1979, SHA-2 in 2001) and peer to peer protocols (Napster in 1999) have existed for decades. Satoshi combined those elements together with clever financial incentives to specifically solve the problem of double-spending for peer-to-peer electronic cash. A blockchain is not a new technology, it is a new design. If those tools could be used to make more profits, they would have been implemented by banks a long time ago, instead of developing architectures based on trusted third parties.
A blockchain is not a new technology, it is a new design
If a blockchain is not a cheap panacea, the interesting question to ask is the following: when do the benefits of a blockchain are worth the extra cost?
This is the end of part 2— continue to read part 3 here.