Ceci n’est pas une pipe

Those who know me in the FinTech community know that one of the topics which gets me into a fierce debate is the “banking dumb pipes”.

I feel that this term should be defined with more accuracy. Why? Because it wants to be a metaphor for the possible future of the banking institutions and it encourages endless discussions about how Financial Services will be delivered when banks will be “just dumb pipes”.

I came across the term of “pipes” within a banking context in a talk by Nadeem Shaikh of Anthemis at TEDxLondonBusinessSchool in 2012. Nadeem presented a number of start-ups and ideas I was quite familiar with from my work as an architect with the Financial Services Innovation Lab at IBM. However, the term “dumb pipes” (minute 3:45) took me by surprise. Having worked with systems and parts of architectural landscapes located in the area designed in Nadeem’s talk as “pipes” territory, my view was that they are not dumb at all and more importantly they were not pipes at all. I could not quite foresee where the complexity and functionality would be transferred when things will become “dumb”. Years later, at a talk at Finextra I restrained myself for civility’s sake from shooting “There are no pipes in banking!” and to this day Liz Lumley teases me with this phrase.

I’ll explain in a long-winded way what I mean when I say that there are no pipes in banking. In time, I hope to develop a shorter way of saying it.

The Stupid Network

The “dumb pipe” debate originates from telco industry and there is a lot of literature on the subject. The grandfather of the debate is David Isenberg who in 1997 published the seminal paper . Isenberg observed that the assumptions which stood at the base of the design of the telephony networks do not hold anymore. The “Intelligent Network” built under the old assumptions was being superseded by a “Stupid Network” which functioned in a new reality of user behaviour and technological enablement. Isenberg continued his argument in another paper by exploring how abundant infrastructure, underspecification, internetworking and user control led to the emergence of “Stupid Networks” which liberated a huge innovative potential in the economy. Sounds familiar?

The Dumb Pipe

Isenberg though does not use the term “dumb pipe” at all, and if you read his papers, it is not obvious to me that his “stupid network” can be interchanged with “dumb pipe”. I came across the dumb pipes term only in a paper published in 2006 by Bernstein Research and authored by Craig Moffett and Amelia Wong:. The paper also looks at the telecommunications industry and more specifically at the cable operators. Assuming that a cable company sells two services, network and programming, they argued that if the cable company would sell its network as a “dumb pipe” and would allow another entity to sell the programming, the cable firm would be more profitable. Without going in more detail, I am sure you recognise the concept of “unbundling” which is debated so much in banking nowadays.

While it is always interesting to look at other industries, unbundling telco and unbundling banking does not leave you with the same Lego pieces. More accurately you will likely end up having some “dumb pipes” when you unbundle the bank but this is fundamentally not what makes the bank, holds the bank, gives you the back office infrastructure.

The Pipe

Wikipedia defines a pipe as “a tubular section or usually but not necessarily of circular cross-section, used mainly to convey substances which can flow — liquids and gases (fluids), slurries, powders and masses of small solids”

Not only that what goes through a pipe can “flow” but:

  • What goes through the pipe it is of the same nature. You have pipes for water and pipes for oil not a mixture of the two.
  • The substance transmitted is not transformed through the pipe or by the pipe. If transformation processes occur they by definition happen in intelligent components connected to those pipes.
  • Any part/quantity of the substance transmitted is of equal importance and fungible. One litre of water which enters the pipe is assumed of the same quality and nature as one litre of water at the exit of the pipe.
  • Thresholds of acceptable losses are expected and they do not affect the end result or the processing. The losses are also definitive through leaks or natural processes (e.g evaporation)

When people talk about banking pipes the assumption is that the substance that flows through is the data. Data is the new oil or so people say. Indeed, data might be the new oil and it flows, however let’s try to see if the above criteria for the use of a pipe holds for banking data:

  • Is banking data of the same nature? One can claim that all data is ‘financial data’ however, it is not additive and thus cannot go through the same pipes. Current account transactions will be in separate pipes than cashflow buckets for financial instrument valuations and even from mortgage transactions. It is like the difference between natural gas Bluestream pipe and the distribution of lead free petrol to the petrol stations. While both are fundamentally fuels, mixing them is non-sensical. Sets of data will go through pipes as long as they are similar, additive, and the path is unchanged for a foreseeable future.
  • Is data at the beginning of a process the same as at the end? This holds true for limited but admittedly numerous cases. However, the business of banking relies on transforming the data into forms that can be used for taking decisions at the balance sheet and PnL level.
  • Is the financial data of equal importance and fungible? Definitely not. Transactions are unique in time, location, value and inextricably linked to an author and a merchant.
  • Is unaccounted loss of financial data through “pipes” allowed? In the physical world one could expect some losses of substances going through pipes, due to leaks, natural processes or measurement errors. As a bank however, it is unacceptable to say that you’ve lost 1% of the transactions through the pipes and things are fine. A bank cannot tell a customer that they’ve lost some of their transactions today but will send others tomorrow. They are not interchangeable.

Financial data is not fungible. Even if it flows it does not flow through “pipes” because it does not have the behaviour or the characteristics of a substance going through pipes.

In conclusion,

  • Can we talk about banking dis-intermediation, unbundling and banking as utility? Yes, not only are we able to do this but we must. These are very realistic scenarios.
  • Should talk about banking “dumb pipes” simply because we don’t know what is happening in the belly of a bank? No, because this is dumb.

Banking Systems Architect. Curious. Antifragile.

Banking Systems Architect. Curious. Antifragile.