Platform Banking: An Email Thread

An old mail thread from about 6 months ago between myself and Antony Evans.


Amusing video below:

Do you think some of these issues are faced by banking consumers?



I think that the internet has altered consumer behaviour a lot, but it's much more that consumers now talk to each other in public. Providers can see what's being said about them directly. (As can anyone else.) I don't really want to have a personal relationship with my bank. The bank does because they can hassle me with cross-sales promotions. But I'd just like them clear my cheques on time, etc.

What I do like is for them to be transparent and explain themselves. From personal experience, financial service customers get VERY upset about "missing" money, but if it's just stuck in processing, or a BACS window got missed, they don't really mind.

I like comparing the financial industry with the web hosting industry. Hosts are forced into a very open world by all the forums that cover them. But the end result seems to be that most sophisticated buyers prefer the utility model of places like Those hosts seem able to grow mostly by good word of mouth, but they only offer a linux "box" with no support. People want a simple well defined thing that "just works". Hosts who offer a more personal touch, openly resell the services of one that does "just works".

Similarly, the really big winners in retail finance have been reliable no-frills online stockbrokers and dumb platforms like PayPal. It's easier to do a simple job well, make money on volume and rely on word-of-mouth to bring in new business. Commodity business are sexy for technologists!

Really, my bank account is a Visa/LINK account far more than it is a Nationwide one. If I fly to the US, I can just wander over to a cash machine and see my account balance in $'s. I probably would consider using a bank with no physical presence and no investment arm. [Note: I've since met Dutch people who bank back home, so they effectively do this.] To me, Nationwide's sole purpose is to tie together BACS, Visa, SWIFT, etc over a regrettably small store-of-value account. Nationwide doesn't make much money on that. They probably make most of their money selling loans, insurance, mutual funds and so on.

Banks have a real issues stemming from being semi-monopolies with dire customer service, but much more by it becoming easier for customers to divest them of their profitable sidelines. Perhaps the solution for the is to do less and simply act as a channel for PayPal et al. Of course, it could go the other way - E*Trader set up a bank (which did almost kill them...)

*ech* The site I'm might be building is getting cold feet and wanting to do a revenue share. Do you think I could become a "Thought Leader" instead? Wink


I really like your comparison of the financial services industry to the hosting business.

Essentially you are proposing that somebody ties together an extremely basic network/account and then other players provide services on top of that.

Customers can decide which services they want/need and which they don't. How complex do you think it would be to develop a really basic 'platform' service like that?

Maybe it would be helpful to summaries the list of issues in the current banking world. Then we can see how such a service could address each of these issues.

  1. Customers dislike/distrust of big banks
  2. Separation of risk ownership from loan origination
  3. Complexity of products and processes


The hosting business analogy is a really useful one. They are (abstractly) extremely similar business. There are even "problem states" that "launder" spam and warez traffic!

Trying together basic services is pretty much what Egg did. Most of the retail side distributes at least some other people's products. At Morgan Stanley, the derivatives desk used to underwrite all the "110% of FTSE 100 growth" ISA bonds for people like the Post Office. (I think the Post Office ones are done by RBS though.)

So in a sense, the change to a platform model would be a brand change. Instead of hiding multiple white label providers under a single brand, you would be co-labelling a competitive range of products that you only take limited responsibility for.

It would be far more effort than you'd initially imagine to build a bank platform. But it is do-able. I think they'd be a few junction "services" you'd need to bundle into a "cash nexus".

  1. Connectivity - Not just a website, but 2-factor fobs, snail mailing statements, SMS notifications, spotting dodgy IP addresses...
  2. Identity - It's legally impossible to act for an unknown party. You might also need to check qualified investors status, FSA bans lists, etc.
  3. Store of Value - Somewhere to keep $, EUR, £, etc. Any commercial paper manager could create backing for the client accounts, but you would still need records.
  4. Notary [2] - Have to attest to contracts between users and providers to make them enforceable. Requirement could vary from a web button to a visit to an affiliated public notary.
  5. Settlement [2,3,4] - SWIFT, BACS, etc transferes, and clearing arrangements with product providers. Transfers need to be linked to notarisations.
  6. Fraud cover [5,1] - The platform should resolve & "make good" fraud. It "sees" everything, so it should have the lowest cost in resolving frauds.
  7. There's possibly a (7) which would be attesting the user's income streams to providers. Eg. "We say this person is receiving XXX $'s of income per month from IBM. Signed, The Platform".

[X,Y] == dependent on X and Y

This would form a wrap account for retail banking, similar to the ones that are available for investments and pensions. Providers of credit cards, loans, spread betting, etc would apply to join the platform. Users would choose and interact with them through the platform's channels. There's a good precedent for this in the asset management world with things like FundNet. And Tesco Finance is wrapped RBS. [Note: Tesco have now bought out RBS.]

On the downstream, a boutique bank could resell the platform with a "lifestyle" services and tax advice. Or strip the phone support and investment features to create a free service for the poor (prob. funded through selling leads for high margin products)..

Regarding the issues it would solve:

1. Customers dislike/distrust of big banks

To a large extent, people dislike banks because they want to not pay for things, and because banks say 'no' to people who want money. And they distrust, because banking is fairly abstract concept. People will feel this way about any middleman *shrug*

But banks do have a genuine agency problem. Often they're trying to both advise, broker and create their products. A bank will review your finances, decide you need a loan to "invest in your future", sell you the loan, underwrite the loan, fund the loan, advise you to take PPI on the loan, and underwrite the PPI on the loan. Cross-selling is inherently dangerous in this way. On the upside, integrated banking, can potentially cut costs *massively*. Especially marketing costs. (The key to success for a personal loans company is actually cost-effective marketing.)

A pure platform "bank" would not have loan/insurance writing, pension (mis-)selling functions, so would operate with reduced incentives to evil. I'd say reduced, because they could still engage in "churning" client accounts, plus all the other bad broker behaviours. (A platform might choose to adopt a binding voluntary code to show its commitment to goodness?)

Competition between platform would probably be keener too. Assuming that every major provider is available on multiple platforms, you could simply instruct each of your providers to transfer your products to the new platform account.

2. Separation of risk ownership from loan origination

Oh - a platform would make the separation worse! But I'm not convinced that separation itself is the problem. (Or securitization.) It was a lack of oversight. If you have Platform->Originator->Owner, the owner can check basic facts with the platform. It's a small improvement, which will still fail miserably if people just do not care.

3. Complexity of products and processes

This is where a platform would shine. You could literally click-to-refinance: the platform could pre-pay one loan from the money raised by another. In the current system you'd need to have bridge financing, and the new provider will assess your debt twice. They'd be no problems doing instant ISA withdrawals - the platform "remembers" the money going in, so it can advance it back to your current account risklessly (no more 3 day BACS). My Mum's spent the last month trying to move between different Nationwide ISAs, so this is a huge "normal people" benefit.

For the less normal, the platform can offer all the investment facilities of a wrap account: pretty tools, backtesting, data feeds, etc. With due [read 'extreme'] caution, Facebook-style widgets could even be installed by the users themselves.


I like your description of the 'cash nexus', although I'm not entirely sure you need all those services at launch - I'm a big fan of the Y-combinator model of getting a basic product out the door fast. The way I would envisage this working is as some kind of open source model, a bit like Ruby (or Insoshi). You need a core team who develops the absolute minimum services to get it launched and then grow a community around this core who develop additional features and functions. Features like ID or income verification would be ones I would see the community develop when they need to.

To me the core platform is really really basic to start with. Just a website (or maybe facebook app) which brings together a store of value with a payments mechanism and then an API system which lets community users develop add-ons and modules (this would be the hard part to develop). You could use ripple pay (already open source) for the stored value and paypal or moneybookers for the payments system (at least initially). Some kind of encryption to uniquely identify and validate transactions between parties. Fraud control you build in later (as you point out control of the data enables that). One key challenge is the legal contracts, but the beauty of this approach is that these can be outsourced to the community on an as needed basis - although for launch it would probably make sense to tackle this in one or two key markets (UK and US perhaps?).

We could wrap the whole entity in a Limited Liability Partnership, which provides incentives for developers to create modules as well as a structure for raising finance.

And I think it does solve the risk separation issue, because the way I see this working is that investors (i.e.. those holding the risks) would use the platform to directly access the customers, completely dis-intermediating the banks - the investors would be the ones taking the decision to originate the loan or not, rather than the platform. This could be quite viable if the right tools were developed to reduce the costs of credit monitoring, and this is the way I think we are going to see the market evolve post-credit crash (this is why credit is going to be more expensive in the future - have you heard the term 'peak credit' being bandied about?). It is this feature, solving the agency issue, which would at the least help generate a significant amount of free PR around the platform launch - it's the killer app if you like. The way I see it working, this wouldn't appeal to private individuals like Zopa does, it would be somewhere that hedge funds and even banks can go to source loans - the banks focus on what they should always have been good at, people very good at assessing credit risk. The value to them is a low cost customer acquisition tool (though I guess customers would have high churn on the platform).


I was thinking of the Platform Bank as a logical end-point, more than something you could build now. I'd expect to see it arising out of Citigroup's break-up (it has to happen, there are smaller countries), or maybe in a virgin market like Iran (think mobile phones).

The Y-Combinator model probably doesn't work for financial firms: if you don't do ID for a money transfer service, you become a drug money conduit and get prosecuted! [Note: I'm not as scared by money laundering regs now, but they're still very awkward to deal with.]

I don't think hedge funds (or pension funds, etc) want to process individual loans in any way. (Actually, I know this...) They might well want title to them, and that would be the advantage. Really the subprime crisis is a human problem of too much money and lack oversight.

I can't see a "community" doing lots of work for you, unless there's something in it for them... what would that be. Who are they going to be, and what's the incentive?

(Also confused by the Ripplepay reference, because it's netting system for privately issued credit balances. There's no Ripple Account with $'s in it. For tech - I think your looking more at Oracle Financials and less at Ruby. Why does everyone want to build stuff in Ruby Smile ?)


There was a suggestion floated around Zopa UK that you could consolidate the loan process from the moneysupermarket quote through to the human underwriter, as shared infrastructure. You'd collect all the credit models from the loan firms to run against the borrower's record in parallel - then show the borrower the best quotes from the models.

When you pass the borrower's lead to the loan firm, they will come as a pre-vetted "package" with identity checks, pre-screened credit, valid payment details and a money-back guarantee (fraud).

Since you're providing a service to the loan company [credit scoring + identification] - you can charge a large flat fee to them. So no conflict of interest over varying commissions between loan providers.

Possibly not the best time for it now though... Islington waiters are talking about the credit crunch while they're packing the tables away :-/


It's precisely because Islington waiters are talking about the credit crunch that now is the right time to launch something. A few reasons:

  1. Clearly identified issues with the existing model, hence able to sell to investors that you are solving a big problem. Whoever solves this problem is going to make billions, at least!! Why can't it be a startup?
  2. Developing/proving the business model (and technical architecture) will take at least six months (whom I kidding, probably more) for the technicals and years for the business model. Thus by the time the market sorts its head out, you are ready to capitalise on the full upswing rather than getting to the party just as the famous people are starting to leave.
  3. 3) Tough times mean less competition. Any monkey can launch a website in a good market
  4. Investor caution means that the business model will be much more heavily challenged, therefore less likely to develop something with is a load of rubbish!!

Was thinking about ID verification... Paypal handles this very simply by making sure you are linked to a bank account - effectively they rely on the bank to do the job for you. I reckon you could even piggyback onto their verification through their API system for launch. [Note: I've now discovered this is common practice in Estonia.] I'm sure with some ingenuity that a Y-Combinator model could be made to work (though I think their model is so brilliant I'm probably biased).

I also remain unconvinced that hedge funds (and banks) wouldn't be interested in originating loans of such a platform. I used to work for two hedge funds and my experience suggests they will do anything which makes them money. The key challenge is that the absolute return has to be worth their time - a highly paid hedge fund manager isn't going to spend ten minutes reviewing a loan application for a small loan. He wants to look at big deals and ideally wants to be able to mark-to-market and maintain liquidity. They have less stringent capital requirements (mostly imposed by their prime broker lenders - they need to leverage up a loan book to make decent returns) so are competitively more advanced.

You are right about the technical language, people only want to use Ruby because it's so agile (and there is a reasonable community of users). For financial transactions you probably want a more fundamental language. I was only suggesting using Ruby to develop a very quick community site, to test to see if it is possible to build a community around an idea like this. I honestly don't know if you could do this, but the only way to find out is to try.