Financial-Industry Experience
For a leading European bank, I ran a project which allowed it to gain a comprehensive view of each of its customers.
This was something that the bank thought it had tackled with technology.
It had linked up its databases for:
- Will-writing
- Trusteeship and
- Sharedealing
and pressed ‘Go’.
The results were unreliable and extremely dangerous.
The reason was that the information wasn’t safe to exchange since it held nuanced, local meanings which had never been intended. And they were not even documented.
Here is an example of what happens when real people have to operate IT systems that cannot quite deliver what is really needed.
The above business areas will have IT systems that can or should record concepts like ‘do not mail’ and ‘deceased’. But Sharedealing had no concept of ‘do not mail’, as the request from a customer was never expected or catered for. The only way Sharedealing would stop mailing was if the customer were flagged ‘deceased’. Which seems like a reasonable work-around.
But ...
... if a customer of the bank decides to go travelling whilst renting out their home, they may well decide that they don’t want any communication about their wealth sent to their home address. Remember that bright end-users at Sharedealing — doing their best: don’t knock it — have a fix for this: they flag the customer as ‘deceased’.
And then, of course, the information is shared with Will-writing who, following due process, write to the next of kin — whilst the testator is goodness-knows-where on the planet — and commiserate with them for their loss ...
This clearly results in:
- A very distressed customer
- Perfect fodder for local and national press
- General anxiety about ‘data protection’
Putting it right
Fixing this problem — particularly in this environment — is tricky. The problems I faced arose because:
- The IT systems were old
- Because of bank mergers, systems and data had been grafted onto older ‘legacy’ systems; with all that that implies
- Contact with the customer of one of the sub-businesses was infrequent, and a delicate matter
- Infrequent mailing meant that address details were unreliable
Commercial data cleaning did quite a bit to help, but in the end we concluded that we had to go back to source — to the customer — and begin a long process of information verification. We developed concepts of ‘Trusted’ and ‘Quarantined’ data, and pools of information which could be used for certain purposes, but not all.
In the meantime we set up Business rules for the exchange of data, allowing only what was safe and verified to be accessible in the shared pool. It meant that there was a lot that could not be seen, but what could be seen could be trusted and safely acted upon. And we knew where to look next to build up information that the business could trust.
There was no quick fix here.
Indeed, the ‘quick fix’ had already been tried: ‘buy some clever middleware and another big data warehouse, and “you’re done”.’
That approach — ignoring human factors and business legacy — had been a disaster.