Anything by Paul Mcglone is a top read on pensions. Paul has an understanding for big data, the power of using it right and the dangers of getting it wrong. Here is an outstanding article that explains why Chris Curry and his  dashboard delivery team have been handed the toughest of job.

If they get it right, no-one will thank them, if they get it wrong, they could become objects of public opprobrium.  This is why they need our help and why we must be as diligent in making data available as the dashboard infrastructure team in providing the search functionality.

My favourite topic of pensions dashboards again. I was asked this week how small the matching error rate has to be for dashboards to work well. So here are some numbers …

Let’s say we have 25m savers with 4 pensions each (this is probably low), that’s 100m records sitting on the dashboard somewhere

What happens if 5m members use it in the first year?

Each of the 5m members is checked against the 100m records
That’s 500,000,000,000,000 (500 trillion) checks a year

What we HOPE will happen is:
– 20m return a positive match (5m members x 4 records each)
– 499,999,980m return a negative match
But matches will not be perfect

There will be some cases where the member DOES have a pension but it isn’t found.

If a search is 99% successful at finding your records then you get:
– 19.8m matches (member correctly told that they have a pension)
– 200,000 cases where a member can’t find a pension that is theirs

---

Missing 200,000 records is clearly not ideal, but it’s also not terrible.

But what about the other way round? 

The other “almost” 5 trillion searches that should return a negative match. Even if schemes get that wrong just one time in a million then we get:
– 499,999,980 cases a year where a member is told they have a pension when in fact they don’t!

Put another way, a typical member would be given details of 4 pensions that really are theirs, plus 100 more that are not theirs at all. And members may not be able to spot the real pensions amongst the false ones.

To be clear, I DO NOT expect this to happen.

The point is that even an error rate of 1 in a million isn’t good enough. For dashboards to work the error rate needs to be something like 1 in 100 million, or even less. That’s why getting the data right and getting the matching rules right will be so important.

As an aside, I often get asked why pension are to hard to match when open banking works.

The point is that open banking works by me telling my bank where to find my other accounts. It does not involve my bank sending a message around every bank and branch in the UK asking them to check every account they have to see whether it’s mine

---