In this blog, Ella Holloway explores how to build a robust data matching strategy as we move closer to the pensions dashboards deadline. 

We now have a final deadline of 31 October 2026 for all schemes in scope to be connected to dashboards, albeit with more detailed staging guidance to follow. It is likely the revised staging timetable will represent a delay for most schemes, and trustees should be using this extra time to tackle some of the more challenging aspects of connection – one of which is generating their matching policy.

When it comes to matching users to their pension entitlements, something on the surface that seems like a simple task can actually be quite challenging once you scratch the surface and think about how it will work in the real world. Users provide a variety of personal data that gets used to match information held by schemes. Some of the personal data individuals provide will be verified, and we can therefore assume it is correct. However, some will be self-asserted, so it could be mistyped. In particular, National Insurance numbers will be neither a mandatory nor a verified field, so a member may not provide it or mistype it if they do.

The publication in March regarding further guidance on data matching by the Pensions Administration Standards Association (PASA) was a welcome step forward in an important aspect of pensions dashboards. But it’s being left to trustees to decide which bits of data to use to match members to the records they hold. And doing so is much like picking a football team – you need the right players in the right positions. So, let’s look at PASA’s latest guidance in footballing terms and consider how you can make sure your matching policy has the best chance of scoring.

Choosing your team

There’s general agreement among pension data providers that your star strikers are going to be a surname, date of birth and National Insurance number. If a member matches on all of those, you can be confident you’ve got a positive match.

But given all the issues data matching raises, you will likely have to widen your team and look at other data items to match with. There are other obvious big hitters, such as postcodes and alternate surnames. PASA’s guidance goes further and suggests looking at other potentially unique identifiers, such as email addresses and mobile numbers. These are not data items traditionally used – or even held – widely by pension schemes, but they are arguably very useful for matching members on dashboards. Alongside some of the more obvious data items, these could be the perfect midfielders to create opportunities for further positive matches.

Choosing your formation

The way these items work together is important. Look at all the possible combinations and consider which you would be comfortable making a positive match and which you might want to flag as possible matches. This will help generate a list of criteria for positive matches and a separate list that would produce possible matches. But your strategy can’t be too defensive and rely too heavily on possible matches – trustees are obliged to match with as many requests as possible. Consider having fuzzy matching. It is where an item may be considered to match if, for example, two figures are transposed on the bench, ready to come into play if the data isn’t quite hitting the mark.

Getting match fit

Deciding which data items to use may be driven by the current quality of the data you hold. There is little point in building your matching criteria around data items you know are not accurately held. But if you do see the value in using less traditional fields such as email addresses, then now is the time to consider collecting the information from your members, as well as cleansing the data items you already hold.

Post-match analysis

One of the points raised in PASA’s guidance is how trustees might measure the effectiveness of their matching criteria. Sufficient coverage should be considered – it is looking at whether there is a high enough percentage of possible matches turning into positives. Sufficient focus is another key factor  - this involves examining whether data items are defined tightly enough so that a minimal number of non-members are picked up as possible matches. This, in reality, won’t be possible to assess until dashboards are released to the public. But reviewing how well your matching criteria is working will be required to make sure you are fulfilling your dashboard obligations and is expected by the Pensions Regulator.

As PASA concludes, there is not a one-size-fits-all approach. All teams will be different and depend on the available resources and chosen management style. I strongly recommend reading their guidance when undertaking your dashboard preparations.

LCP’s Dashboards Team can help with data cleansing for dashboards and all aspects of dashboard preparation.