Value for Money is the first query, not the whole database

The FCA’s Value for Money framework is expected to bring the first assessments in 2028, where schemes will be assessed on investment performance, costs and charges, alongside the quality of their service.

Many pension organisations are facing into uncertainty as multiple reforms compete for the same finite operational capacity. But that complexity doesn’t change the reality.

VfM asks three separate questions of your data, and sets out why the organisations that build for all three now will be holding a real data capability once the deadline passes, rather than a simple compliance artefact.

Knowing what we know about regulation-informed business change, there is probably a version of the VfM programme that goes like this. The board allocates budget, engineering builds the data pipelines, compliance signs off the first assessment. The project closes, the budget is spent, and the organisation has exactly what it set out to build: a mechanism for producing VfM reports.

That is the minimum viable version. It is also a very expensive point-in-time reporting mechanism. Is there a better route to business value and VfM compliance?

VfM is three data problems wearing one name

The instinct is to treat “VfM data” as a single normalised layer: clean it once, report from it forever. The FCA’s proposals do not support that shortcut. Look closely and they ask for three different things.

• Investment performance, chain-linked across a saver’s whole time with a firm and broken out by three retirement age cohorts: 30 years to retirement, 5 years to retirement, and at retirement. 
• Costs and charges, unbundled into investment charges and service costs for 1-year reporting periods, and reported at the overarching arrangement level rather than split across the same retirement age cohorts used for investment performance.
• Service quality, measured through a mix of transactional data, complaint volumes and satisfaction surveys triggered at specific member events.

Three data models, three sources of truth, three sets of stakeholders spanning investment operations, finance and member services. If you don’t design for the future shape of these three different data models then you will build a brittle Frankenstein’s monster which will make iterative change harder than it needs to be.  

The build is five layers, not one dataset

Accept that VfM is three data challenges, and the next question is what to build. The reflex is still a single clean pipeline with all three routed through it, but that is the wrong shape twice over. The key to making progress is building five layers:

• Source systems, the raw feeds each data job already has: investment accounting, the finance ledger, the complaints system. Parse, validate, and standardise fragmented third-party extracts, so the rest of your architecture doesn’t stall before it starts.
• Domain data marts, one structured mart per data job (investment performance, costs and charges, service quality), each with its own owner and refresh cycle.
• The assembly layer, where the FCA’s specific arithmetic actually happens: cohort mapping across the 30/5/0-year cohorts, chain-linking for merged arrangements, the cost split and the FLM assumptions log.
• The comparator interface, the submission connector into the central VfM database the FCA is building.
• Consumption, where the IGC or trustee board turns the numbers into a rating, and where disclosure and member-facing tools draw on the same normalised data.

None of this is trivial to begin with. The Pensions Regulator’s own November 2025 engagement exercise found that fewer than three in five were confident in the accuracy of the member data they hold, a shortfall TPR described as an industry-wide “data debt”. That research is more about dashboard readiness, not VfM specifically, but the underlying problem, accurate record-keeping and data quality that has drifted for years without a business reason to fix it, is exactly what VfM will now put a spotlight on. 

With TPR’s analysis in mind, before commencing with any build activity, a data readiness audit is required to check data completeness. From there, the objective is to get a real number quickly, not a perfect one.

One working pipeline beats three planned ones

The temptation at this stage is building for every case before you have a proven single step. Waiting for three or five years of clean historical data before starting, designing a cost split for employer-varied arrangements that do not exist in your book, or treating the comparator connector as day-one work, all create complexity the deadline does not require. Instead, we recommend you prove the simplest pipeline first (likely fund performance data) then decide, deliberately, how much of the rest to build ahead of 2028.

The fastest route into fund performance data is one year of gross performance for savers in the simplest, static allocation funds, exactly the case the FCA flags as low burden. That proves the cohort engine and the pipeline end to end before chain-linking, lifestyling or the FLM model are added. Three, five and ten-year averages, merged arrangements, cost splits and service quality metrics all layer on once that first slice works, rather than being built in parallel with it. Each phase is validated against real data, so any issues with the model or the data surface sooner rather than later.

Some of the FCA’s proposed service quality metrics are pure operational telemetry: response times, complaint rates, things you can pull from a system that already exists. Others are not.

The proportion of savers using retirement planning tools, apps or calculators is a product adoption metric. It only exists if you have built the tool and mapped the KPIs and underlying analytics. Satisfaction surveys are event-triggered: they need to be issued within a set window of a specific member action, not sent out on a fixed annual cycle.

None of that is instrumentation you bolt onto an existing pipeline. It requires decisions about member journeys, guidance content and communication timing that sit with product and member services as much as engineering. 

These metrics are supposed to help you analyse  how well you actually serve members, not just how well you record it.

Your IGC can only be as good as the data you give it

The consultation puts the Independent Governance Committee, not the firm, at the centre of the judgement. 

IGCs and trustees benchmark each arrangement against a commercial market comparator group, compiled through a central database rather than selected by the scheme itself. Ratings run on a four-point scale, red, amber, light green or dark green, so solid, fair value isn’t rated the same as clear outperformance.

A governance body asked to run that process against thin, inconsistent or late data can’t discharge the function properly, however capable its members are.

Poor data infrastructure doesn’t just produce a weak VfM report. It produces a governance body that cannot do its statutory job. That is a different, and arguably more serious, failure mode than a compliance miss, and it is the one that should concern trustees and IGC chairs.

Build the three streams properly, rather than bolting them together for a single report, and what you are left with is more than a compliance artefact, it is a normalised member data layer that can support personalised engagement and contribution modelling. 

Regulators expect third-party league tables to emerge from Framework data, and a provider with clean, comparable data is better placed to shape how it is read than one that simply appears in someone else’s table. It also raises, cautiously and without committing to it, the possibility that future iterations of the Pensions Dashboard could surface some VfM information directly to savers. 

Neither is guaranteed. Both are more useful to you if your data is already in the right shape when they happen than if you are starting from scratch.

The two types of organisations responding to VfM

There is a useful precedent for what happens when a significant regulatory moment arrives in financial services and organisations get to choose between a strategic response and a defensive one.

The Open Banking rollout of 2018 is instructive. In a previous article, we drew on Deloitte’s survey of European banks responding to PSD2, which found two distinct archetypes. Challengers treated the directive as a platform opportunity, investing in architecture and capability beyond the compliance requirement. Minimalists took a defensive posture, met the legal requirement and stopped there. The majority fell into the second category.

The analogy holds if you push it one step further than most VfM commentary does. In Open Banking, the challengers’ winning move was building APIs and customer-facing products on top of PSD2 infrastructure. The VfM equivalent is specific, not general: member engagement tools built on the same normalised member data, a data model clean enough to make consolidation straightforward rather than a project in itself, and a position ready to take advantage of dashboard or league table exposure if and when it arrives. 

The regulatory returns from both approaches looked similar in 2019. They looked very different by 2022. The first VfM assessments in 2028 will look broadly comparable between compliant and strategically capable organisations. The real difference will become visible from 2030.

The question worth asking now

The practical question for any senior leader in a pensions organisation is not whether to build the VfM infrastructure. Under the FCA’s proposals, you have to. The question is what you are building it to do.

If the answer is “meet the VfM reporting requirements”, you will build reporting infrastructure for three separate data problems. If the answer is “create a data capability that produces the assessment as its first use”, the architectural decisions that get you there are made early, by the people designing the data model.

The cost of clean data, structured investment performance reporting and defined service quality metrics does not change enormously no matter how ambitiously you scope the downstream use. 

Both routes will produce a rating in 2028. One of them will still be building in 2030. The other will be moving.