Value-for-Money Under the Retail Investment Strategy: The Benchmarking Data You Will Report

The Retail Investment Strategy turns “good value” from a marketing claim into a data claim your firm has to substantiate on demand. If you cannot reproduce the cost and performance figures a supervisor asks about, the honest position is that you cannot evidence value for money at all.

On 18 December 2025 the Council and the European Parliament reached a provisional agreement on the Retail Investment Strategy — an omnibus directive amending MiFID II, the Insurance Distribution Directive, UCITS, AIFMD and Solvency II, alongside a separate regulation amending PRIIPs. The final legal text is still being cleaned up and both institutions have yet to formally adopt it, so every date below should be read as provisional and checked against the version published in the Official Journal. But the architecture of the value-for-money test is now stable enough to build against, and the sensible firms are building against it now rather than after the text lands.

This is a data problem before it is a compliance problem. The regulation does not ask you to believe your products offer value; it asks you to demonstrate it against evidence, repeatedly, in a form a national competent authority can interrogate. That is a question about which fields you capture, where the peer data comes from, and whether the number you reported six months ago can be regenerated today from the same inputs.

What the value-for-money test actually requires

Strip away the negotiation history and the agreed mechanism has two limbs. First, the manufacturer or distributor has to identify and quantify all costs and charges borne by the investor, and assess whether those costs are justified relative to a peer group of comparable instruments. Second, ESMA and EIOPA will develop supervisory benchmarks — statistical reference points drawn from market data — that national supervisors use to challenge products that sit outside the pack. The firm-level peer comparison is your obligation. The supervisory benchmark is the lens the regulator points back at you.

The consequence sharpens the product governance process rather than adding a disclosure line. A product that fails the cost-justification test is not meant to reach the retail market at all. So the value-for-money assessment moves upstream, into product approval, and it has to be evidenced at the point a product is signed off and re-evidenced when it is reviewed. That is the shift that catches firms out: this is not an annual report you assemble in arrears, it is a gate your own data has to clear before anything is sold.

The fields you actually have to hold

“All costs and charges” is deceptively simple. In practice the assessment needs a cost model decomposed to the components a supervisor can compare across products, not a single ongoing-charges figure. At minimum you need, per product and per share class:

• Management and administration fees, separated — not netted into one ongoing charge.
• Transaction costs, including implicit costs, which are the ones most firms cannot reconstruct.
• Entry, exit and any performance fees, with the calculation basis, not just the headline rate.
• Distribution costs and any inducements attached to the product.
• Net and gross performance over standardised horizons, so cost is read against what the investor received.

The inducements limb of the same reform makes the distribution cost line load-bearing in its own right — a point I have taken separately in instrumenting the new disclosure and best-interest tests. For value for money, the discipline that matters is that each of these components is a stored, dated field with a known source, not a number recomputed on a spreadsheet each quarter by whoever is free.

Where the peer cohort comes from

The comparison is only as defensible as the cohort. A peer group of “similar instruments” has to be defined on attributes a supervisor would accept — asset class, strategy, risk indicator, currency, target market — and it has to be stable enough that you are not quietly reselecting a flattering set of peers each period. That means sourcing external product data (fund universes, cost datasets, the PRIIPs KID population) and aligning it to your own taxonomy, which is a data-interoperability exercise as much as a compliance one. The same standards discipline I have described for interoperable data spaces applies here: if your cost fields and the market’s cost fields are not defined the same way, the percentile you report is noise dressed as evidence.

Store the cohort definition and its membership as of the assessment date. When a regulator asks why a product priced in the top cost quartile was approved, the answer is a query against a frozen cohort, not a reconstruction from memory. This is exactly the reproducibility argument for versioned, snapshot-capable table formats I made in the case for Apache Iceberg — the value-for-money file is a time-travel problem, and storage that cannot show you the data as it stood on the approval date will not survive a supervisory challenge.

The calculation, as a query rather than a scramble

Once the fields and cohorts are in place, the value-for-money indicator is a straightforward computation: where does each product sit against its peers on cost, and where does it sit on net performance. The point of expressing it as a query is that it is reproducible — the same inputs give the same output, dated and auditable — rather than a slide someone rebuilt for the committee.

-- Value-for-money indicators: product cost and net performance
-- positioned against a dated peer cohort. Percentiles are computed
-- within cohort so the comparison is like-for-like.
WITH cohort AS (
    SELECT product_id, share_class, cohort_id,
           total_ongoing_cost_bps,      -- all recurring costs, in bps
           net_return_5y_annualised_pct -- net of every charge above
    FROM vfm_product_snapshot
    WHERE snapshot_date = DATE '2026-06-30'  -- frozen assessment date
)
SELECT
    product_id,
    share_class,
    cohort_id,
    total_ongoing_cost_bps,
    net_return_5y_annualised_pct,
    -- 1.0 = most expensive in cohort; flag the top cost quartile
    PERCENT_RANK() OVER (
        PARTITION BY cohort_id ORDER BY total_ongoing_cost_bps
    ) AS cost_percentile,
    -- 1.0 = best net performer in cohort
    PERCENT_RANK() OVER (
        PARTITION BY cohort_id ORDER BY net_return_5y_annualised_pct
    ) AS net_return_percentile,
    CASE
        WHEN PERCENT_RANK() OVER (
                 PARTITION BY cohort_id ORDER BY total_ongoing_cost_bps
             ) > 0.75
         AND PERCENT_RANK() OVER (
                 PARTITION BY cohort_id ORDER BY net_return_5y_annualised_pct
             ) < 0.50
        THEN 'REVIEW: high cost, below-median net return'
        ELSE 'within tolerance'
    END AS vfm_flag
FROM cohort
ORDER BY cost_percentile DESC;

The flag is not the decision — a high-cost product can still offer value, and the regulation expects you to explain why. But the flag forces the explanation to exist, and to sit next to the numbers that provoked it, at the moment of approval.

Build the pipeline before the text is final

The reported timeline gives member states roughly two years to transpose and around thirty months to application, with the PRIIPs changes on a shorter clock — all subject to the final text and the eventual Official Journal dates. That looks like room. It is not, because the work that decides whether you pass is the unglamorous part: defining cost components consistently, sourcing a defensible peer universe, and storing both so the assessment is reproducible on the date it was made. None of that depends on the final wording, and all of it takes longer than the drafting of a policy.

A firm that stands the data model up now turns the value-for-money assessment into a query it can run on demand. A firm that waits will be reconstructing five years of cost and performance under supervisory pressure, from spreadsheets nobody can quite reproduce — which is the exact position the regulation was written to end.