Targeted support went live in April. Twenty-three million people are the reason.

The advice gap has been named in every industry conference programme for a decade and measured almost never. The FCA has now put a number on it that it is prepared to defend: around 23 million consumers are currently underserved by the markets for advice and guidance. That figure appears in PS25/22, and it is the reason the rules exist.

The mechanism is narrow and worth stating precisely, because the industry shorthand for it is already wrong. Targeted support is not simplified advice. It is not guidance with a better name. It is a permission to make a suggestion designed for a group of consumers who share a characteristic — not a recommendation to a person.

The word that does the work is "group"

Everything difficult about this regime lives in one operational question: how did this person end up in that group?

A suggestion aimed at "customers aged 55 to 60 holding more than 80% in cash within a pension wrapper" is only defensible if you can show, for a named customer on a named date, that they were in that segment, on what data, and that the data was right. If you cannot reconstruct that, you have not given targeted support. You have given a personal recommendation by accident, with none of the suitability work that a personal recommendation requires.

This is why we think the regime is, in practice, a data-lineage problem wearing a policy hat. The consultation language is about design, delivery and purpose. The thing that will be examined in two years, when the first complaints arrive, is the record.

Six things to be able to produce, per customer, per suggestion

1. The segment definition as it stood on that date. Not as it stands now. Segment rules get tuned; a tuned rule quietly rewrites history unless the version is stamped.
2. The input data and where each field came from. A platform feed, a fact-find, an inferred value. Inferred values are the ones that hurt.
3. The evaluation. Which rule matched, which did not, and the result.
4. The suggestion as it was shown, including the wording, because the wording is the product.
5. Who, if anyone, reviewed it, and what they changed.
6. What it declined to suggest, and why. The refusals are the most persuasive record you will ever keep, and almost nobody keeps them.

That is the same six-record shape we build into every system that touches a customer decision — it is the audit layer, and the argument for it is in the note on why the trail is the product rather than a by-product of it.

Where this collides with Consumer Duty

Targeted support does not suspend the Consumer Duty; it operates inside it. A suggestion aimed at a group still has to deliver good outcomes for the individuals in that group, and a firm still has to be able to show it considered foreseeable harm. The interesting tension is that segmentation is a blunt instrument by design — the whole point is that you are not assessing the individual — while the Duty asks about outcomes for individuals, including vulnerable ones.

We do not think that tension is fatal. We think it means the monitoring matters more than the design: a firm should be watching what happened to the people who received each suggestion, in cohorts, and be willing to retire a segment that is producing bad outcomes. That is a reporting pipeline, and it should be running from the first suggestion, not built after the first complaint.

What this does not do

It does not close the advice gap. It gives firms a way to say something useful to people they previously could not economically serve, which is different. It does not remove suitability obligations where a personal recommendation is actually being made, and the boundary between the two is exactly where the supervision will land. And it does not make the segment data good — a firm whose customer records have not been reconciled in four years now has a compliance exposure where it previously had only an annoyance.

If you are deciding whether to apply, the first question is not "can we design segments". It is "can we prove, for any customer we contacted last month, which segment they were in and why". If that answer takes more than a day to produce, the answer is not yet — and the fix is a pipeline and the evidence work, not a bigger policy document. Advisers reading this may also want the note on what consumers are already doing instead of asking you.