Operations
Part of Running a finance-content operation on release packets, not chat threads
Finance service standards: score evidence, count release defects
Report two finance-content service standards with records: an evidence-completeness score and a release-defect count, kept honest by sampling.
Two service standards can be proved from records a small publisher already keeps: evidence completeness and release accuracy. Each needs a definition, a denominator and a sample, or the number it produces is only a claim. The measures here are proposals, not industry benchmarks.
Review hours, correction tiers and incident reporting belong to the wider operating model, set out in running a finance-content operation. Two measures a client can check against the publisher's own files, and the wording that keeps them from becoming guarantees, are the subject here.
What to take away
- A standard states what will happen, who owns it and how performance is evidenced.
- Do not promise instant review or guaranteed compliance.
- Evidence completeness counts every candidate presented for release, including rejected ones.
- A high evidence score does not prove sources support claims, so sample the actual reasoning.
- Release accuracy compares the live item with the signed candidate.
- The first report is a baseline, not a marketing claim.
Write the promise so it can be checked
State what will happen, who owns it and how performance is evidenced. Avoid promising instant review or guaranteed compliance, because neither can be shown from records. A promise worded around a measured process can. The standards are contractual and operational choices, and they need review for the particular business.
| Promise to avoid | Checkable replacement | Record that proves it |
|---|---|---|
| Every post is fully compliant | Every release candidate is scored for evidence completeness and the score is reported | Evidence-completeness log |
| Instant review | Review time starts when the reviewer receives the fixed candidate and required evidence | Review clock record |
| Nothing goes live unchecked | Live items are compared with the signed candidate and differences counted | Release comparison record |
UK financial and advertising rules may affect the underlying work, so the wording of each promise should be read against them. Every standard also needs a named owner, and who signs off finance content separates those decisions from the person who wrote the piece.
Score evidence completeness
Measure the share of release candidates whose material claims have a primary source, access date, applicable geography and reviewer outcome. The denominator is every candidate presented for release, including those rejected, so a strict reviewer cannot improve the score by hiding failures.
Evidence completeness flow
- Candidate presented for release
- Material claims have primary source
- Access date and geography recorded
- Reviewer outcome logged
- Missing source after window: hold or remove
Set an ageing measure for unresolved evidence. A missing source after the agreed research window should move the item to hold or cause the claim to be removed. It should not disappear when the deadline arrives, so count held items in the report as well as passed ones.
A high score cannot show that the sources support the claims. Sample the actual reasoning as a separate control: pick released claims, open the saved source and ask whether it states what the piece says. Report how many claims were sampled beside the score.
For possible financial promotions, add the lawful route to the record. FCA FG24/1 says unauthorised persons, such as social media influencers, who promote a regulated financial product or service without approval of an appropriate FCA-authorised person may be committing a criminal offence. Record who made the perimeter decision and what exact version any authorised person approved.
Count release defects
Track whether the live item matches the signed candidate. Compare words, graphics, disclosure, destination and audience settings shortly after release, and count each difference as a defect, even where the uploaded master was correct. Platform truncation, a changed landing page and an omitted caption all count.
Element compared / A defect looks like
- Words
- A caption cut short by the platform
- Graphics
- A chart replaced or cropped after sign-off
- Disclosure
- A label hidden by cropping or fast scene changes
- Destination
- A landing page changed after sign-off
- Audience settings
- A wider audience than the signed candidate named
The comparison needs a fixed candidate, set out in the finance-content quality checklist. Report defects against the number of live items checked, and list each defect with its cause. A count of zero means something only if the comparison was made on every item.
Do not read silence as success. The FCA's consumer-understanding review found that several firms continued to rely on sales data or the absence of complaints as evidence of understanding, and says this does not provide reliable assurance. Other publishers can adopt the evidence discipline without suggesting that FCA duties apply to them.
Report a baseline, not a boast
Give every metric a definition sheet: question, population, exclusions, clock, data owner, calculation and reporting frequency. Show missed targets and their reasons. Revisit targets after staffing, channel or risk changes, and sample the raw records each quarter or at another justified cadence.
The first report should be a baseline, not a marketing claim. Publish the score and the defect count with their denominators, state how many claims were sampled, and set a target only after observing what the team can sustain.
Common questions
How should a publisher score evidence completeness?
Count the release candidates whose material claims all carry a primary source, access date, applicable geography and reviewer outcome, and divide by every candidate presented, rejected ones included. Then sample the reasoning behind a few claims, because a complete record does not prove the source says what the piece says.
What counts as a release defect?
Any difference between the live item and the signed candidate in words, graphics, disclosure, destination or audience settings. Platform truncation, a changed landing page and an omitted caption count, even when the uploaded master was correct, because readers saw the live version.
Can a quiet complaints inbox show that readers understood?
No. The FCA's consumer-understanding review says relying on sales data or the absence of complaints as evidence of understanding does not provide reliable assurance. Use sampled records and defined measures instead, and report them with their denominators.
