# Not Using It Is Not A Refund Right: The Money Is In Trader Compliance Failures, The Regime Everybody Is Waiting For Does Not Arrive Until Spring 2027, And The Pool Of Claimable Failures Is Largest Now And Shrinks Once It Does

**version** v0.33.61
**date** 20 August 2026
**from** Human (project lead)
**to** Strategy, Product, Legal, the sgit.ai site team

**type** Strategy brief

*Sixteenth of 20 August, and the first in this corpus aimed at a consumer market rather than an enterprise one. The legal position is checked against a government written statement, a legislation tracker, official trading standards guidance and the regulator's own scope pages, and is quoted from them rather than recalled, because one central assumption turned out to be wrong on the date. Limitation: this is a mapping of published rules and is not legal advice, no solicitor has reviewed it, the position on one lever is flagged as needing advice rather than resolved, and the whole legal surface changes in spring 2027 so every claim here carries that date.*

---

## What This Is

A service that recovers money from subscriptions, the correction its headline needs, and the timing argument that makes it worth starting now: **the ask is a UK service that cancels subscriptions and recovers money for services not used, sharing documents through vaults and using data protection rights to get evidence out of companies, with the founder as the first client; the first correction is the headline itself, because not using something you agreed to pay for is not a legal basis for a refund anywhere in the rules that apply, so a service promising that will fail on its first contested claim, and what is actually recoverable is money taken where the trader made a compliance error, which is a completely different proposition and a much stronger one because those errors are findable, provable and arithmetic rather than argued; the second correction is a matter of fact about the date, since the subscription regime under the 2024 Act that has been widely written about, with its cooling-off periods, renewal reminders and mandatory easy exit, is not in force and a government written statement in April 2026 gives spring 2027, having already slipped from autumn 2026, so a service launched today cannot rest on it and anybody who assumes otherwise will build against rules that do not yet bind anybody; that turns out to be an argument for starting now rather than waiting, because the strongest lever currently available extends a cancellation right by up to twelve months where the trader failed to give the required cancellation information, and official guidance states that in that case the consumer owes nothing for services supplied during the extended period even where they asked for the service to start, so today's claimable pool reaches back over a year into a market that has not yet been forced to tidy its sign-up flows, and it will shrink once every trader is compelled to fix them; the discovery mechanism is a subject access request, which must be treated as if made by the person themselves when sent by an authorised representative and which the regulator says has a high threshold for refusal, and its most valuable output is the usage record, which converts an unprovable feeling that nothing was used into evidence, though the request must be a genuine access request rather than leverage, since a request made with no intention of exercising the right is the one case the regulator names as refusable; the document pipeline is not new work, because a debrief from March records this exact workflow already run on the founder's own credit card statements, sharing a vault snapshot with an agent by a short token, working inside it, committing and pushing back, so only the destination changes; and the honest limit on the vault story is the one this corpus has recorded twice already, that vaults buy portability and controlled sharing rather than zero knowledge, because an agent reading statements to find subscriptions is reading plaintext.** It is the sixteenth document of 20 August (cross-ref: the March accountant pack debrief, the v0.24.56 folders-with-version-control brief, the v0.33.58 WhatsApp bridge brief, the v0.33.59 comparison pages brief, and the v0.33.56 token gateway brief). New contributions: **the headline promise corrected, the commencement date established against the assumption, the closing-window argument, the levers ranked by strength with the evidence each needs, the access request identified as the discovery engine with its one trap, the pipeline located in March, and the regulatory perimeter mapped with the single lever that may cross it.**

## The Headline Promise Needs Correcting Before Anything Is Built

The ask is money back for services not used. **That is not a right, and no rule in the applicable set creates it.**

If a person agreed to pay for something, had it available, and did not use it, they owe the money. Gym memberships, unwatched streaming, forgotten software: the trader supplied what was promised, which was availability. Non-use is the consumer's own conduct.

**This matters more than a wording quibble**, because a service whose promise is that will lose its first contested claim, in writing, to a trader's legal team, and that letter will be the thing customers see.

The recoverable money is somewhere else and it is a better place:

> **You do not get money back because you did not use it. You get money back because the trader did something wrong when they signed you up, renewed you, or took your money.**

And the reason that is the stronger proposition is that it converts an argument into a calculation. This corpus reached the same shape on 2 August, where the most defensible finding in a worked example was the one that was arithmetic rather than judgement: a fact set against a written obligation. **A sign-up flow that did not carry the required cancellation information is that kind of fact.**

Non-use does not disappear from the case, though. It stops being the claim and becomes **evidence** supporting misrepresentation, unfair terms, and negotiation. That is a demotion in the pleading and a promotion in the file.

## The Regime Everybody Is Waiting For Is Not In Force

The single most consequential fact for this product, and it is a date.

The subscription contracts regime under the Digital Markets, Competition and Consumers Act 2024 is the one described everywhere as click to cancel: pre-contract information, cooling-off rights when a trial converts or a long contract auto-renews, reminder notices before those moments, and an obligation to make ending a contract straightforward including online where sign-up was online.

**It is not in force.** A government written statement of 13 April 2026 says the regime is expected to commence in **spring 2027**, that no provisions are currently in force under that framework, and that the detail depends on secondary legislation to be introduced when parliamentary time allows. A legislation tracker records the same, and records that the timeline had previously been autumn 2026 before the April 2026 announcement moved it.

**What is in force is a different part of the same Act.** The unfair commercial practices provisions and the regulator's direct enforcement powers commenced on 6 April 2025, with penalties reaching ten percent of global turnover for the most serious consumer law breaches.

So the position today:

| | Status |
|---|---|
| Cooling-off when a trial converts or a long contract renews | **Not in force. Spring 2027** |
| Mandatory renewal reminder notices | **Not in force. Spring 2027** |
| Mandatory easy exit, online where sign-up was online | **Not in force. Spring 2027** |
| Unfair commercial practices, and direct enforcement with turnover-based fines | **In force since 6 April 2025** |
| The distance selling cancellation regime | **In force since 2014, and it is the workhorse** |

**Anybody building this without checking that date would have built the product around rules that bind nobody**, and would have discovered it in a reply from a trader rather than in a design review.

## Which Makes The Claim Pool Largest Now

The consequence, and it inverts the instinct to wait for the law to land.

The strongest lever available today extends a cancellation right where the trader failed to give the required information. Official guidance states the mechanism precisely: the cancellation period is fourteen days from the day after the contract was made; where the trader has not given the required cancellation information it becomes fourteen days from the day after the information is eventually given; and **the longest that period can extend to is twelve months from the day after the normal cancellation period would have ended.**

And the part that produces money rather than merely an exit:

> **Where the trader failed to provide the required cancellation information, the consumer owes nothing for services supplied during the cancellation period, even if they expressly asked for the service to start.**

Set that beside the commencement date and the shape of the opportunity is unusual:

| | Today | After spring 2027 |
|---|---|---|
| Traders compelled to fix sign-up and exit flows | **No** | Yes |
| Prevalence of information failures | **Highest it will ever be** | Falling |
| Reach-back of a live claim | **Up to roughly thirteen months** | The same rule, over a tidied market |
| Competitors in the space | Few | Many, once the rules are famous |

**So the pool of claimable failures is largest now and shrinks once the regime forces every trader to correct its flows**, and the twelve-month reach-back means today's pool includes contracts entered a year ago. That is a closing window rather than an early start, which is the opposite of how this looks from the outside.

It also gives the service a second act rather than a cliff. When the regime commences, the method transfers unchanged to a larger surface with clearer obligations, and the service arrives with a year of worked cases while everybody else is reading the guidance.

## The Levers In Force Today, Ranked By What They Actually Yield

The mapping the ask requires. Ranked by strength, with the evidence each one needs, because the evidence requirement is what the workflow has to produce.

| Lever | Basis | What it yields | Evidence needed |
|---|---|---|---|
| **Cancellation information not given** | Distance selling regime | Cancellation up to twelve months beyond the normal period, **and nothing owed for services supplied in it** | The sign-up flow as it was, the confirmation, the terms served at the time |
| **Payments taken after cancellation** | The payment was not authorised | Full return of those payments | The cancellation instruction, its date, the payments after it |
| **A term that is unfair** | Consumer rights regime, unfair terms | The term does not bind the consumer | The terms accepted, and how they were presented |
| **Service not supplied with reasonable care and skill** | Consumer rights regime, services | Repeat performance, or a price reduction | Records of what went wrong and when |
| **Misleading action or omission at sign-up** | Unfair commercial practices, in force since April 2025 | Primarily an enforcement and complaint lever rather than an automatic refund | The sign-up flow, screenshots, the omitted information |
| **A claim against the card issuer** | Consumer credit, linked transactions | A claim against the lender as well as the trader | The statement, the amount, the purchase |
| **Card scheme chargeback** | Scheme rules, not law | Reversal, within tight scheme time limits | The statement, and speed |
| **Direct debit indemnity** | Scheme guarantee, not law | Return by the bank | The mandate and the payments |
| **You did not use it** | **Nothing** | **Nothing on its own** | Useful as evidence, never as the claim |

Two things to take from the table rather than from the rows.

**The top row is the product.** It is the only lever that both ends the contract and returns money already taken, it is arithmetic once the evidence exists, and the evidence is a screenshot of a sign-up flow rather than a matter of opinion.

**And the last row is the one to put on the website**, in those words, because a service that says plainly what it cannot do is the only kind anybody should trust with their bank statements.

## The Access Request Is The Discovery Engine, And It Has One Trap

The ask names data protection rights as a way to get information out of companies, and that is the right instinct. It is the mechanism that turns a suspicion into a file.

The regulator's guidance settles the operational questions. A request made by an authorised representative **must be treated as if it had been made by the person themselves**. There is a **high threshold** for refusing on the grounds that a request is manifestly unfounded or excessive, a request is not excessive merely because it asks for a large amount of information, and a repeat request is not excessive merely because an earlier one was made. The response is due within **one month**.

What a request should be asking for, in order of usefulness to a claim:

| Ask for | Why it matters |
|---|---|
| **Usage and access logs** | Converts nothing was used from a feeling into a record |
| The sign-up record: date, route, what was presented | The top lever depends on this |
| Consent and preference records | Whether anything was agreed, and when |
| Billing and payment history | The amount, and payments after cancellation |
| All correspondence, including support tickets | Cancellation attempts that went nowhere |
| The terms in force at the time of contracting | Not the current ones on the website |

**The trap is real and it is the one thing that could discredit the service.** The regulator names as refusable a request where the person clearly has no intention of exercising their right of access, or where it is being used to harass. **A request sent as pressure is that request.** So the rule for the service:

> **Every access request is a genuine request for the person's own data, sent because they want it. It is never sent as leverage, never bundled with a demand, and never sent in a volume calculated to burden.**

That is a constraint on the product's most scalable mechanism, and observing it is what keeps the mechanism available.

One more note worth carrying. A separate right allows a person to receive data they provided, in a portable form. That is narrower than access and, where it works, it produces machine-readable output rather than a bundle of documents, which matters if this is ever to run at volume.

## The Pipeline Was Built In March, For A Different Destination

The document half of this needs almost nothing new, and the evidence is in the corpus.

A debrief of 29 March records the founder preparing statutory accounts materials with exactly this workflow: a pile of personal credit card statements and an export, turned into a structured, navigable pack, with an agent working **inside the vault** rather than through files pasted into a chat. The mechanism it describes is the one this service needs, verbatim: a snapshot of a vault shared with an agent using a short token, the agent clones it, works inside it, commits, and pushes back, and the human pulls and sees exactly what changed. The debrief calls that a genuine shared workspace, persistent and versioned across sessions.

**Only the destination changes.** In March the pack went to an accountant. Here it goes to a claim.

| The March workflow | This service |
|---|---|
| Credit card statements as raw input | The same |
| Parse and structure into a ledger | Parse and structure into a subscription inventory |
| Agent works inside the vault, versioned | Unchanged |
| Shared by a short token | Unchanged |
| Output: an accountant-ready pack | **Output: a claim file per subscription** |

And the positioning was written in April, for this exact audience: professionals who have never used version control, who send bundles of documents to accountants and mortgage brokers, described there as folders that remember everything. That brief also recorded that this pattern is light on key infrastructure and is a serverless, low-cost business.

**So the novel work is the legal mapping and the claim file, not the plumbing**, which is a much smaller build than the ask implies.

## Vaults Buy Portability And Controlled Sharing, Not Zero Knowledge

The honest limit, stated because this corpus has recorded it twice and it will otherwise be overstated in marketing.

An agent that reads bank statements to find subscriptions is **reading plaintext**. It has to. So the privacy claim for the processing step is operational, meaning we do not retain it, rather than architectural, meaning we cannot see it. That is exactly the distinction made on 6 August for the token gateway and again on 14 August for the messaging bridge, where the conclusion was that vaults replace the store and not the processing, and buy portability rather than zero knowledge.

What the vault genuinely delivers here is still substantial and should be claimed precisely:

| Property | True? |
|---|---|
| The client holds their own evidence pack, and can take it elsewhere | **Yes** |
| Sharing with a specific party is a key rather than an account | **Yes** |
| The store cannot read it | **Yes** |
| Every change is versioned, so what was submitted when is answerable | **Yes**, and it matters in a dispute |
| The agent that reads the statements cannot see the contents | **No** |

That last row is the one to publish rather than omit, and this corpus already has the discipline: a page that names where its own approach loses is what makes the rest of it credible.

There is also a finding from earlier today that transfers directly. **A completed claim file is a concentrated dossier of one person's financial life**, and the rule reached this morning for a different product applies here too: it belongs in the client's own vault, never in browser storage, and the operator should hold only what a live claim requires.

## The Regulatory Perimeter, And The One Lever That May Cross It

Checked rather than assumed, because getting this wrong is the difference between a business and an enforcement action.

Regulated claims management in the United Kingdom covers six sectors: financial services and products, personal injury, housing disrepair, specified benefit, criminal injury, and employment, with separate permissions for advising, investigating and representing in each, plus one for lead generation. The regulator's scope material is explicit that **general consumer complaints or refunds from traders such as subscription companies fall outside** that framework.

**So the core service is outside the perimeter.** Chasing a streaming company for a refund is not regulated claims management.

**One lever in the table may cross it, and it is the one worth advice on before it is offered.** A claim against a card issuer is a claim about a financial product against a financial firm, and pursuing that on a customer's behalf, for a fee, looks much more like the first of those six sectors than like a complaint to a subscription company. That is a boundary question, it turns on the precise permissions and on how the service is structured, and this brief flags it rather than answering it.

**The recommendation is to launch without that lever and add it after advice**, because it is the only row in the table that carries this risk and the top row does not need it.

Two smaller perimeter notes worth carrying into that conversation. A contingency fee arrangement changes how the service looks to a regulator regardless of sector. And acting as a person's representative for access requests is ordinary and requires evidence of authority, which the workflow has to capture properly rather than assume.

## The Site

The ask includes a site defining the workflows and the services. **The discipline for it already exists in this corpus and it fits this domain unusually well**, because the legal position here has a known expiry date.

| Rule, already settled | Why it matters here |
|---|---|
| Publish the method before the findings | A service handling bank statements has to show its working |
| Date every claim and give a re-run method | The entire subscription regime changes in spring 2027 |
| State who is writing it | A participant selling the remedy is mapping the law |
| Publish where the approach loses | The plaintext row above, and the levers that yield nothing |

Two things specific to this site.

**Every legal page carries an as-at date and a commencement status.** A page describing cooling-off rights that does not say not in force until spring 2027 is actively harmful, because a reader will assert a right they do not have and be corrected by a trader.

**And the workflows should be published as the agentic workflows they are**, which is what the ask says and what makes the site more than a law summary: this is a mapped pipeline from statements to inventory to evidence to claim, with the agent's role at each step stated, including what it reads in plaintext.

## The Business Model, And The Joke That Will Be Made

Three shapes, and one of them is unavailable for a reason worth stating.

| Model | Note |
|---|---|
| Percentage of what is recovered | Aligns effort with outcome, and **incentivises pursuing weak claims**, which is how this category acquired its reputation |
| Fixed fee per subscription investigated | Honest, and asks for money before anything is proven |
| **A subscription** | **Charging a monthly subscription to cancel subscriptions is a joke that will be made, in public, on the first day** |

The third is genuinely unavailable as positioning even if the economics work, and it is worth deciding that now rather than after a launch.

The first is the obvious choice and it needs the constraint that goes with it. A percentage model plus an automated pipeline that can generate hundreds of access requests is a machine for producing volume regardless of merit, and the access-request trap above is where that ends. **So the merit test happens before the request, not after**, and the service should be able to say what proportion of investigated subscriptions it declined to pursue. That number is the equivalent of the declines metric reached earlier today for a different product: a service that pursues everything is not assessing anything.

The metering side needs no new thinking. This corpus already has a consumption ledger designed to be generic on unit type from the first version.

## The First Client Is The Wrong Sample, Usefully

Worth naming because it will shape the product silently otherwise.

The founder as first client is right for iteration: real statements, real subscriptions, real motivation, and no permission needed to try things. **It is also a sample of one, and an unrepresentative one**: technical, high subscription count, mostly software, comfortable with a command line, willing to write to companies, and in a position to keep going when a first reply is unhelpful.

The typical claimant is none of those, and the parts of the workflow that will break for them are the parts that felt easy: understanding what the reply means, deciding whether to escalate, and continuing after the first refusal. **The first client will not surface those**, so the second and third should be chosen for being unlike him rather than for being easy to recruit.

There is also a straightforward incentive note. A founder recovering his own money is testing a product and also being the customer, and the first honest accounting of whether this works should separate what the workflow found from what he would have found anyway because he built it.

## What This Does Not Try To Be

- **Not legal advice.** It maps published rules and quotes official guidance, and no solicitor has reviewed it.
- **Not a promise of refunds for non-use.** That is not a right and the site should say so.
- **Not reliant on the new subscription regime.** That is spring 2027 and the service must work without it.
- **Not zero knowledge end to end.** The processing step reads plaintext, and the site should publish that.
- **Not a settled perimeter question.** One lever may sit inside a regulated sector and needs advice before it is offered.

## Honest Tensions

| Tension | Note |
|---------|------|
| Correcting the headline | The honest promise is narrower and harder to advertise than the one people would click on |
| The closing window | Starting now catches the largest pool, and the pool is largest because the market has not been forced to tidy up, so the product's best year is also its least defensible pitch |
| Access requests at volume | They are the discovery engine and volume is exactly what makes a request refusable, so the mechanism caps its own scale |
| A percentage fee | It aligns effort with recovery and it rewards pursuing weak claims, which is how this category got its name |
| Vaults in the pitch | They deliver real portability and controlled sharing and they do not make the processing blind, and the second half will be dropped from the marketing by somebody |
| The founder as first client | It gets the product moving and it will hide every failure mode that only appears with a claimant who does not already understand the system |

## Open Questions

| Question | Notes |
|----------|-------|
| Does the card issuer lever sit inside the regulated perimeter? | The one row in the table carrying this risk, and it needs advice rather than reasoning |
| How is authority to act captured? | Access requests by a representative need evidence of authority, and doing it badly is how the mechanism gets refused |
| What is the merit test, and who applies it? | It must sit before the request, and a percentage fee pushes against it |
| How is the sign-up flow evidenced retrospectively? | The top lever needs what the flow looked like then, and archives are the only source |
| What happens at spring 2027? | The method transfers and the pool shrinks, and the transition should be planned rather than absorbed |
| Who owns the dated legal pages? | They expire on a known date and a stale page is worse than none |

## Relationship To Previous Briefs

| Date | Document | Relationship |
|---|---|---|
| 29 Mar | `debrief__claude_and_sg-vault__workflow-to-create-accountant-pack.md` | The statements-to-pack pipeline already run once on the founder's own data, with only the destination differing |
| 17 Apr | `v0.24.56__brief__folders-with-version-control-messaging.md` | The positioning for exactly this audience, and the observation that this pattern is a serverless low-cost business |
| 14 Aug | `v0.33.58__arch-brief__whatsapp-bridge-one-vault-per-conversation-not-the-dedup-check-portability-not-zero-knowledge.md` | Portability rather than zero knowledge, which is the limit on the vault claim here |
| 6 Aug | `v0.33.56__arch-brief__sg-send-token-gateway-resale-prohibited-in-line-forced-append-and-settle.md` | Architectural against operational privacy claims, and the generic consumption ledger |
| 16 Aug | `v0.33.59__strategy-brief__sgit-comparison-pages-as-reproducible-tests-privileges-is-the-missing-column.md` | Dated, re-runnable, method before findings, which the legal pages need because they expire |
| 2 Aug | `v0.33.55__arch-brief__sg-send-end-to-end-worked-example-article-26-5-creditworthiness-agent-fact-to-board.md` | A finding derived from a fact plus a provision, which is what makes the top lever arithmetic |

---

## Key Claims

| # | Claim |
|---|-------|
| 1 | Not using a service you agreed to pay for is not a legal basis for a refund |
| 2 | The recoverable money is where the trader made a compliance error, which is provable rather than arguable |
| 3 | The subscription regime under the 2024 Act is not in force and is expected in spring 2027, having slipped from autumn 2026 |
| 4 | The unfair commercial practices provisions and direct enforcement with turnover-based penalties have been in force since 6 April 2025 |
| 5 | The distance selling cancellation right extends to at most twelve months beyond the normal period where the required information was not given |
| 6 | In that case the consumer owes nothing for services supplied during the period, even having asked for an early start |
| 7 | So the claimable pool is largest now and shrinks once the regime forces traders to correct their flows |
| 8 | An access request by an authorised representative must be treated as if made by the person, with a high threshold for refusal and one month to respond |
| 9 | A request sent as leverage is the one case the regulator names as refusable, so merit is tested before the request |
| 10 | The document pipeline was built and run in March on the founder's own statements, and only the destination changes |
| 11 | Vaults deliver portability, controlled sharing and versioned submissions, and the processing step still reads plaintext |
| 12 | General consumer refunds fall outside regulated claims management, and the claim against a card issuer is the one lever that may not |

---

## Sources

- The government written statement of 13 April 2026 on the subscription contracts regime, stating that the regime is expected to commence in spring 2027, that no provisions are currently in force under that framework, and describing the intended cooling-off rights when trials convert or long contracts auto-renew, reminder notices, and straightforward cancellation routes including online where sign-up was online: https://questions-statements.parliament.uk/written-statements/detail/2026-04-13/hlws1503
- The legislation tracker recording that the unfair commercial practices provisions and the direct consumer enforcement regime came into force on 6 April 2025 with penalties of up to ten percent of annual global turnover, and that the subscription contracts regime is not yet in force with implementation moved from autumn 2026 to spring 2027 following the April 2026 announcement: https://www.wiggin.co.uk/insight/digital-markets-competition-and-consumer-act-tracker/
- Official trading standards guidance on distance contracts, stating a cancellation period of fourteen days starting the day after the contract was made, extension to fourteen days from the day after the required cancellation information is given, a cap of twelve months from the day after the normal period would have ended, the proportionate payment rule where a consumer asked for a service to start, and that where the trader failed to provide the required cancellation information the consumer owes nothing for services supplied during the cancellation period: https://www.businesscompanion.info/en/quick-guides/distance-sales/consumer-contracts-distance-sales
- The regulator's claims management scope material, listing six regulated sectors covering financial services and products, personal injury, housing disrepair, specified benefit, criminal injury and employment, with separate permissions for advising, investigating and representing plus one for lead generation, and stating that general consumer complaints or refunds from traders such as subscription companies fall outside that framework: https://www.fca.org.uk/firms/claims-management-regulation
- The regulator's guidance on refusing a subject access request, stating a high threshold for treating a request as manifestly unfounded or excessive, that a request by an authorised representative must be treated as if made by the person themselves, that a request is not excessive merely because it seeks a large amount of information or repeats an earlier one, that a request where the person clearly has no intention of exercising the right or which is used to harass may be refused, that a reasonable fee may be charged instead of refusing, and that the response is due within one month: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/individual-rights/right-of-access/when-can-we-consider-a-sar-to-be-manifestly-unfounded-or-excessive/

---

This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).