In February 2025, administrators confirmed that the boutique retailers who had sold through Trouva (a curated independent-brand marketplace) were owed £954,000 and would receive nothing. This is not a bookkeeping failure. It is a forensic infrastructure failure with no name in any Statement of Insolvency Practice, no Dear IP bulletin and no published methodology from any Big 4 firm.
The mechanism is not unusual. Trouva collected customer payments, deducted commission of around 25% plus VAT and was supposed to remit the balance to each boutique. By the point of administration, the balance had not been remitted. The administrators, Lee Van Lockwood and Gareth Harris of RSM UK Restructuring, were left to manage the claims of sellers who had effectively extended unsecured credit to a platform they believed was holding their money in transit.
Atterley, a Scottish womenswear marketplace with around four hundred independent boutiques, followed the same pattern into winding up in December 2022. Payment was due to retailers on 21 and 22 December. It did not arrive. Deryane Tadd of The Dressing Room told Drapers the effect on independent retail would be devastating. Rachel Hunt of Sass & Edge estimated she was owed more than £5,000. The liquidators, Brian Milne and David McGinness of French Duncan, confirmed it would be weeks before the total owed to creditors could be established. When it came, it appeared not under a field marked “platform-held seller funds.” It appeared because creditors submitted claims after the fact.
These two cases are the cleanest examples in the public record of a phenomenon the UK insolvency profession has not yet named, let alone systematised. When an ecommerce platform fails, the funds it holds on behalf of sellers (collected from customers, not yet disbursed) do not present themselves to office-holders as a distinct asset class. They sit somewhere between “cash at bank” (technically wrong: the funds typically sit in operating accounts of entities registered in Luxembourg, Ireland, or Canada) and “trade debtors” (more accurate legally, but treating the platform as an ordinary receivable misses the reserve, suspension and chargeback dynamics that determine whether the amount is actually collectable). More often than any taxonomy acknowledges, they disappear into “sundry.” The consequence of that disappearance is structural.
The Insolvency (England and Wales) Rules 2016 are the primary structural reason. Rule 3.30 prescribes a charge-based asset taxonomy: assets subject to a fixed charge, a floating charge and uncharged assets, with no sub-category for platform-held funds. Amazon Account Level Reserves, Shopify payout holds, eBay Managed Payments buffers, Etsy payment reserves, FBA inventory in transit and pending settlements must all be fitted by practitioners into pro-formas built from those three headings. The standard form does not cause the silence. It creates the conditions for it.
SIP 7, revised effective 1 April 2021, reinforces those conditions in practice. Paragraph 9 directs that receipts and payments accounts “should show categories of items under headings appropriate for the insolvency appointment, where practicable following headings used in prior statements of affairs.” SIP 7 does not prohibit reclassification; the “where practicable” language permits it. Yet across the nineteen UK ecommerce insolvencies examined in this analysis, no publicly filed progress report reclassifies or isolates platform-held funds that were subsumed at SoA stage. Nothing in the standard architecture creates an affirmative incentive to revisit the initial categorisation, and in practice it is not revisited. The consequence is not that platform-held funds are legally invisible. The consequence is that nothing in the prescribed forms or the SIPs prompts an office-holder to identify and isolate them in the first place.
No SIP in the current canon (not SIP 1, 2, 3.1, 3.2, 6, 7, 9, 11, 12, 13, 14, 15, 16, or 17) references marketplace data, platform reserves, or payout reconciliation. The Insolvency Service’s Dear IP bulletins from 2024 to 2025 do not address it. ICAEW’s Insolvency Committee has issued no technical release on the topic. One piece of published guidance exists at the boundary of this territory: HMRC Insolvency Practitioner Bulletin 11 (2024) directs that insolvency practitioners appointed over a digital platform operator must use HMRC’s digital platform reporting service under SI 2023/817. That bulletin addresses the situation where the platform itself is insolvent and must report its sellers to HMRC. It says nothing about the insolvency of a seller whose assets include funds held by a platform. The profession has answered a different question from the one that is waiting to be asked.
This is not primarily a bookkeeping problem, though it produces bookkeeping consequences. It is a forensic infrastructure problem: the UK professional services ecosystem (forensic accountants, insolvency practitioners, their professional bodies, their standard forms, their training programmes and their tooling) has not developed or published a methodology for verifying what a multi-channel ecommerce estate is actually worth at the point of insolvency, at the point of a tax inquiry, or at the point of acquisition. The surface area of this gap is not small. Internet sales accounted for 28.2% of all UK retail in January 2026, subsequently revised to 28.0% by the ONS, representing approximately £140 billion of annual consumer transaction volume. Wholesale and retail trade is the second-largest source of company insolvencies in England and Wales, behind only construction. The sector accounted for 3,773 insolvencies in the twelve months to November 2025, sixteen percent of cases where industry was captured, in a year when compulsory liquidations reached their highest level since 2012, according to the UK Insolvency Service. The profession has built detailed forensic infrastructure for a much smaller market, cryptocurrency, and nothing equivalent for this one.
What follows is the full account of that gap, where it sits, what it costs and why it has not been closed until now.
The comparison the profession cannot explain
The ratio of published forensic methodology is 100% to zero: every Big 4 firm for crypto, none for ecommerce settlement. This is not a competitive gap. It is a category that has not yet been named.
The reason that category exists for crypto and not for ecommerce settlement is regulatory. The crypto forensic ecosystem was built because the FCA's registration requirements for crypto-asset businesses, the Law Commission's digital assets report and the high-profile insolvency proceedings (FTX, Celsius, Quadriga) forced the profession to develop methodology. Courts and regulators demanded it. SI 2023/817 and Making Tax Digital for Income Tax Self-Assessment have mandated reporting of ecommerce data. They do not mandate verification. A profession that reports what the platform tells it posts. A profession that verifies what the platform tells it reconstructs. The UK has mandated the former. It has not required the latter, and so the latter has not been built.
The asymmetry is not a general impression. It is documented, firm by firm.
PwC UK maintains crypto tracing and recovery services, blockchain analytics for fund-flow mapping, and served as Joint Official Liquidator for FTX Digital Markets, establishing for that engagement a dedicated cryptoassets data-capture workstream. No equivalent workstream or methodology has been published for any UK retail insolvency involving marketplace settlement data.
EY UK built the EY Blockchain Analyzer: Reconciler, a proprietary tool designed to reconcile off-chain books and records to the public blockchain and to extract on-chain balances for independent verification of wallet records. EY has published no equivalent tool for reconciling ecommerce settlement reports to bank records.
KPMG operates a dedicated Blockchain and Digital Assets working group in the UK and has developed Chain Fusion, a patented suite of crypto accelerators for ingesting and structuring data from blockchain protocols. No equivalent tooling or methodology has been published for ecommerce settlement verification by any KPMG entity.
BDO UK offers dedicated Crypto Asset Services including wallet recovery, fraud investigation supported by machine learning, and derivative valuation. BDO has published nothing analogous for ecommerce settlement verification.
ICAEW published "Considerations for Auditing Cryptocurrencies" in September 2024, providing a detailed framework for audit risk, proof of ownership, and valuation methodology. ICAEW has published no guidance on forensic verification of marketplace-held funds or platform settlement data.
The forensic analytics tools have followed the same taxonomy. MindBridge integrates with QuickBooks, Xero and Sage Intacct. It lists no connectors for Amazon, Shopify, PayPal, Stripe, or Etsy. CaseWare IDEA, the industry standard for UK audit analytics, carries no native marketplace connector. Relativity is confined to document and email review. The tools follow the professional taxonomy: if the profession has not categorised ecommerce settlement as a forensic problem, the tools do not address it. UK forensic recruitment advertisements routinely list blockchain forensics, Chainalysis experience and digital asset tracing as required or desirable skills. No advertisement has been identified that lists ecommerce platform settlement reconstruction as a forensic skill requirement.
The profession built a methodology where regulation demanded one. Where no equivalent demand has been made, no methodology exists. That is the structural explanation.
The UK crypto market is not straightforwardly comparable to ecommerce in scale. Consumer holdings of cryptoassets in the UK stand at approximately £13 to £14 billion by FCA survey data. Revenue earned by UK crypto businesses was estimated at $1.89 billion in 2023. Trading volume on centralised exchanges is not systematically recorded at a UK level; global spot trading volume across the top exchanges exceeded $17 trillion in 2024, but that figure measures the same capital circulating at high velocity and is not comparable to consumer expenditure on goods and services. The £140 billion UK ecommerce figure is annual consumer transaction volume: money spent once, by a buyer, on a product or service.
The absence of a forensic methodology for ecommerce settlement does not appear to be governed by market size in any of these senses. The gap is explained by regulatory demand, not by the scale of the market that lacks it.
What the settlement trail actually is
The settlement layer of ecommerce is more opaque than the ledger suggests and that opacity is the source of the verification problem.
When a seller makes a sale on Amazon, the customer pays Amazon. Amazon deducts a referral fee (8 to 15% depending on category), a fulfilment charge if the goods are stored and shipped through FBA (a further 10 to 15%), and, for UK-established sellers since 1 August 2024, 20% UK VAT on those fee amounts under Amazon's UK-branch invoicing arrangement, which replaced the previous Luxembourg reverse-charge structure when Amazon migrated all UK selling-platform services from Amazon Services Europe S.à r.l. to Amazon EU S.à r.l. acting through its UK branch. Amazon may also deduct advertising charges for Sponsored Products campaigns, apply a reserve against the settlement and net out refunds issued since the last settlement period. The net result arrives in the seller's bank account, typically every fourteen days, as a single consolidated BACS credit. That deposit is not revenue. It is the residue after a series of platform-level calculations, none of which the seller controls and not all of which appear on any single report.
Amazon produces three primary seller-facing reports: the Summary Report, the Transaction Report and the Business Report. Each uses different cut-off times, different fee categorisations and different treatment of returns and reserves. They do not reconcile with each other by design. For the professional asking "what did the seller earn and what did they actually receive," no single report answers both halves of the question. The answer requires a reconstruction: the Settlement Report (a V2 Flat File in Amazon's terminology) reconciled, transaction by transaction, to the bank statement.
Shopify Payments works differently. The platform processes card payments and issues payouts to the merchant's bank account, typically on a two-business-day schedule, but subject to reserve holds that Shopify may impose without advance notice. Shopify's own reserve documentation describes a reserve as "a temporary hold on a portion, in some cases a full amount, of transactions processed through Shopify Payments for a specified period." A UK merchant documented £75,000 held under a 90-day payout extension. A second reported £10,000 held for 120 days. Neither event produces an accounting entry at the moment it occurs. The bank statement reveals only the eventual effect.
In November 2024, Shopify introduced a "Track missing payouts" feature, describing the problem it was solving: "If an expected payout hasn't settled in your bank account but is marked as paid on Shopify…" The platform's own engineers were conceding in a public changelog that a payout status of "paid" within their system does not reliably correspond to a settled bank deposit. That concession matters not because it is surprising but because it is documented in primary source: the platform recording a transaction as complete does not mean the money has moved.
PayPal, Etsy, Stripe and eBay Managed Payments each have their own settlement architectures, their own reporting formats, their own reserve and hold mechanics and their own timing conventions. None of them reconciles to any other and none reconciles to the bank by default. For a multi-channel seller operating across three or four of these platforms simultaneously (which is increasingly standard, with nearly 10% more Amazon sellers operating across multiple platforms year-on-year according to Jungle Scout's 2025 seller survey) the professional working from accounting software alone is working from a derivative of a derivative, three transformations removed from the source records that would establish what actually happened.
Four distinct data sources coexist for every multi-channel ecommerce engagement. The accounting software holds what the firm has posted. The integration tool holds what the platform sent. The platform's own settlement reports hold what the platform calculated. The bank statement holds what actually arrived. None of these fully agrees with any other. The disagreements are not errors to be corrected. They are structural features of how the settlement architecture works. The reconstruction that closes the gap requires all four sources simultaneously, worked from the bank statement backwards to the platform report, transaction by transaction, with every exception categorised and explained.
What the profession has documented when it looks
When professionals have actually pulled settlement data at the point of client engagement, the variances documented are material by any standard.
Leandro D'Elia of Eightx published a detailed account of Amazon FBA accounting that describes what one client's onboarding revealed: $500,000 in unresolved Amazon deposits-in-transit. Half a million dollars sitting in limbo because nobody had been pulling the settlement reports and reconciling against the bank. The money was not missing. Amazon had initiated payouts that had not been traced to the bank and the balance sheet was carrying an asset that existed in principle without being demonstrated in practice. A second client, a green cleaning products company with annual revenue exceeding $60 million, produced the finding that Amazon's own reports "don't match, typically 20% up or down." Twenty percent at $60 million is $12 million. That variance is not a bookkeeping error. It is a structural feature of Amazon's reporting architecture, arising from the misalignment between report cut-off times, settlement periods and return processing cycles.
The UK Seller Central forum documents the same dynamic at sterling scale. One UK seller reported that Amazon's transaction report showed sales of £7,018 for a given month while the payments report showed £5,296: a £1,722 single-month variance, accumulated to £4,000 year-to-date on £60,000 of annual revenue. A second documented 483 shipped orders and over £14,000 in sales per Amazon's Business Reports against product charges of only £8,428 in the settlement report, a 40% gap, with a corroborating reply noting that "some of the big accounting packages can't cater for it correctly as the data from Amazon is incomplete or late." A third reported a £50,000 accumulated variance over seven months between Amazon invoice totals and actual settlement deductions, arising because Amazon invoices by calendar month while settlement reports use a different period. Two primary documents from the same platform, internally inconsistent on their own terms.
In October 2024, Amazon introduced a deferred-transaction policy change known as DD+7, under which orders remained in a deferred state for seven additional days before appearing in settlement or transaction reports. On the Amazon Seller Central UK forum, one seller reported over 1,000 orders missing from reports over a six-day window. A second quantified 13 sales worth approximately £3,000 unaccounted for in that period. A third described "thousands of pounds of sales" in the deferral gap, with the Financial Director "having kittens." Accountants booking from settlement reports during the transition window systematically under-recorded sales, not because the integration tool failed, but because the platform had changed the mechanics of what it reported and when.
The tools the profession relies on to transfer platform data into accounting software (A2X, Link My Books, Synder, Dext Commerce) are built to post. They are not built to verify. That distinction matters because the cascade from platform through tool into ledger can create financial variances that are invisible until someone works from source. The tools are not the weak link. The absence of independent verification is.
Synder's public changelog documents, across its own release history, a catalogue of errors that affected financial accuracy before each fix was deployed: Amazon settlement report logic corrections; Stripe payout-failure detection that was previously absent; TikTok orders previously lost due to status mismatches; Amazon fees previously under-captured, pending up to fourteen days; merged payouts where two payouts posted as one; Shopify tax-sync producing mismatched totals; Stripe posting the incorrect field to Xero for transaction amounts; and Shopify POS orders syncing to the wrong clearing account. These are not edge cases from a poorly-regarded product. They are concessions, made in a public changelog, that a widely-used integration tool was producing incorrect financial records on multiple platforms across multiple years before each fix was deployed.
A2X's support documentation records a February 2023 incident in which an Amazon API change caused settlements to be flagged as "On Hold," with VAT-report data unretrievable until a manual disconnect and reconnect was performed. UK and EU VAT-registered sellers were specifically affected. For any UK seller whose VAT return was filed during that window, the figure in the return was derived from a report that A2X itself was flagging as inaccessible.
Marc Dady, a UK Amazon and eBay seller trading as DadyBros at approximately 41,000 orders per month, is the subject of a published case study by Link My Books that illustrates the cascade with precision. Amazon altered its product-upload file format. The third-party inventory uploader, Seller Tool Kit, stopped assigning Product Tax Codes. Amazon defaulted all new SKUs to the standard 20% VAT rate, including items that were zero-rated or reduced-rated. Link My Books faithfully posted the incorrect VAT rate into Xero. Every link in the chain operated correctly on its own terms. The output, £8,829 in overpaid VAT accumulating at approximately £10,000 over two months, was wrong because the platform changed a file format and the change propagated, correctly, through every subsequent system.
No First-tier Tribunal decision, no Upper Tribunal authority, no High Court judgment and no Financial Ombudsman determination has been identified that names an ecommerce integration tool or specifies a quantified tax discrepancy caused by platform settlement report mechanics. The adjudicatory record is empty. The errors exist, are documented in primary source and are material. Not a single case has reached a tribunal.
The direction HMRC is not looking
HMRC's matching matches the wrong figure in the wrong direction. That is the structural finding this section establishes.
From 1 January 2024, under the Platform Operators (Due Diligence and Reporting Requirements) Regulations 2023 (SI 2023/817), HMRC began receiving seller data directly from platforms. HMRC's Tax Information and Impact Note for the measure estimates it affects between 2 and 5 million businesses and states that reported consideration "will be used to identify and risk assess the individual or company." The stated policy objective is to help taxpayers get their tax right and to bear down on evasion. HMRC estimated implementation costs at £36.69 million including 24 full-time equivalents. No parallel estimate is disclosed for any overpayment-identification workstream.
The HMRC consultation on better use of third-party data, whose outcome was published on 21 July 2025, quantifies the Exchequer impact at £30 million in 2028-29, £350 million in 2029-30 and £465 million in 2030-31. These figures are definitionally revenue-raising. The word "overpayment" does not appear as a policy objective. HMRC's 2024-25 Annual Report records £48.0 billion in compliance yield, defined as revenue collected and protected that would otherwise have been lost to the Exchequer. Compliance yield is by construction unidirectional: refunds identified by HMRC are not counted.
HMRC acted on the data. From early 2025, marketplace sellers began receiving "one to many" nudge letters, documented across ICAEW's Tax Faculty, CIOT and BDO. The accompanying Certificate of Tax Position, hosted in full by ATT and CIOT, asks the recipient to confirm whether their marketplace income has been correctly declared, with the declaration signed under a criminal-offence warning. Experienced tax practitioners will recognise that no compliance prompt from any tax authority operates in the overpayment direction; that is not what compliance prompts are designed to do. The significance for the adviser is not in the design of the certificate itself. It is that a seller who correctly declared a figure that was itself overstated by unclaimed embedded platform fees will attest to that declaration and hear nothing further. The statutory machinery to recover overpaid VAT (section 80, the Error Correction Notice process) and overpaid income tax (Schedule 1AB to TMA 1970) exists and is well understood. No equivalent official prompt, however, initiates the recovery for trading income from digital platforms. The onus to identify and act on overpayment sits entirely with the adviser, and HMRC's compliance architecture does nothing to surface it.
The structural reason for this directional bias is architectural. Under SI 2023/817, platforms report consideration net of their own fees and commissions, with those fees reported separately, a structure broadly aligned with the OECD Model Rules and the EU's DAC7 (Council Directive 2021/514). The definitional starting points are therefore similar. The divergence is operational. HMRC's TIIN states that reported consideration "will be used to identify and risk assess the individual or company." The matching operates on that figure. For any seller who has correctly reported revenue at or above the platform-reported consideration, but has failed to decompose and claim the corresponding platform fees as allowable deductions, HMRC's matching produces no discrepancy to investigate. The seller's filed figure meets or exceeds the platform-reported figure. The overpayment leaves no fingerprint in the data. The DAC7 architecture does not solve this problem either; the gross-versus-net confusion at the seller's end exists regardless of what the platform reports. The EU regime's single net figure reduces the surface area for mismatch, but the UK's requirement to report fees separately creates a second data field that HMRC's operational matching does not appear to use in the overpayment direction. HMRC's matching architecture is oriented toward underpayment. The direction of overpayment is structurally invisible to it.
LITRG's September 2024 position paper on online platforms identifies the gross-versus-net confusion with a worked example (gross £1,050, net £892.50 after a 15% platform fee) and warns that platform data will not be able to distinguish between trading and non-trading activities. LITRG frames this as a taxpayer education problem. No UK professional body, tax publication, or practitioner commentary has publicly named it as a directional-bias problem in HMRC's compliance architecture. That distinction matters, because it changes the advice.
The practical consequence for the adviser is immediate. Platform commissions run from 25 to 40% of gross merchandise value for Amazon FBA sellers, 13 to 16% for eBay UK business sellers and 10.5 to 28% for Etsy UK sellers depending on the Offsite Ads programme. Each of those fee categories is an allowable business expense. Any category that goes unclaimed because it was bundled into the net settlement payout and never decomposed means the seller overpaid tax on that amount. The statutory machinery to recover it exists: under VATA 1994 section 80, through the online VAT Error Correction Notice process that replaced Form VAT652 on 8 August 2022, or under Schedule 1AB to TMA 1970 for income tax. The four-year claim window is open for most UK ecommerce sellers on activity from 2022 onwards. The P800 process that automatically notifies PAYE employees of overpaid income tax has no equivalent for trading income from digital platforms. The statutory machinery to refund exists. The official prompt does not.
One qualification the practitioner advising on overpayment claims must hold in mind: the unjust enrichment defence under section 80(3) of VATA 1994. HMRC may argue that a merchant who overpaid output VAT passed the cost to customers and should not be reimbursed at Exchequer expense. However, HMRC's own VAT Refunds Manual at VRM12240 identifies the factors that tend against the defence, and in a typical B2C marketplace scenario they tend overwhelmingly in the claimant's favour: the customer paid the same gross price regardless of the seller's accounting treatment; the error is idiosyncratic to the seller rather than sector-wide; and the economic burden falls wholly on the merchant rather than having been passed on through the price. HMRC bears the burden of proof on the civil standard. No First-tier or Upper Tribunal decision has applied section 80(3) to an ecommerce gross-versus-net overpayment fact pattern, but the structural analysis under VRM12240 does not favour HMRC where the ecommerce pricing mechanism means the consumer never bore a different cost.
The published evidence of recovery is not speculative. Social Commerce Accountants has documented a case in which £14,500 in unclaimed input VAT was recovered for a single UK Amazon seller over one financial year through HMRC's formal error correction process. Dan Little, CEO of Link My Books, documented on LinkedIn a client standing to reclaim over £100,000 in overpaid VAT over two years. Your Ecommerce Accountant states its aggregate client recovery from Amazon fee overpayments exceeds £500,000.
A specialist firm carrying unverified gross-versus-net exposure on a client file in 2026 is carrying two risks simultaneously. The client may owe HMRC. The client may be owed money by HMRC. The direction of the discrepancy is unknown until the gross-to-net bridge has been reconstructed from source. The reconstruction is the same engagement either way. The asymmetry belongs to the adviser because it belongs nowhere else.
The public insolvency record
Before turning to the case record, the hardest version of the contrary argument deserves to be stated. A senior insolvency practitioner will observe that the absence of a published sub-category for platform-held funds does not necessarily represent a professional failure: these amounts are, in legal terms, trade debts owed by the platform to the seller and the existing taxonomy captures them correctly under that characterisation. On that view, the profession has not missed a category. It has applied the right one.
The answer requires examining what that characterisation actually produces in practice. If treating platform-held funds as ordinary trade debtors were sufficient, the MatchesFashion Statement of Affairs would not have understated submitted creditor claims by 62%. The Olsam CVL would not have proceeded without a document identifying the Amazon Account Level Reserve position. The Debenhams progress report would not stand as the single published instance of an administrator acknowledging merchant-acquirer retentions as a live recovery workstream. The legal characterisation is not wrong. The question is whether it is doing the forensic work the situation requires. The case record answers that question.
Against that background (documented settlement variances, known integration tool errors, architectural HMRC asymmetry and a profession without a published methodology for ecommerce settlement verification) the UK insolvency record of the past six years is exactly what one would expect.
Across nineteen verified UK ecommerce insolvency cases from 2020 to 2026 (encompassing marketplace operators, direct-to-consumer retailers, a luxury consignment platform, an Amazon aggregator and a used-car marketplace that pivoted to marketplace model weeks before collapse) the public record contains no instances of source-level platform settlement reconstruction, no Statements of Affairs identifying platform-held funds as a distinct asset class and no administrator communications discussing Amazon, Shopify, eBay, or Etsy counterparty positions in the context of fund recovery. This is a finding about what appears in the public record. It does not exclude the possibility that office-holders have performed such work privately and without public documentation. The absence in public filings, press releases and professional publications is, however, total and it is the kind of absence that has a structural explanation rather than an accidental one.
The most striking single case is Debenhams, not because it is the most recent but because it is the only case in which an administrator-authored document directly acknowledges the problem in primary source. FRP Advisory's progress report dated 8 November 2021 (covering the period 9 April to 8 October 2021, produced by joint administrators Geoffrey Rowley and Alastair Massey) states: "The Administrators continue to liaise with the merchant providers in respect of ongoing retentions which will be released on a piecemeal basis in future periods."
That sentence is the only administrator-authored statement in the entire researched corpus that directly acknowledges platform-held funds as a live recovery workstream. The amounts are not disclosed. The counterparty names are not disclosed. One sentence in one progress report is the complete published professional literature on ecommerce settlement reconstruction in UK insolvency.
MatchesFashion Limited entered administration in March 2024, with Benjamin Dymant and Julian Heathcote of Teneo Financial Advisory as joint administrators. The case illustrates not the scale of a Statement of Affairs variance, since experienced retail practitioners will recognise that submitted claims routinely exceed directors' initial estimates, but the mechanism by which the gap was identified: passive creditor submission rather than active administrator reconstruction. The platform operated mixed wholesale and e-concession arrangements with Kering, Gucci (£553,338), Burberry (£467,525), Saint Laurent (£323,648), Bottega Veneta (£326,564) and Prada (£281,069) among the creditors. Those brand partners were classified as undifferentiated trade creditors or retention-of-title claimants, not as a distinct class of marketplace-model counterparties with settlement positions requiring reconstruction separate from ordinary trade debt. The directors' Statement of Affairs estimated unsecured creditors at £31 million. By the 2025 progress report, submitted claims had risen to approximately £50 million, a 62% increase on the directors' initial estimate, reported consistently across Retail Gazette, TheIndustry.fashion and Companies House filings. The gap between what directors believed was owed and what creditors said was owed closed only because creditors submitted claims. There was no intermediate independent reconstruction to determine which figure was correct.
Cazoo pivoted from first-party online car retailer to pure marketplace model in March 2024, weeks before entering administration in May 2024 with Matt Mawhinney and David Soden of Teneo as administrators. The timing matters precisely. From 1 January 2024, under SI 2023/817, Cazoo in its marketplace form became a "platform operator" with reporting obligations to HMRC in respect of its dealer sellers. The administration commenced within five months of Cazoo acquiring those obligations. Combined unsecured debt reached approximately £259.3 million across 10,107 creditors. No public documentation indicates any engagement with the platform-operator reporting obligations, or any reconstruction of dealer settlement positions from source data. Teneo's public communications were organised around the pivot narrative (Cazoo had stopped selling cars directly to the public in April and all customer vehicle deliveries had been completed), a consumer-facing framing that disappears the dealer-side settlement position entirely.
Olsam Group Limited, the sole confirmed UK Amazon aggregator insolvency, entered creditors' voluntary liquidation on 4 November 2024, with Robert Goodhew and Geoffrey Bouchier of Kroll Advisory as liquidators. Olsam's entire asset base was Amazon-dependent: its portfolio of FBA brands generated revenue through Amazon settlements, held Amazon Account Level Reserves and had FBA inventory in transit at any given point. Because the insolvency proceeded as receivership plus CVL rather than administration, no Statement of Administrator's Proposals was ever filed. The insolvency procedure itself prevented the creation of the document that would have revealed whether platform settlement reconstruction was performed.
These cases, alongside Trouva and Atterley described above, define the pattern. The absence they represent is not an absence of effort or competence on the part of the named practitioners. It is an absence produced by forms that contain no prompt, standards that create no incentive to revise the initial categorisation and a profession that has not yet developed the vocabulary to name what it is not collecting. In a survey of public statements made by administrators and liquidators across the nineteen cases, the professional voice is organised entirely around four themes: causation narrative, stakeholder reassurance, sale mechanics and creditor expectations. Platform settlement data has no voice in the UK insolvency public record. The vocabulary for it does not yet exist.
The legal infrastructure that has not been used
The absence from the case record is not because the legal powers are unavailable. They are available. They have not been applied.
Section 236 of the Insolvency Act 1986 gives an office-holder the power to compel production of information from any person capable of providing it. The question of whether that power extends to platforms operating from outside the UK requires care in the post-Brexit environment: in Re Akkurate Ltd [2020] EWHC 1433 (Ch), Sir Geoffrey Vos C held that section 236 does not itself have extraterritorial effect, reaching its outcome in that case only through Council Regulation (EC) No 1346/2000, a route no longer available against EU-resident respondents following the UK's departure from the European Union. The practical route for a UK administrator has therefore become more straightforward than the cross-border authorities might suggest. From 1 August 2024, Amazon migrated all UK selling-platform services from Amazon Services Europe S.à r.l. to Amazon EU S.à r.l. acting through its UK branch. eBay (UK) Ltd, Stripe Payments UK Ltd and PayPal's UK operations are similarly UK-domiciled or UK-branched. Service on a UK subsidiary or branch on a sufficient connection basis is domestic service. Crowell & Moring's 2024 analysis confirms that section 236 has been successfully used against technology platforms to obtain records held in third-party systems, and that the power extends to situations where company records are held by a platform rather than by the company itself.
Schedule 36 to the Finance Act 2008 gives HMRC equivalent compulsion powers over third parties, with extraterritorial effect confirmed in [R (Jimenez) v FTT and HMRC [2019] EWCA Civ 51](https://www.rpclegal.com/thinking/tax-take/jimenez-extraterritorial-reach-of-schedule-36-information-notices). Norwich Pharmacal relief (compelling a third party to disclose information that enables a wrongdoer to be identified or pursued) has been upheld against an online platform at Supreme Court level in [RFU v Consolidated Information Services (Viagogo) [2012] UKSC 55](https://supremecourt.uk/uploads/uksc_2012_0030_press_summary_158423871a.pdf).
None of these powers has been deployed against Amazon, Shopify, eBay, Etsy, PayPal, or Stripe for merchant settlement data in any reported UK proceeding. No BAILII judgment exists on the point. No Dear IP guidance addresses it. No law firm client alert describes a successful application. The powers are available. The question of what is worth recovering through them is what has not been established, because the underlying reconstruction has not been performed.
In December 2025, the Canadian Federal Court of Appeal provided the nearest available international comparator. In Canada (National Revenue) v. Shopify Inc., 2025 FCA 232, the Court granted a preservation order requiring Shopify to retain data from inactive Canadian merchant accounts that would otherwise have entered the platform's standard two-year deletion cycle. The significance for UK practitioners is double: first, that governments understand this data to be worth litigating for; second, that Shopify's retention architecture operates on a standard cycle that destroys it. Shopify confirmed on 27 January 2026 that the Admin GraphQL Event interface and Admin REST events resource would be restricted to one-year retention. The platform's data architecture is tightening on its own commercial timeline. An office-holder's window for contemporaneous capture is a day-one decision.
Amazon Payments UK Limited is an FCA-authorised payment institution, firm reference 799814, authorised under the Payment Services Regulations 2017. The methodology the profession has developed for payment institution insolvencies under the Payment and Electronic Money Institution Insolvency Regulations 2021 (SI 2021/716) (safeguarded customer-fund asset pools, daily reconciliation, bar dates, distribution plans) has been applied in detail to regulated firms, most recently in the special administration of Argentex LLP (joint special administrators Daniel Conway, Anthony Wright and David Hudson of FRP Advisory Trading Limited, appointed 21 July 2025 under PESAR 2021), and is the subject of FCA Finalised Guidance FG25/2. None of that methodology has been transposed to the marketplace seller context, notwithstanding that Amazon Payments UK Limited is itself an FCA-authorised payment institution whose regulatory perimeter includes payment services to marketplace sellers. It is the most immediately available doctrinal bridge in the professional literature. It has not been crossed.
FBA overcharges and a closing window
The specialist FBA reimbursement industry exists as separate evidence that a specific class of error within Amazon's settlement infrastructure is real, deterministic and recoverable, but only by those who know to look.
GETIDA and Carbon6's Seller Investigators, the two largest specialist FBA reimbursement firms, estimate that Amazon FBA sellers lose between 1% and 3% of annual revenue to FBA fee and inventory discrepancies: dimensional fee overcharges based on incorrect weight or size measurements, lost inbound shipments, damaged inventory attributed to the wrong settlement period and pick-and-pack errors. No independent audit or Amazon-published aggregate confirms these figures, but the estimate is commercially validated: the reimbursement industry operates at recovery-contingent fee models of 10% to 25%, and the continued viability of multiple firms at those margins over several years is consistent with systematic, material underlying error rates. The existence of a commercially viable reimbursement industry at those recovery percentages is itself documentation that the underlying error is real and recurring. These businesses exist because the errors are large enough and consistent enough to support a market.
On 23 October 2024, Amazon shortened the manual claim eligibility window for FBA warehouse lost and damaged claims from 18 months to 60 days. From 1 November 2024, Amazon began automatically reimbursing sellers for certain fulfilment-centre-lost inventory, but the automated reimbursement does not cover removal claims or shipment-to-Amazon claims and is not retroactive beyond approximately 2 September 2024. UK-specific windows differ from the US: six months for inbound discrepancies and 90 days for pick-and-pack errors. The historical 18-month window that sustained the reimbursement industry has been compressed to 60 days. For sellers and their advisers who were not tracking these claims proactively, the window for the categories still eligible for manual claim is now 60 days from the event date.
No equivalent reimbursement infrastructure exists for Shopify, eBay, Etsy, or Stripe. The error class (fee and settlement discrepancies between what the platform calculated and what it disbursed) exists across every platform. It is systematically caught on one. On the others, it accumulates silently.
Why the gap has not been closed
The problem sits at the intersection of three disciplines and no firm organised around any one of them alone can claim the other two.
Ecommerce accounting specialists (Elver, Social Commerce Accountants, Sterlinx Global, Outserve and the other firms operating in this vertical) understand platform mechanics in the granularity the work requires. They do not characteristically produce evidence-grade deliverables structured for professional working papers, signed off by a named professional independent of the ledger and formatted for adoption into an administration or forensic engagement without modification.
Forensic accountants and insolvency practitioners understand evidence, procedure and professional standards. They operate general-purpose forensic frameworks that can in principle be applied to any dataset. They do not have native platform knowledge (the difference between Amazon's Summary Report and its Transaction Report, the mechanics of a Shopify reserve release, the timing dynamics of a PayPal PPWD withdrawal) and no published methodology or training programme exists to provide it.
Integration vendors understand platform data. They are software products, not professional services firms. A software product cannot attest to its own output. A subledger built by the same tool that posts the data cannot independently check what it posted, not as a criticism of the tools, but as a description of what independence means in professional services terms.
The economics reinforce the gap. Minimum engagement fees at Big 4 forensic practices typically start at £50,000 and can exceed £100,000; mid-tier practices operate from £25,000 upward. For an ecommerce merchant at £1 million to £10 million in annual revenue, forensic verification at those price points is economically unviable. The firms with the capability to build ecommerce settlement methodology have no economic incentive to serve the segment where the reconstruction problem is most concentrated.
The professional taxonomy is the deepest structural constraint. No open-access or publicly indexed methodology, technical release, SIP, guidance note, academic paper, or conference programme in the UK addresses ecommerce platform settlement verification as a distinct professional service, for forensic, insolvency, assurance, or agreed-upon-procedures purposes. The FRC's three IFRS 15 thematic reviews (November 2018, October 2019 and the 2020 follow-up) contain no ecommerce-platform case study despite revenue recognition being a consistent top-ten finding. No academic paper addresses ecommerce platform settlement reconstruction in the Journal of Business Law, Insolvency Intelligence, Insolvency Lawyer, or Corporate Rescue and Insolvency. No conference programme has hosted a session on it.
What does not appear in the taxonomy does not appear in the tools, the training, the standards, or the case law. The profession has not closed this gap because it has not yet named it.
What reconstruction requires
Independent settlement verification starts from the two sources that neither the seller nor the integration tool controls and that a reviewing professional can independently examine: the raw platform export and the bank statement.
For each platform (Amazon V2 Flat File Settlement Report, Shopify Payments payout history CSV, PayPal transaction history, Etsy Deposits CSV, Stripe transaction download) the reconstruction identifies every settlement event, decomposes the gross-to-net calculation into its constituent parts (sales, fees by type, refunds, chargebacks, reserves, adjustments) and matches each net settlement to the corresponding bank deposit by amount and date. Where the match is exact, the item is reconciled. Where it is not, the exception is categorised with a documented reason and a recommended action: settlement timing lag, reserve hold, chargeback reversal, integration tool cut-off error, or unresolved discrepancy requiring further investigation.
The bank statement composition analysis (classifying every transaction on the bank statement by source platform before reconciliation begins) produces, from the bank statement alone, a complete decomposition of what the merchant actually received across all revenue channels during the period. For an insolvency appointment, this decomposition is the starting point for a SIP 7-compliant receipts and payments account that distinguishes, by platform, between verified settlements and unverified platform-reported figures.
The deliverable that matters for a forensic or insolvency engagement is not a report that says "we looked and everything reconciled." It is a report that documents the scope of the inquiry, the methodology applied, the sources relied on, the exceptions identified with categorised reasons and the coverage qualifications where the bank statement was incomplete or a platform export was unavailable. A report in which coverage is bounded and limitations are stated is a report that can be adopted into professional working papers, withstand scrutiny from creditors or opposing experts and inform further professional judgment. Software produces outputs. A named professional signs the report.
Timing governs the value of the work. Contemporaneous capture of platform settlement exports at the moment of insolvency appointment (before reserves are released into a general payout, before the Seller Central account is suspended, before the Shopify store is shuttered, before PayPal freezes the account) preserves the full picture. Reconstruction from historical exports remains possible for platforms that maintain accessible reporting history and the statutory four-year window for VAT overpayment claims means historical reconstruction has real commercial value. The Canada v. Shopify case and Shopify's own tightening retention architecture are, however, clear on the direction of travel: the window closes. The data that exists today may not exist at the same granularity in twelve months.
What this means in practice
For the forensic accountant or insolvency practitioner: the absence of a line item for Amazon Account Level Reserves in a UK Statement of Affairs does not mean the reserve does not exist. It means no one was prompted to ask. The prompt does not appear in the prescribed forms, is not required by any SIP and has not been provided by any professional body guidance or court judgment. The absence is structural, and it falls to the first competent professional on the engagement to supply it.
That professional has the legal infrastructure available (Section 236, Schedule 36, Norwich Pharmacal) and, following Amazon's August 2024 migration to a UK-branch structure for UK sellers, the most practical route to platform data is service on UK-domiciled or UK-branched entities rather than complex cross-border applications. The underlying reconstruction establishes what is worth recovering before any legal route is engaged.
For the ecommerce accounting firm: a client whose clearing account does not close against the bank is not necessarily a client with a bookkeeping problem. It may be a client with a verification problem, a gap between what the integration tool posted and what the platform actually settled. The distinction matters because the remediation for a bookkeeping problem is a journal entry. The remediation for a verification problem is a source-level reconstruction that determines which of the four data sources is correct, documents the exceptions and produces an evidence bundle the adviser can rely on.
Under Making Tax Digital for Income Tax Self-Assessment, mandatory from 6 April 2026 for sole traders and landlords with qualifying income above £50,000, quarterly digital submissions compound unverified platform exposure four times a year. Under SI 2023/817, HMRC receives platform-reported data and has demonstrated it will write to merchants whose filings do not match. The direction of HMRC's inquiry will be toward underpayment. The direction the adviser needs to check independently is both ways.
For the due diligence and quality of earnings practitioner: the question in any acquisition of a multi-channel ecommerce target is whether reported revenue ties to what actually settled to the bank across every channel. The target's management accounts, prepared from accounting software populated by integration tools, are three transformations removed from the raw settlement data that would answer that question. The clearing account balance is not an asset. It is a residue whose composition (what proportion is reconciled settlement, what proportion is reserve, what proportion is unmatched exception) is unknown until the reconstruction is performed.
The infrastructure that fills the gap
What the preceding analysis describes is a gap in professional infrastructure. The gap exists because the taxonomy does not name the problem, the forms do not prompt for it, the standards create no structural incentive to address it, the economics do not reward the generalist who attempts it and no regulatory mandate has forced it into standard practice. That is a description of a category that does not yet exist as named professional practice, not a description of a problem that cannot be solved.
Ledger Forensics Group is the specialist firm organised to verify the ecommerce settlement trail. LFG reconstructs the complete path between platform payouts and the bank, from raw source exports (the SP-API settlement file, the Shopify Payments CSV, the PayPal transaction download, the Etsy deposits export) working backwards from the bank statement, across every revenue channel simultaneously.
The methodology starts from the two sources that neither the seller nor the integration tool controls: the raw platform export and the bank statement. The reconstruction is rules-based and deterministic. Every match is an exact-amount match with a documented rationale. Every exception is categorised with a reason and a recommended action. The independence is structural: the system has no read or write access to the client's accounting software and is wholly independent of the ledger and integration layer. The engagement is informed by ISRS 4400 (Revised) as a structural reference for scope, methodology and findings reporting. It does not provide assurance and does not constitute an audit in the attest sense.
The output is a report that documents the scope of the inquiry, the methodology applied, the sources relied on, the exceptions identified with categorised reasons and the coverage qualifications where the bank statement was incomplete or a platform export was unavailable. A named professional reviews and signs off before the report leaves. A report in which coverage is bounded and limitations are stated is a report that can be adopted into professional working papers, withstand scrutiny from creditors or opposing experts and inform further professional judgment.
For any professional who relies on platform-reported figures without independent reconstruction, the exposure is the gap between what the settlement trail shows and what it actually contains. That gap is the cost borne by the adviser, by the creditor and by the estate when the reconstruction has not been performed.
The absence of a line item for platform-held funds in a UK Statement of Affairs does not mean those funds do not exist.
It means no one was prompted to ask.