Home · Solutions · Finance & accounting
Solution · Finance & accountingEvery gateway payout unpacks itself into orders, fees and refunds
Gateway payouts matched to orders
A robot collects the gateways’ and marketplace’s transaction reports, matches every line to a shop order, settles the fees, links the refunds and compares the totals with the bank statement. Discrepancies become a short case list, and the accountant receives a statement instead of a riddle.
Executive summary
Gateways and marketplaces pay out in bulk, fees deducted and refunds mixed in; matching that to orders is days of manual spreadsheet work, and a residue still ends up “to be clarified”.
The robot fetches the gateways’ transaction reports, the shop’s orders and the bank statement, reconciles every line three ways, settles fees and refunds, and raises discrepancies as concrete cases.
Reconciling shrinks from days to minutes of review, fees are computed to the line, refunds stop vanishing, and the accountant receives a set that books itself.
gateway and marketplace reports; shop orders (export or API); the bank statement; statements and cases in Microsoft Teams
Business problem
A bulk payout versus a hundred and forty orders
An online shop sells in single orders, but the money arrives wholesale: the gateway pays one amount every few days, fees deducted, refunds and adjustments mixed in. The marketplace does the same, only on its own calendar and its own report logic. What lands on the bank statement is a handful of bulk transfers that must be unpacked back into hundreds of events.
That unpacking is manual work across three files at once: the gateway report, the order export, the statement. The amounts disagree by definition, because of the fee, because of last week’s refund, because the 11:59 p.m. transaction fell into another payout. At several hundred orders a month the reconciling eats days, and a residue still remains: lines “to be clarified” that nobody will ever clarify.
Unreconciled lines are not cosmetics: hiding among them are refunds charged twice, fees taken wrongly, orders paid but unmarked, and money that simply never arrived. A shop that does not reconcile does not know whether the gateway settles it honestly; it only knows things “roughly add up”.
The accountant receives a riddle at the end: bulk transfers with no breakdown, which she must book onto something. In practice she books them onto a suspense account that grows month by month, until year-end close requires untangling it. By then it is archaeology.
How it works today
Below is what the work looks like before anything is automated.
- PersonSomebody opens three files at once: the gateway report, the order export, the statement
- Risk of errorThe amounts disagree by definition: fees, refunds, transactions straddling payouts
- WaitingReconciling eats days; it happens monthly or less
- Risk of errorThe “to be clarified” residue grows; nobody returns to those lines
- Risk of errorFee errors and double refunds stay undetected
- PersonThe accountant books bulk transfers onto a suspense account that swells until year end
Why the current process costs more than it appears
The bill that never shows up in a budget.
- Days of manual reconciling a month is a staff job done in the owner’s evenings.
- Unreconciled lines are real money: double refunds and wrong fees surface only in line-level matching.
- A suspense account growing all year returns at close as archaeology billed at the accountant’s rate.
- A shop without reconciliation does not know whether the gateway settles it honestly; it trusts instead of knowing.
Cost of inaction
The second row is a cautious model: e-commerce settlement audits almost always find fee and refund errors worth a fraction of a percent of turnover; at several hundred thousand euros a year that is thousands. Only line-level reconciling detects them, because in bulk totals they vanish.
The model does not price the accountant’s calm or a year-end close without archaeology; anyone who has untangled a suspense account once will price that themselves.
A model organisation with realistic proportions – the numbers exist so you can run the same maths on your own data; they are not a client result.
An online clothing boutique: around 900 orders a month through its own shop (two gateways) and a marketplace, returns in the low teens of percent, an external accountant, Microsoft 365.
Manual reconciling once a month, two days of work; a “to be clarified” folder gaining a few lines monthly.
Fees taken on faith; refunds spot-checked; the accountant’s suspense account growing.
The owner knows revenue “by inflows”, not by orders; per-channel margin is guesswork.
The robot fetches the gateways’ and marketplace’s transaction reports, the shop’s orders and the statement daily; reconciles three ways: order to transaction, transaction to payout, payout to statement; settles fees and refunds to the line; raises discrepancies as cases with context; the accountant’s statement generates in her format.
In the modelled case reconciling falls from two days to a quarter-hour case review, and fee errors and double refunds surface on the day they arise. Model numbers, not the shop’s records.
Proposed solution
We start with a map of the flows: which gateways, which marketplace, what each one’s payout cycle looks like, and what the accountant needs. Every source has its report format; the robot learns them once and reads them nightly from then on.
The reconciling is three-way: every shop order links to a transaction in the gateway report, the transaction to a bulk payout, and the payout to a line on the bank statement. Fees are computed to the line and compared with the rates in your agreement; refunds link to their original orders. Whatever does not close does not land in a “to be clarified” folder; it becomes a case with context: what, how much, from which day, what is missing.
The accountant receives a monthly statement in her format: revenue per channel, fees as a cost, refunds, transit balances; the suspense account ceases to exist. The owner sees in Teams what he never saw: the real revenue and payment cost per channel, week by week, and a case list that usually holds a few items, not a few dozen.
UiPath Orchestrator: nightly fetch schedules, the reconciliation queue, retries and an audit trail; UiPath Integration Service connectors for Microsoft Teams and Outlook 365
Gateway and marketplace report reading, three-way line-level reconciling, fee settlement against agreements, refund linking, cases with context and the accountant’s statements
Gateway and marketplace reports (API or export), shop orders (API or export), the bank statement; the reconciliation register in SharePoint
How the automated process works
- AutomationThe robot fetches gateway reports, shop orders and the statement nightly
- AutomationThree-way reconciling: order, transaction, payout, statement
- AutomationFees computed to the line and compared with the agreement’s rates
- SystemRefunds link to their original orders; doubles surface at once
- PersonDiscrepancies become cases with context; the review is a quarter hour, not two days
- SystemThe monthly statement for the accountant generates in her format
Human-in-the-loop model
Automation handles
- Nightly fetches, three-way reconciling and line-level fee settlement
- Refund linking and discrepancy detection with cases
- The accountant’s statements and the per-channel revenue view
People decide
- Resolving the cases: a claim to the gateway, an order correction, a write-off decision
- The gateway and marketplace agreements; the robot compares rates, a human negotiates
- The statements’ format and scope, agreed with the accountant
Before and after
Systems and integrations
The stack is short on purpose: one engine, one execution layer, one place where a person decides.
Inputs
- gateway and marketplace transaction reports
- shop orders (API or export)
- the bank statement
- fee rates from the agreements
Automation layer
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
Target systems
- the line-level reconciliation register
- discrepancy cases with context
- the accountant’s statements and per-channel view in Microsoft Teams
Human touchpoints: the case review in Microsoft Teams; the monthly accountant statement; a fee rate review at agreement changes
Technologies used
nightly schedules, the reconciliation queue, retries, a trail for every line
Acases, statements, reports to the accountant
Athe reconciliation register with history and permissions
Atransaction reports via API or export; popular gateways have APIs
Borders via API or export; popular platforms have both
BIllustrative economic model
Numbers you can check against your own data.
The model assumes sources with reports available via API or export, which is standard in popular gateways and platforms. Volumes and rates belong to the scenario; your own numbers go into the calculator alongside.
Run the maths on your data
An illustrative estimate based on your inputs. It models freed capacity, not promised savings.
Business benefits
- Reconciling happens nightly; the review takes a quarter hour instead of two days
- Fees are checked against the agreement, not taken on faith
- Refunds stop vanishing and doubling in the settlements
- The accountant receives statements that book themselves; the suspense account disappears
- The owner sees real revenue and payment cost per channel, weekly
The management view
- Payment settlement becomes a controlled process, not an act of faith in the gateway
- Every line carries its full chain: order, transaction, payout, statement
- A new sales channel is a new report format to learn, not two new days of work
Board-level KPIs
lines reconciled automatically · open cases and their resolution time · detected fee and refund errors · monthly reconciling time · the age of the oldest unreconciled line
Security and governance
The automation has exactly the permissions it needs. Not one more.
- The robot works on transaction data: order numbers, amounts, fees, dates; customer data at a minimum
- Gateway and platform access lives in a credential vault, in report-only mode, with no right to execute operations
- Every reconciled line has a trail: sources, time, result
- The accountant’s statements contain what is bookable: no surplus customer data
- Data stays in your Microsoft 365 tenant; the robots run in the EU region of UiPath Automation Cloud
Why now
Payment channels multiply: gateways, instalments, marketplaces; every integration is a new report format and a new slice of manual work.
Fees grow and grow more complex; checking them to the line stopped being a luxury and became margin hygiene.
Digitised reporting demands order in the revenue breakdown; the suspense account no longer gets away with it.
Relevant executive roles
Knows what each channel really earns, and gets two days a month back
Receives broken-down, tied-out statements instead of bulk transfer riddles
Resolves a short case list instead of diving into three files
Common questions and objections
A platform report usually sees its own slice: the shop’s transactions, without the gateway’s bulk payouts, line-level fees or the statement. Three-way reconciling ties all three ends together; the platform report is one of its sources, not the answer.
That is exactly the problem the robot solves: each format is learned once, then read nightly. People are no worse than a robot at reading reports; they are worse at doing it every night, without errors, across three formats at once.
The access is report-only: the robot reads transaction reports, executes no operations and touches no payouts. Credentials live in a vault, and every fetch leaves a trail. It is the same access level the person doing manual reconciling has today, minus the files copied to a desktop.
When this is not the right solution
- A shop with a few dozen orders a month on one gateway: a careful spreadsheet suffices
- Sales purely by cash on delivery and direct transfer: the bulk payout problem does not exist
- Expecting the automation to litigate with the gateway: it prepares the claim with evidence, a human leads the conversation
A question for the next management meeting
How many lines sit in our “to be clarified” folder, and how many of them are money we simply let go?
Implementation approach
Scope without ambiguity, before anything is signed.
We deliver
- Report reading for all gateways and marketplaces
- Three-way line-level reconciling with a register
- Fee settlement against agreements and refund linking
- Discrepancy cases with context and the accountant’s statements
- The per-channel revenue view and two weeks of parallel running
We need from you
- Report-level access to the gateways, marketplace and shop platform
- The fee rates from your agreements, for comparison
- The statement format agreed with your accountant
Stages
Discovery
Channels, payout cycles, report formats, the scale of “to be clarified”
Rules
Linking rules, fee rates, the accountant’s format, case thresholds
Build
Fetches, reconciling, fees, refunds, cases, statements
Parallel run
One full monthly cycle beside the manual one; results compared
Go-live
Reconciling moves to the robot; a rate review after the quarter
A quick win. Effort depends on the number of sources and their interfaces; popular gateways and platforms have APIs or decent exports, so the first source runs within a week.
The gateway payout: €8,214.37. Orders in the period: one hundred and forty-two. Refunds: eleven. The fee: “as per agreement”. Good luck.
Send us one payout’s report and the order export for the same period. We return a trial reconciliation: what ties out, what does not, and how many lines would need a case. It is usually the shortest road to a decision.
Check your “to be clarified” folderThe neighbouring process usually has the same problem
On the fifth the accountant asks about missing invoices. The inbox hunt begins.
See the solution Finance & accountingInvoices chased before they turn difficultThe client pays after 40 days because nobody reminded after 14. The cash sits with others.
See the solution Customer serviceOne inbox instead of four channelsEmail, Messenger, phone and a form. Four places where the customer could ask the same question.
See the solutionIndustries where we deploy this most oftenSmall business & services