Bank and imports

The bank is the witness. Once connected through Plaid, every posted transaction on that account flows in: deposits land as revenue with the payer worked out where it can be, and card and bank charges land as expenses already sorted into categories. DeskOps also uses the bank to confirm that money a processor says it paid out actually arrived, and to match each payout to a bank line. The connection row shows the window of dates it covers and any gap it knows about, so "the bank says" is never quietly "the bank said, until three weeks ago". Bank connection is the single largest source of transactions in DeskOps today.

One bank connection per business today. Connecting a second bank replaces the first rather than adding to it — the first connection is removed. If you run a second business from a separate account, that business is a separate DeskOps business with its own connection. Accounts within one bank login can be excluded individually, so a personal account under the same login stays out.

A statement CSV covers what a connection cannot: a bank Plaid does not support, or the days a paused connection missed. Every dropped file is recognised by its own columns, so a Chase or Capital One export is read without you naming it, and other banks' layouts are mapped from their headers. Lines the connection already holds are matched and counted once, never twice, and the coverage line on the connection closes the gap it filled. Statement deposits are not attributed to a payer the way connected ones are, and re-downloading next month's statement is on you until an email-forward intake exists. Early access: no real customer's statement has been imported yet.

When a processor like Stripe or Square pays you out, the bank sees one lump deposit that is really several customer payments already counted when they were paid. Counting the deposit again would double your revenue, so DeskOps files it as a processor payout and matches it to the payments inside it. What counts as revenue is the customer's payment, once, at the moment it was taken. A deposit DeskOps cannot place — no payer it recognises, no processor behind it — is shown for you to name rather than guessed at.

When DeskOps matches a processor payout to a bank deposit and the arithmetic works out but a person should confirm it, the match waits in a short queue with the sums in plain words — these payments, minus these fees, equals this bank line. You confirm it or reject it; a rejected line is freed for other matches. Undoing a confirmation never deletes anything: the history reads "confirmed" then "returned", both true, in order. Owner-only, because bank settlement is the same class of information as the Wealth page. An empty queue is the normal state.

One drop zone, any file. Export from the old system as CSV, drop the files in, and each one is recognised on its own — Jobber, Housecall Pro, ServiceM8, Workiz, FieldPulse, Kickserv, Service Fusion, GorillaDesk, Joist and Tradify exports are known by their columns, and an unfamiliar file is mapped from its headers with you confirming the columns. You see a preview before anything is written: how many will be added, and what will not be counted twice because your books already hold it. Then the import runs in chunks, can resume if the connection drops, and can be undone for seven days. Customers, jobs, paid invoices, estimates and recurring plans all come in; DeskOps never messages your customers because of an import.

A recurring schedule in the export becomes a service plan in DeskOps when its repeat rule is one DeskOps can read — weekly, every two weeks, monthly, quarterly or yearly. A rule it cannot read is named on its own row in the preview rather than silently imported as a one-off, so nothing quietly stops recurring. Plans book their visits onto the schedule like any other.

One screen for the whole move: what came in, what still needs you and where to fix it, and next week as it now stands in DeskOps. It reads the most recent sitting of imports — everything committed in the last 24 hours — and only dated jobs and understood recurring schedules appear in next week; anything the importer could not read is listed under What needs you. The undo for the whole batch lives here for seven days. After an undo the screen stays put and says so, and you choose whether to import again.

Two records merge on their own only when two real email addresses agree. A customer who booked through one system and paid through another under two addresses stays as two files — DeskOps will not guess that a shared phone number means one person, because spouses and offices share phones too. So it asks. "Same person" merges the records and moves their payments onto one; "different people" is remembered so the pair is never suggested again. Suppliers get the same treatment on the Expenses page.

Books import brings in what your accounting system already holds — invoices, payments, expenses — and DeskOps shows what agrees with your processors and bank, what conflicts, and what is missing on the supported path. Nothing is counted twice: a Stripe payment that QuickBooks also recorded is one payment in DeskOps. Books are an importer, not a requirement; a business with no accounting system connected loses nothing, because DeskOps keeps its own records.