The volume and variety of documents that a cross-border business generates — and how quickly they become unmanageable when you have no system for them — is something nobody in the startup guides mentions. It is unglamorous, it is not a strategic problem, and it does not appear on any pitch deck. But it costs real time and real money when it goes wrong.
What documents actually look like
In the first six months, we received documents from: the German commercial register (Handelsregister), two different notaries, our German tax advisor, the Finanzamt (twice), our German bank, our Spanish bank, two municipal councils in Andalusia, the Spanish tax authority (Agencia Tributaria), three utility companies, four contractors, our insurance broker, and a property management company we eventually fired.
Some arrived as PDFs by email. Some came by post as certified letters. Some were handed over at notary appointments in paper form only. The property purchase contract alone was 78 pages long, in Spanish, with an addendum in German prepared by our lawyer.
We stored them in a shared cloud folder, organised — initially — by date. Then by sender. Then by subject. Then, honestly, by nothing, because we kept disagreeing about how to structure it and the documents kept arriving faster than we reorganised.
The specific problem with multilingual documents
When documents arrive in two languages — and in two different legal and cultural registers — manual categorisation becomes surprisingly difficult.
A Grundbuchauszug (land register extract) is a property document. A nota simple (its Spanish equivalent) is also a property document. They look entirely different, are written in different legal styles, and the relevant information appears in different places. If you are scanning and filing, you need to either know this intuitively or look it up. Every time.
Invoices in Germany include the VAT breakdown on the face of the document, in a predictable format. Spanish facturas follow a different convention and often include information that in Germany would appear elsewhere. Matching these for accounting purposes is tedious even when you are fluent in both languages.
Then there are the documents that are themselves bilingual — a contract prepared by a cross-border law firm, a letter from a dual-jurisdiction notary, an insurance policy with German terms and Spanish addenda. These resist any categorisation scheme based on language alone.
What we tried before finding a system
We tried a shared filing convention in Notion. It lasted about three weeks before our virtual assistant created a category called "miscellaneous" that eventually contained 40% of everything.
We tried tagging by date and sender only, accepting that categorisation would happen at tax time. Tax time was unpleasant.
We tried splitting the work by language: I handled the German documents, my co-founder handled the Spanish. This worked for incoming documents but fell apart the moment anything required cross-referencing — which it frequently did, because the properties generated documents in both languages simultaneously.
We tried a virtual assistant who specialised in document management. She was good. She was also expensive, and she was not available on the day the Finanzamt letter arrived requiring a response within ten working days.
The actual cost of not having a system
The costs are mostly invisible until they are not.
One invoice was paid twice because neither of us could quickly find the original and assumed the other had not yet paid it. That was €340 lost to a refund process that took six weeks.
One contractor dispute required us to produce a written quote that had been emailed, printed, signed, scanned, and emailed back. Finding the signed copy took ninety minutes across four different storage locations.
At our first tax filing, our accountant billed us for two additional hours of "document organisation" — her polite description of reconstructing Q1 from the fragments we sent her.
None of these were disasters. Collectively, they cost about €800 and a significant amount of goodwill with our accountant.
How we changed the workflow
What eventually worked was treating first-pass categorisation as a scanning problem rather than a filing problem. The insight was that the effort of deciding where a document goes should happen at upload time, not at retrieval time — and that the upload-time decision does not need to be perfect, just directional.
We started scanning everything on arrival — letters, printouts from notary appointments, even handwritten notes — and uploading in weekly batches. The categorisation output gave us a first-pass label (contract, invoice, property document, bank correspondence) that we then used to route each document into the correct shared folder. The OCR preview meant we could do a five-second sanity check without opening each file individually.
The multilingual handling was better than we expected. German and Spanish documents were categorised by content, not by language. A Spanish factura and a German Rechnung both landed in the invoice bucket.
What we would do differently from day one
Build the document system before you need it. Not before the first documents arrive — they will arrive before you are ready — but before you have more than a month's worth of backlog. A month of backlog is a manageable afternoon. A year of backlog is a project.
The physical originals still exist — many are legally required to. But we know what we have, and we can find any document without calling each other to ask where we filed something in October.
That alone, in a business with a distributed team and documents in two languages, is worth more than we expected when we started.