Short answer. Archiving is not tidying: it is deciding, for each document, its name, its record, its place and who may see it, then being able to search inside its text. Start with a test: ten documents drawn at random, a stopwatch. Then four moves, in this order: a filing plan of ten to fifteen types; a naming rule, with the date first, written year-month-day; paper sorted into three piles, only one of which is scanned; search inside the text, including scanned pages. A copy outside the building, a named owner, and the test run again at the end. Eight weeks in the example.
This page is for the manager whose cabinets and shared folder have grown without a rule, and for the founder based in France who needs, the same day, a contract or an invoice stored somewhere in Algeria. The definitions come from public sources: document management system, metadata, optical character recognition, full-text search, the PDF/A format, the ISO 8601 standard, the 3-2-1 rule, digital preservation, and the documentation of a recognition engine for the advised resolution. The company of 25 people, its six cabinets, its ten documents, its 40,000 pages and its calendar are a constructed example. This page gives no legal keeping period and does not cover the standing of a digital copy: those rules depend on the type of document, so ask your accountant or adviser. Everyday file sharing, server, NAS or cloud, is the subject of another guide.
Ten documents to find again: the test before any decision
The figure shows the same test, before and after.
The test. Write ten documents on ten slips of paper: a supplier invoice for a given month, a contract, a signed delivery note, a report, a warranty. Draw them one by one, start a stopwatch, and let the person who usually searches do the searching. Note the time, and what is not found. In the example: seven out of ten, 78 minutes in all, and three documents never found, including a machine warranty. What the test says. Where the time goes: in the cabinets, in the shared folder, in mailboxes. And what is missing: a document that cannot be found is a document to redo, to ask for again, or a lost warranty.
What an archiving tool is. A document management system is a computerized system used to store, share, track and manage electronic files. It typically stores, for each document, metadata, data that describes other data: the date it was stored, the identity of the user who stored it. But the tool comes last. The order. First the test, then the filing plan and the naming rule, which take a week of thought and no purchase; the tool only applies what has been decided.
A file name that can be found again
The figure shows the same document under four names.
The date first. In the ISO 8601 standard, date values are ordered from the largest to the smallest unit of time, year, month, day, so that the lexicographical order of names corresponds to chronological order. A folder of invoices named this way sorts itself. The date is that of the document, not of the day it was scanned. A closed list. Ten to fifteen types, no more, always written the same way: invoice, contract, quote, delivery note, letter, report. If two people hesitate between two words, the list is too long.
The third party and the number. The customer or supplier code is the one in the management software: a single spelling, and a possible link between the archive and the software. The number is the one printed on the document. Length. Keep names short and folders shallow: many Windows applications are limited to 260 characters for the full path of a file, folders included. The version. A figure, never an adjective; the full history is the tool's job, which records each time a document is checked out and back in.
«A good name is read without opening the file.»
An archive, in five layers
The accented layer is the record.
Search inside the text. Optical character recognition converts the image of a text into text: a scanned page becomes searchable by its words, and no longer only by its name. Full-text search refers to the techniques that search the content of the documents themselves. For holdings in French and Arabic, recognition must read both: the open-source engine Tesseract, for example, announces more than a hundred languages, Arabic among them; quality is checked on your own pages, stamps and handwriting included. Searching in two languages raises the same questions as search in an application.
Rights. An employment contract, a bank statement and a delivery note are not shown to the same people. Rights are given by type of document, not folder by folder as you go, following the principles of the guide to roles and the audit log. Time. Digital preservation aims to keep information accessible and usable in the long term. Two rules: a format made to last, PDF/A, the ISO-standardized version of PDF for archiving, which prohibits what hinders future reading and requires fonts to be embedded; and the 3-2-1 rule, three copies, on two types of storage media, one of them off-site, as in the guide to backups.
Scanning the paper, in six steps
The first step is accented: sort.
Three piles. Active paper, still consulted: it is scanned and indexed. Paper to keep without consulting it: it goes into numbered boxes, and it is the box that is indexed, not each page. Paper that no longer has to be kept: it leaves, following a retention schedule, the document that says how long each type of record must be kept before it can be destroyed. Those periods are not invented: ask your adviser. In the example, a third of the six cabinets is active: 40,000 pages.
Quality. The recognition engine mentioned above states that it works best on images of at least 300 dots per inch: below that, the text is read badly and search fails silently. The pace. In the example, two people prepare, scan, index and check 2,000 pages a day: four weeks for the active pile. The check. One batch in ten is reopened: missing pages, pages upside down, unreadable text. The original receives its box number, written in the record: the day the paper is needed, you know where it is. Going further. Reading the amount or the number of a scanned invoice automatically belongs to extraction by AI, useful when volumes justify it.
«You do not scan a cabinet: you scan the active pile.»
A well-kept shared folder, or an archiving tool?
The figure compares the two, on six points.
Often, the folder is enough. With a filing plan, a naming rule and rights by folder, a well-kept shared folder settles most searches in a small company. It is the first level, and it requires no purchase. When the tool is justified. When you must search inside the text of scanned pages, give rights by type of document, keep the history of versions, or know who consulted what. That is what the index is for: it exists mainly to support query and retrieval.
Three families. A document management package on the market; the library functions of the office suite you already use; or an archive module added to your management application, when documents must open from the customer's or the order's record. The neighbours. The archive is not the intranet, where current notes and procedures are published, described in the guide to the intranet; nor the customer area, where each customer finds their own documents, described in the guide to the customer area. It can feed both.
Eight weeks, from the first test to the second, in 2027
Eight weeks in the example, the first for the test and the inventory.
- What the manager receives: the result of the test, before and after; a filing plan of ten to fifteen types; a naming rule on half a page; the shared folder taken up; the active paper scanned, read and indexed; the numbered boxes; rights by type of document; a copy outside the building.
- What the company no longer does: spend half a day looking for a contract, redo a document it already has, name a file “final (2)”, scan a whole cabinet without sorting it, leave employment contracts readable by everyone.
- What is measured: documents found out of ten and time per document, pages scanned and checked per day, files that follow the naming rule, documents with no record, date of the last off-site copy.
- What Symloop does: the test and the inventory, the filing plan and the naming rule with your teams, taking up the shared folder, organising scanning and text recognition in French and Arabic, search, rights, the off-site copy, then, if the six points justify it, setting up a tool or an archive module connected to your application.
Digital first, paper at the same time. The shared folder is taken up in batches: the current year's folders first, then going back. Files are renamed, moved, duplicates removed; what has not been opened for years goes into an “old” folder, without being renamed. The active pile of paper is scanned meanwhile. The rule, from the start. Every new document follows the rule from the second week: otherwise the stock to take up grows while it is being tidied.
The second test. The same ten documents, the same stopwatch, the same person. In the example: ten out of ten, under six minutes in all. What makes it last. A named owner, who keeps the list of types and settles hesitations; the rule displayed; and the test run again twice a year, with ten other documents.
- How do you archive documents?
- In this order: a test of ten documents to measure, a filing plan of ten to fifteen types, a naming rule with the date first, rights by type of document, search inside the text, and a copy outside the building. The tool comes after these decisions.
- How do you archive paper documents?
- In three piles. Paper still consulted is scanned, read by character recognition and indexed. Paper to keep without consulting goes into numbered boxes, indexed by box. The rest leaves, following periods to ask your adviser for.
- How do you archive digital documents?
- With a built name, type, date, third party, number, version, a folder per type, and a lasting format such as PDF/A for what must be read again in ten years. Take up the current year first, apply the rule to every new document, then go back in batches.
- How should files be named?
- The date first, written year-month-day, then the type, the third party's code, the document number and the version in figures, separated by underscores, with no accent or space. Example: 2027-03-14_invoice_customer-0418_F-2291_v2.pdf.
- Must everything be scanned?
- No. Scan what is still consulted. The rest is kept in numbered boxes, with the number written in the index, and scanned on demand the day someone needs it. Sorting before scanning divides the work by three in the example.
- Can the paper be thrown away once scanned?
- It depends on the type of document and on rules outside this page: ask your accountant or adviser before destroying anything. By default, keep the original in a numbered box, with the number written in its record.
- How long should documents be kept?
- This page gives no period: they depend on the type of record, and it is for your adviser to set them. What the archive must do is carry a period per type of document and flag what reaches its term.
- Do we need archiving software, a document management system?
- Not always. A well-kept shared folder is enough for many small companies. The tool is justified when you must search inside the text of scanned pages, give rights by type of document, keep the history of versions or know who consulted what.
- Can a scanned document be searched, in Arabic and in French?
- Yes, if the pages were read by optical character recognition in both languages, from images of at least 300 dots per inch. Check the quality on your own pages: stamps, tables and handwriting are read less well.
- How long does an archiving project take?
- Eight weeks in this page's example, for 25 people: one for the test and the inventory, one for the plan and the rule, then taking up the digital files and scanning 40,000 pages in parallel, and a second test to finish.