Moving an archive from paper to the cloud

The most common failure in a paper-to-cloud project is not a technical one. It is arriving at the end with a correctly scanned, correctly indexed archive sitting in a repository that nobody uses, because the destination was chosen after the scanning started and the structure was inherited from a filing cabinet.

Every decision below is cheaper to make before capture than after.

Decision one: what actually goes

Start by sorting the archive into four groups, because they get different treatment and different budgets. Records within retention that are actively used. Records within retention that are dormant. Records past retention, which should be disposed of under your own policy rather than migrated. And records whose status nobody knows, which is usually the largest group and needs a decision-maker rather than a scanner.

Migrating everything because sorting feels harder is a false economy. You pay to convert it, you pay to store it, and you inherit an obligation to manage and produce it.

Decision two: where it lands

Most organisations end up with two of these: a working repository for records people use, and cheaper storage for a preservation copy. That is a reasonable outcome as long as it is chosen rather than accumulated.

Decision three: the metadata model

This is the migration. Everything else is logistics. The metadata model determines whether the archive answers questions in five years or becomes a searchable pile with better lighting.

1. 1List the questions people will ask of the archive, in plain language. Find every invoice from this supplier. Show me the signed lease for this unit. Produce all HR records for this employee.
2. 2Derive fields from those questions. Each field must serve at least one question, and each question must be answerable by some combination of fields.
3. 3Fix the format of every field before capture: date format, whether reference numbers keep leading zeros, whether names are captured as recorded or normalised. Inconsistency here is what makes searches fail later.
4. 4Decide which values come from controlled lists rather than free text. Free-text department names will produce six spellings of the same department.
5. 5Decide the retention class per record category now, so retention can be applied at load rather than retro-fitted across a million files.
6. 6Keep it as small as it can be while answering the questions. Every additional field is capture cost and an opportunity for inconsistency.

Do not replicate your existing folder tree as metadata. A folder path encodes one hierarchy, usually the one that suited the department that built it in 2009. Fields let a document be found down several paths at once, which is the entire advantage you are paying for.

Decision four: who can see what

Paper archives are secured by the storeroom door. Digital archives are secured per record, and migration is the moment that model has to be designed. Work out the permission groups before the load, map them to record categories rather than to individuals, and decide what happens to material that is currently secured only by obscurity. HR files, legal correspondence and board papers that lived in a locked cabinet will become instantly searchable, and that is a change with consequences if nobody planned for it.

The migration itself

1. 1

   Pilot one category end to end

   Prepare, capture, index, load and then use it. Have real staff search for real documents in the destination system. Every wrong assumption in the metadata model surfaces here for the price of one category.

2. 2

   Freeze the rules

   Fields, formats, splitting rules, naming, permissions and retention classes are signed off. Changes after this point are reprocessing, and everyone should know that before they ask for one.

3. 3

   Capture and index in batches

   Production runs by category, with quality control on both image and index data. Batch boundaries should match something meaningful, so a failed batch can be reprocessed without disturbing the rest.

4. 4

   Load and verify

   Files are loaded with metadata attached, then verified in place: counts reconciled against the manifest, a sample of records opened and checked against the paper, permissions tested by logging in as an ordinary user rather than an administrator.

5. 5

   Reconcile completeness

   Box manifest counts against loaded document counts, with every discrepancy explained rather than absorbed. This report is what lets you state later what the archive contains.

6. 6

   Cutover

   A dated point after which the digital record is the one people use. Announce it, retire the physical retrieval process, and remove the shortcut back to the storeroom. Without this step, both systems run in parallel indefinitely and the digital one stays incomplete.

What happens to the paper

There are three honest options and one bad habit. The options are secure return to you, transfer to offsite physical storage, or certified destruction. The bad habit is leaving the boxes exactly where they were, which means you now pay for storage and cloud licensing and have solved nothing.

Destruction deserves a slower conversation than it usually gets. Some record categories must be kept in original form, and that determination is one for your own legal or compliance adviser rather than for a scanning supplier. What a supplier should give you is the evidence you need to make the decision safely: a completeness reconciliation, a quality report against a defined standard, and a certificate of destruction listing exactly what was destroyed and when, once you instruct it.