Is digital continuity the same as keeping old files?

No. The retained information must remain usable for its required purpose.

Must the original software always be preserved?

Not necessarily. Migration or other supported approaches may meet the usability need, but the result must be checked.

Does continuity guarantee exact AI reproducibility?

No. Reproducibility can depend on additional technical and methodological conditions that must be identified for the specific task.

AI Glossary / Data governance and quality / Intermediate

Digital continuity

Definition

Digital continuity is the ability to keep digital information usable for the purposes and period for which it is needed. It requires more than keeping a file's bytes intact. People must be able to locate and open the information, interpret its meaning, work with it appropriately, and establish enough confidence in its identity and history. The concept is particularly important when systems, organizations, formats, or responsibilities change.

What it means

A file can survive a migration yet become practically useless. Its format may require unavailable software, its labels may lose their definitions, or its relationship to other records may disappear. Digital continuity focuses on the required use of information through change. It overlaps with preservation and continuity planning, but does not mean retaining every digital item forever or keeping every old application running indefinitely. The intended use and retention requirements define the practical objective.

How it works

Identify the information needed for important activities and describe how it must remain usable. Record the relevant context, metadata, relationships, software dependencies, and responsible owners. Before a migration or organizational change, assess what could be lost and define acceptance checks. Afterward, test actual use with representative information rather than counting files alone. Maintain those checks as requirements change, and make documented decisions about migration, preservation, access, and disposal. Technical teams and information owners need to work together.

Three practical examples

System replacement: a customer-record migration preserves the links between cases, attachments, dates, and responsible teams.
Research records: an archive retains the units and code definitions needed to interpret a dataset after its original analysts leave.
AI reproducibility: a project preserves the model artifact together with preprocessing information, relevant dataset references, and evaluation context needed for the intended later review.

Think of it this way

Think of keeping a recipe along with its ingredient names, measurement units, and instructions for the available equipment. Saving only a list of numbers may preserve the writing but lose the ability to cook the dish. The analogy is limited because digital information can depend on complex software, permissions, and linked records, and future uses must be assessed rather than assumed.

Worked example

An original migration exercise moves a dataset with 10,000 observations. The destination contains all 10,000 rows and matching checksums for the exported files. However, the source's status-code dictionary and measurement-unit field are omitted. Analysts can open the table but cannot reliably distinguish a status of 1 from 2 or determine whether a distance is measured in meters or kilometers. The file checks pass, but the intended analysis fails. The acceptance plan therefore includes interpreting selected records against their metadata, tracing relationships, and reproducing an agreed calculation. Restoring the missing dictionary and units addresses this specific gap; it does not eliminate the need to assess other dependencies.

Illustrative case study

Fictional example: Willow Research retains an older model file to review a past decision. The file opens, but the team cannot determine which text-normalization rules were applied to its inputs. It locates the versioned preprocessing configuration and adds it to the preserved package with a clear relationship to the model. This invented case shows that a technically readable artifact can remain unusable for a particular purpose until its context is recovered.

Common misconceptions and mistakes

Do not equate digital continuity with a successful backup job, an unchanged checksum, or a complete file count. Those checks answer useful but narrower questions. Do not assume converting everything to a readable image preserves the ability to search, calculate, or analyze it as required. Also, an accessible format does not automatically preserve meaning, relationships, or evidence of provenance. Avoid retaining material indefinitely by default; continuity concerns the information and usability that are actually required.

Business use

Digital continuity supports reliable handovers, system changes, and later accountability. For AI projects, it can prevent a model archive from becoming a collection of unexplained weights, labels, and spreadsheets. Define the future review tasks in practical terms: inspect a prediction, understand a label policy, or reproduce a specified evaluation within known limits. Assign an owner to each dependency and test representative packages. This gives the business a clearer measure of success than storage volume or migration completion alone.

Check your understanding

Every file in a project archive opens, but no one knows what its numeric labels mean. Has the archive met a requirement to interpret the past analysis?

Answer: No. The missing label definitions are part of the context needed for that use, even though the files remain readable.

Related terms

Metadata
Data lifecycle
Data integrity
digital preservation
provenance
format migration
reproducibility

Learn next

select one dataset you may need in two years and write a short future-use scenario. List the files, definitions, permissions, and software needed to carry it out. Ask a colleague unfamiliar with the original project to attempt the task using the preserved package and record what is missing.

Sources

Sources: The National Archives, Understanding Digital Continuity, February 2017, https://cdn.nationalarchives.gov.uk/documents/information-management/understanding-digital-continuity.pdf , introductory and usability sections read; What is digital continuity?, https://www.nationalarchives.gov.uk/information-management/manage-information/policy-process/digital-continuity/what-is-digital-continuity/ , search excerpts consulted because direct access returned 403; Digital cataloguing practices, https://www.nationalarchives.gov.uk/professional-guidance-and-services/cataloguing-practices/digital-cataloguing-practices/ , relevant excerpt consulted. These sources ground the continuity concept; the AI applications and examples are original synthesis. Historical legal passages are not adopted as current legal advice.

Previous: Digital cloning Next: Digital curation