Part of The company brain: where your organization's knowledge lives
Effective data lifecycle management
Data has a life: created, used, retained, removed. Skip a phase and you pay twice - in storage, and in wrong answers.
Data lifecycle management means managing data across its whole life: from the moment it is created to the moment it is deleted. Not as a clean-up afterwards, but as an agreement up front about who is responsible in each phase.
It sounds like admin. In practice it is the reason a report does or does not hold up.
The phases
Every phase has an owner and a decision. Skip one and the problem moves to the next.
Creation. A scan at the gate, an order in the ERP, a reading in a cold store. Quality is decided here: what goes wrong at this point cannot be repaired later, only masked.
Capture and enrich. The data gets a home, a definition and a lineage. Without that step every later analysis is a reconstruction.
Use. Reporting, planning, a model, an agent. This is the phase everyone sees, and usually the only one that gets attention.
Retention. What must you keep because the law says so, and what do you keep out of habit? That is a decision, not a technical detail.
Disposal. Delete or anonymise, provably and on time.
Why it goes wrong
Almost every organisation is good at phase three and weak everywhere else. Everything is collected, nothing is ever thrown away, and nobody owns the meaning. The result is familiar: three versions of the same revenue figure, a data lake nobody can navigate, and a privacy obligation met on paper only.
The cost is double. You pay for storage nobody uses, and you pay for decisions taken on the wrong version.
Who owns what
A phase without a name is not a phase.
The practical test is simple: can you name a person for every phase? Not a department, a person. If you cannot, that phase is unowned, however much policy has been written about it.
How to start
Take one dataset that genuinely matters - stock position, delivery reliability, customer revenue. Walk the five phases and note the owner, the retention period and the definition for each. Record it as a data contract so it becomes an agreement instead of an assumption.
Then scale it out per domain, exactly as in a data mesh. The technology underneath - catalogue, lineage, retention rules - belongs in the platform, not in people's heads. What that layer looks like is described on the engine.
Evidence
These claims do not stand alone. They lean on our own research, which we keep updating.
About the publisher
Datahub
Datahub editorial team
Pieces without a personal byline are written and reviewed by the Datahub team. We build governed data foundations for logistics, retail and manufacturing, and only publish figures we measured ourselves or read in a primary source.
Why this source
- Every publication is reviewed before it goes live
- Figures follow the methodology at /research/methodology
Writes about: Data foundations · AI readiness
More about the team
