The problem arises when information loses its correspondence, validity or context as it moves from one system to another.
A company may have reliable data, well-integrated applications and controls over its processes. Yet the result can still be incorrect if the information being used no longer corresponds to the operation it is intended to support.
Consider an order whose delivery address changes after it has been confirmed.
The sales system records the change. The warehouse prepares the correct products. The logistics provider receives a valid address. But that address belongs to an earlier version of the order.
Each system has processed recognisable information. The problem lies in the relationship between them: the address being used no longer corresponds to the current state of the operation.
This situation illustrates an aspect of data integration that deserves particular attention. The quality of the individual elements matters. So does their ability to retain their meaning as they move from one process to another.
Information needs to preserve its relationships
When analysing a data, analytics or AI solution, we tend to identify its components: information sources, applications, models, connectors, providers and existing systems.
That inventory tells us what is involved.
But to understand how a result is produced, we also need to examine the relationships connecting those elements.
A query must retrieve information about the correct customer. Data must remain valid for the relevant point in time. An operation must use the rules applicable to its current state. An update must reach the system responsible for executing the next step.
These are different conditions, and each can affect the validity of the result.
In March 2023, OpenAI documented a ChatGPT incident involving the reuse of cached connections. Certain requests could receive information associated with another user. In some cases, the data was of the expected type and appeared valid.
The fix introduced additional checks to ensure that the retrieved information corresponded to the requesting user.
The documented mechanism concerned data and connection management, rather than the model’s reasoning.
The case illustrates why validating the format of a response is insufficient when we also need to verify who the information belongs to and which operation it is valid for.
Time is also part of the data
Let us return to the order.
The previous address may be correct as a historical record. The problem arises when it is used to execute a delivery for which a different address has already been authorised.
This introduces an important distinction for any organisation working with distributed information: data can be accurate and still be outdated for its intended use.
In retail, banking, insurance and services, different systems hold information about the same operation. Each may be updated at different times and serve different needs.
That is why the analysis must consider which version of the data corresponds to each stage, when a change becomes valid and what confirmation the next system requires.
Technical synchronisation helps transmit information. Process coherence also requires preserving the conditions that determine which information should be used.
The validity of data over time depends on the operation that needs it.
When multiple capabilities share the same dependency
Relationships can also connect services that appear independent from a business perspective.
In June 2025, Cloudflare documented an outage affecting Workers KV, a storage service used by different functions across its platform.
The incident affected services related to identity, configuration and AI functionality. Different products depended on shared information to complete certain operations.
The case shows how a configuration, an access policy or a shared resource can participate in several operational workflows.
For an organisation, understanding this convergence helps identify where the same condition supports different processes.
An application inventory can show which resources exist.
Analysing their relationships explains what they need to share, preserve or confirm in order to operate coherently.
From system inventories to the data journey
Many organisations already use enterprise architecture tools, data catalogues, data lineage, configuration inventories and service maps.
The challenge is to connect that knowledge to the result the business needs to achieve.
In March 2026, the UK’s Financial Conduct Authority (FCA) acknowledged progress in operational resilience mapping practices. It also identified opportunities to improve how information, processes, people and third-party dependencies are incorporated.
For Cognodata, this perspective has a direct application in data management projects.
We can follow the journey of an operation and examine the conditions that allow information to retain its usefulness at every exchange.
In the order example, the analysis would include the unique identification of the operation, the authorised version of the address, the time of the update, confirmation of receipt by the logistics provider and consistency of the final status across the systems involved.
Each condition can be linked to a specific check.
In this way, the map describes both the resources and the relationships that support the result.
The introduction of AI adds new information flows and dependencies that must be incorporated into this analysis. The Generative AI Risk Profile published by the US National Institute of Standards and Technology (NIST) in 2024 specifically addresses the documentation and oversight of dependencies between data, components and third parties.
Data quality must also be verified across systems
Data quality is often assessed through properties such as accuracy, completeness, consistency and timeliness.
When information moves between applications, these properties must be interpreted in relation to the process using the data.
An address may be correctly formatted. An insurance policy may exist. A permission may be valid. An operation may be marked as completed.
What matters is verifying that each element corresponds to the appropriate customer, time, status and action.
The validity of the result also depends on those relationships being maintained.
Understanding them allows organisations to assess more precisely what is happening between systems and where information needs to be verified.
Because an organisation may know every component involved in an operation and still need an answer to one essential question:
Are we using the right data for the right operation at the right time?
Frequently Asked Questions About Data Integration and Data Quality
Why can correct data produce incorrect results?
Because data accuracy does not guarantee that the information corresponds to the operation using it. Data may be valid but belong to another customer, an earlier version of a process or a different context.
How does data integration affect the quality of business results?
Data integration enables information to move between systems. For the result to be valid, the data must also retain its correspondence, timeliness and context throughout that exchange.
What does it mean for data to be current?
It means the information is valid for the time and operation in which it is used. Data can be correct as a historical record but unsuitable for executing a current operation.
How can organisations verify data consistency across systems?
By tracing the information flow and checking conditions such as the identity of the operation, the authorised data version, the time of the update and confirmation that the next system has received the appropriate information.
