QUALITY WITH A PURPOSE
Improve the data that can actually change the outcome.
Not every imperfect field deserves the same attention. Start with the data that can change a customer decision, an AI response, an audience, a service action or another business outcome.
That keeps data-quality work tied to value instead of turning it into an endless cleanup program.
MEANING BEFORE COMPLETENESS
A populated field is not necessarily reliable data.
A field can contain a value and still be unsafe to use. The value may be outdated, copied from another system, entered for a different purpose or based on a definition nobody shares anymore.
Reliability depends on context.
The same field may be reliable for one decision and too weak for another. Quality should be judged against the use case, not against an abstract idea of perfect data.
Ask: reliable enough for what?
EXAMPLE
Example: One customer status. Three different meanings.
A status field can look consistent across systems while representing completely different business facts.
The field name alone is not enough. Before using a status in Data 360 or AI, define which business fact it represents and which system is authoritative for that fact.
- CRM status
May describe the current relationship with the customer.
- Service status
May describe whether a support case, entitlement or service condition is active.
- Contract status
May describe the legal or commercial state of an agreement.
- Meaning
- Source & ownership
- Quality & access
- Use & feedback
The goal is not zero defects. It is known limitations, visible ownership and reliable handling of the data that matters most.
START WITH ONE PROCESS
Start with the data one process cannot afford to misunderstand.
Pick one process, audience or AI use case and identify the fields that materially affect its outcome.
Critical data is defined by consequence, not by how often the field appears in the model.
- Which fields can change the decision or action?
- Which source owns the business meaning?
- How fresh does the data need to be?
- Which missing or conflicting values create real risk?
Further steps (1)
- What evidence makes the value trustworthy enough to use?
FOCUS THE WORK
Do not fix every data issue at once.
Prioritize the issues that change business outcomes, create operational risk or make the use case impossible to trust.
Fix first: data that changes the decision
Incorrect values would send the customer, agent or automation down the wrong path.
Fix next: data that creates repeated operational friction
Teams constantly investigate, correct or work around the same quality issue.
Monitor: imperfections with limited impact
The data is imperfect, but the limitation is known and does not materially affect the current use case.
The right priority depends on the process, not on a universal data-quality score.
DATA FOUNDATION
Four layers of a usable data foundation
Useful data needs more than populated fields. Teams need to understand what the data means, who owns it, why it can be trusted and where it is allowed to be used.
01
Meaning
What business fact does this field represent, and how is that definition shared?
02
Ownership
Which system and team are responsible for the value and its business definition?
03
Trust
What evidence, freshness and quality checks make the value reliable enough for this use case?
04
Activation
Where may the data be used, under which permissions, consent or business rules?
If one layer is missing, the data may still exist — but teams will struggle to use it confidently.
To assess whether Data 360 fits the resulting activation path, read Salesforce Data 360: From Use Case to Activation.
AI cannot infer missing business meaning safely. It needs data whose definition, ownership and limitations are understood.
Frequently asked questions about data foundations
Does all data need to be perfect before we use Data 360 or Agentforce?
No. Focus on the data that materially affects the use case. Known limitations can be acceptable if they are understood, monitored and do not create unacceptable risk.
What matters more: completeness or consistent meaning?
For most important decisions, consistent meaning comes first. A populated field is not useful if different teams or systems interpret it differently.
Who should own data quality?
Ownership usually spans both business and technology. The business owns the meaning and acceptable use; technology teams help ensure the source, movement and controls support that definition.
Sources and further reading
Primary Salesforce documentation and reference material for Data 360, Agentforce and data-quality practices.
- Salesforce – What is Data 360?Checked on
Supports data, trust and governance fundamentals for Data 360 and Agentforce; recheck current product requirements before publication.
- Salesforce – How Agentforce worksChecked on
Supports data, trust and governance fundamentals for Data 360 and Agentforce; recheck current product requirements before publication.
- Salesforce Well-Architected – TrustedChecked on
Supports data, trust and governance fundamentals for Data 360 and Agentforce; recheck current product requirements before publication.
Topics in this article
THREE WORKING VIEWS
Three views that make data quality tangible in operations.
Illustrative configured workspaces make the decision, evidence and operating state visible. All records are fictional.
Check missing identity evidence
Illustrative review where a missing key prevents a match.
Swipe with your finger or drag with the mouse button held down. You can also select an image or use the arrow keys.
Image 1 of 3: Check missing identity evidence
Continue reading
AI in Salesforce: Which tasks are worth it?
Not every Salesforce task gets better with AI. The useful starting point is a specific piece of work where interpretation, summarization or drafting creates a real advantage over a simple workflow or automation. The best first use case is narrow enough to test, useful enough to matter and controlled enough to operate.
Read article
Salesforce Data 360: Start with the Use Case, Not the Data
Data 360 becomes useful when the business outcome comes first. Instead of asking which sources can be connected, start with the decision, audience or process you want to improve — then work backwards to the data actually required. That usually leads to a smaller first scope, clearer identity rules and an activation you can measure.
Read article
Escaping the Monolith: Designing Scalable Multi-Agent Ecosystems in Agentforce
A practical guide to replacing one overloaded Agentforce agent with a lean router, bounded specialists, explicit handoffs, and evidence that the architecture scales.
Read article