Certification Practice: Data, Quality & Privacy

Question 1

Why is 'Data Quality' crucial for the success of a Generative AI implementation in Salesforce?

Accepted Answer

Because AI models trained or grounded on poor quality data will produce inaccurate or biased results ('Garbage In, Garbage Out')

Answer

Because Salesforce prevents enabling Einstein if it detects a single duplicate record

Answer

Because poor quality data increases network latency

Answer

Because AI consumes more license credits if data is dirty

Question 2

Which of the following are standard dimensions for measuring Data Quality? (Select 3)

Accepted Answer

Answer

Complexity

Answer

Completeness

Answer

Elasticity

Answer

Accuracy

Answer

Consistency

Question 3

An administrator observes that the 'Country' field in customer records has inconsistent values like 'EE.UU.', 'USA', 'United States', and 'U.S.'. Which data quality dimension is failing?

Accepted Answer

Consistency

Answer

Timeliness

Answer

Completeness

Answer

Security

Question 4

What is a 'Duplicate' in the context of CRM data and how does it affect AI? (Select 2)

Accepted Answer

Answer

It is beneficial because it gives AI more data to train with

Answer

It occurs when multiple records exist representing the same real entity (person or company)

Answer

It only affects marketing reports, not AI

Answer

It can cause AI to provide fragmented or contradictory information about a customer

Question 5

What Salesforce Data Cloud feature helps solve the problem of having customer data scattered across multiple systems with different IDs?

Accepted Answer

Identity Resolution

Answer

Data Deletion

Answer

Apex Triggers

Answer

Flow Orchestration

Question 6

What does 'Data Harmonization' mean in the context of preparing data for AI?

Accepted Answer

Mapping data from different sources to a common standard data model (like CIM/Data Cloud Model)

Answer

Deleting all old data older than 5 years

Answer

Encrypting all data so no one can read it

Answer

Translating all text fields to English

Question 7

What is the consequence of the 'Garbage In, Garbage Out' principle in AI model training?

Accepted Answer

If input data is inaccurate or biased, AI predictions or generations will be equally flawed

Answer

The model will take longer to train, but the final result will be perfect

Answer

The model will refuse to train and give an error message

Answer

The model will automatically clean bad data during the training process

Question 8

What should be done with fields containing 'Missing Values' before using them for AI? (Select 2)

Accepted Answer

Answer

Leave them empty, AI will guess the correct value magically

Answer

Impute the values (fill them with mean, median, or a logical default)

Answer

Duplicate the record to compensate for the missing value

Answer

Delete the records if the missing data is critical for the model

Question 9

What is the difference between 'Training Data' and 'Grounding Data' in the Salesforce context?

Accepted Answer

Training data teaches the model general patterns once; grounding data gives specific real-time context for each query

Answer

There is no difference, they are synonyms

Answer

Training data is owned by Salesforce; grounding data is public

Answer

Grounding data is used to create the model from scratch

Question 10

What is 'Data Leakage' in the context of predictive model training?

Accepted Answer

When the model accidentally has access to data during training that reveals the outcome it is supposed to predict (cheating)

Answer

When the model forgets what it learned

Answer

When data is stolen by hackers

Answer

When data is lost during transfer