This lesson is all about Data Quality. You need a solid understanding of the dimensions of data feeding AI systems because the data has a direct impact on the performance and reliability of AI models and any prediction models you create in Salesforce. Why are we spending so much time on data? That’s because it’s one of the core elements to a successful AI implementation. Remember: AI + Data + CRM + Trust
Episode 16: Data Quality Dimensions
Definition: The currency or recency of data, which refers to how old the data is.
Importance: AI models perform better when they are trained on up-to-date information. Outdated data can lead to poor predictions and decisions.
Example: A sales forecast based on customer purchase data from five years ago might not be as relevant as data from the past quarter.
Definition: The degree to which all necessary data points are present.
Importance: Missing data can skew AI models, making them less reliable. Completeness ensures that the dataset has all the required fields for effective analysis.
Example: If customer records are missing key fields like email addresses, Einstein’s email marketing predictions will be less accurate.
Definition: The correctness or precision of the data.
Importance: Inaccurate data can lead to wrong conclusions, making AI recommendations faulty. Ensuring accuracy means the data reflects real-world scenarios.
Example: A contact’s address being listed incorrectly could affect the accuracy of Einstein’s lead scoring model.
Definition: The uniformity of data across different systems or datasets.
Importance: Inconsistent data can cause errors in AI models. Consistency ensures that the same information appears the same way wherever it is stored or accessed.
Example: A customer’s name being stored as “John Smith” in one system and “J. Smith” in another can lead to confusion and duplicated predictions.
Definition: The presence of repeated or redundant data entries.
Importance: Duplicates can skew AI models and inflate metrics, leading to inaccurate predictions. Data deduplication helps in maintaining clean, usable datasets.
Example: Two records for the same account could result in Einstein forecasting more revenue than is actually likely.
Definition: The frequency and context in which the data is utilized.
Importance: Understanding how often and in what scenarios data is used helps ensure that only relevant data is considered for AI models, enhancing model performance.
Example: Regularly updated data from a CRM is more useful for predictive modeling than rarely accessed archival data.
Definition: The conformity of data to predefined formats and rules.
Importance: Invalid data can cause errors in model training and prediction. Validity ensures that the data is correct in structure and meaning.
Example: Ensuring that phone numbers are stored in the correct format (+1 XXX-XXX-XXXX) ensures accurate communication with clients via automated tools like Einstein Bots.
Definition: The singularity of data points in the dataset, ensuring no duplication of records.
Importance: Uniqueness is vital to avoid redundancy, ensuring each data point contributes distinctly to AI models without skewing results.
Example: Having two identical customer profiles could lead to double counting in sales projections.
Definition: The relevance of data at the time it is accessed or used.
Importance: AI models need timely data to generate relevant predictions. If the data is outdated, the models may produce irrelevant insights.
Example: Real-time data from active sales campaigns can help Einstein Analytics provide more relevant sales trend insights than data from past campaigns.
Each of the dimensions discussed – Age, Completeness, Accuracy, Consistency, Duplication, Usage, Validity, Uniqueness, and Timeliness – plays an important role in ensuring that Einstein and other AI tools provide accurate and actionable insights. It will also drive the responses from Agentforce whenever an agent action is using data for making decisions or providing responses.
Want to see how your data quality stacks up? Here is an app resources from Salesforce that does just that! Data Quality Analysis Dashboards App
Take this quiz to test your knowledge!
