Improve Accuracy with Insights from the Model Scorecard
To recommend articles that can help agents close cases, Einstein
requires lots of rich case and knowledge data to learn from. To help you understand your data
and the effectiveness of your predictive model, Einstein produces a model scorecard. At a
glance, you can learn where there are opportunities to refine your data and improve
recommendations.
Where: This change applies to Lightning Experience in Enterprise, Performance, and Unlimited editions.
How: The scorecard updates when Einstein retrains the model or when you rebuild it.
From Setup, under Einstein Article Recommendations, open the Model
Scorecard tab. On the left, the scorecard shows model accuracy and quality (1).
On the right, you can learn more about your data: how much term overlap there is between cases
and articles (2), how well populated your selected case fields are (3), and how many
case-article attaches there are per language (4). The results help you understand where your
data contributes to accurate recommendations and where you can make improvements.