A performance score indicates how an individual employee contributes to your customer satisfaction level and revenue. At ‘manage staff’, performance scores are revealed in graphs which show how an employee’s performance changes over months.
Performance scores are calculated by an AI model. In essence, the model considers an entire team’s performance in a day and extracts the individual contribution of each employee. A feature engineering process ensures the model works on only relevant and significant data. External factors impacting the revenue in a day, such as weather or demand change on workdays/holidays/seasons, are carefully excluded. Also, the model considers the balance between the available workforce and the number of guests, for the purpose of not blaming employees if a team is understaffed or overstaffed.
Compared to our old model, the new model implements a significantly improved feature engineering process and a completely new scoring function. As a result, the new model generates more accurate, realistic, and consistent scores. Also, the new model does not present scores for managers (although they are considered in the model) since their impact on customer satisfaction and revenue is different (and more complex) than other employees and unsuitable for the current model.
What is important for more accurate performance scores
The data is the heart of the performance score model. Every information we can collect helps us engineer more relevant features. For instance, registration of the number of guests each day is essential to eliminate external factors such as weather or demand change on workdays/seasons. It is also critical to obtain correct data (teams, shifts, guests, sales) to generate accurate scores.