Metrics

The analysis of a model generates several metrics used to prioritize the issues reported by DAX Optimizer.

This section describes each metric, but consider that DAX Optimizer performs a static analysis of the DAX expressions, generating an estimation of the cost that cannot match the actual cost of each element.

Therefore, the comparison between metrics makes sense only in relative terms. So, do not make any assumptions about the absolute values reported in each metric.

Furthermore, to display the metrics more intuitively, DAX Optimizer automatically rescales the numbers with a meaningful number of digits.

Overview metrics

After the analysis of a VPAX file, DAX Optimizer displays an Overview page reporting four metrics.

The open issues found by DAX Optimizer affect the VPAX models as follows:

• Overall. It’s the mean value of the other three metrics and represents a global recap of how the model is performing. If the model is performing well globally, this metric will be near 100.
• CPU. It indicates how hard are the open issues stressing the CPU. Efficient models have a value near 100.
• RAM. It indicates how hard are the open issues stressing the RAM. Efficient models have a value near 100.
  

  NOTE: This metric has even an indirect importance, as a high RAM consumption of your models determines the execution capacity of your Power BI subscription: the more RAM you consume, the more you need a greater capacity of the underlying system and, therefore, spend more. Therefore, DAX Optimizer helps you spend less through this metric.

• Impact. This metric indicates that the open issues found by DAX Optimizer are related to measures that depend on other measures. The more branched this phenomenon is, the worse your model is in terms of performance. Efficient models have a value near 100.

Measures metrics

The Measures page regroups the measures of a VPAX model referring to the following metrics:

• Relevance. It indicates how relevant is a metric for the model. In other words, it represents the estimated impact of the measure on the overall performance of the model. The relevance value is normalized using a normalization factor and the tooltip on the number shows the actual value computed in the analysis.
• Exec. It shows how many times the measure is executed in the model. This number is obtained by estimating multiple executions of the measure when it is evaluated in an iterator. The estimation is not necessarily accurate, but it provides a good relative estimate to compare a measure with other measures.
• Dir Ref. The direct references metric reports the number of measures that have a direct reference to the measure.
• Ind Ref. The indirect references metric reports the number of measures that depend on this measure but do not have a direct reference to it.

Expert metrics

By clicking on View > Expert View in the pages Issues or Measures you can visualize the following metrics:

• CPU Opt. The CPU optimizable metric represents the possible optimization (%) of the CPU cost of the measure. The estimate is based on the relevance of the issues detected compared to the overall CPU cost of the measure.
• CPU cost. It represents the estimated CPU cost of the measure inclusive of referenced measures. The value is normalized using a normalization factor reported in the tooltip on the column header. However, the absolute value is not important; use CPU Cost only for relative comparison with other measures.
• RAM Opt. The RAM optimizable metric is the possible optimization (%) of the RAM cost of the measure. The estimate is based on the estimated materialization of the issues detected compared to the overall materialization required by the measure.
• RAM cost. It represents the estimated RAM cost of the measure based on the maximum materialization expected. The value is normalized using a normalization factor reported in the tooltip on the column header. However, the absolute value is not important; use RAM Cost only for relative comparison with other measures.