NOTE: "Variance" discussed in this blog is not related to statistical variance.
Any BI presentation layer that contain reports, dashboards and applications have one fundamental and widely reported measure, which is Variance. It is addressed with different names such as “Change“ or “Difference“, but the essence of showing a degree of deviation of a metric between two values remains the same. Variance is a clear indicator of performance and are extensively used in conjunction with benchmarks. (Previous posts about Different Types, Guided and External benchmarks)
Variance can be calculated between
- Two logically relevant metrics (e.g. Actual vs Budget)
- Two semantically related dimensional elements (e.g. Current Quarter vs Last Quarter)
In both contexts listed above, scope of analysis can be further enhanced by adding more dimensionality.
Two popular ways of showing variance are as absolute numbers or percentage values. There is no hard and fast rule on which format is better and in majority of cases is boils down to user preferences.
In the presentation layer that incorporates “Variance”, user attention gives precedence to this measure. It is an automatic cognitive switch because Variance calls out the attention of users. Due to this ability, variances have additional visual cues built into the reports such as colors or icons, which is widely adopted and more of an integral element in today’s context.
Variance can be looked at as a “derived” measure, where the calculation happens only in presentation layer. Facts in a data warehouse keeps changing on daily basis, which makes it hard to store them.
In conclusion, Variance is a great of way of showcasing the performance of several KPI’s in small real-estate. Decision making efforts can be more focused in addressing problematic areas rather than spending time on digging through multiple data points.