Non-Aggregatable Measures

KPIs have to make business sense!

Published in

Crux Intelligence

3 min readDec 3, 2019

Context

Data is used day in and day out by business users to make important decisions. But for a business user to make decisions based on data, the data has to make business sense.

In some cases, business insights might not make sense because of certain aggregations associated with them. In this article, I’ll explain when these cases occur and how it can become problematic to consume data without having proper aggregation restrictions.

The Need

Certain KPIs do not make business sense when they are aggregated across certain dimensions or periods.
This happens because data providers might be providing information at a certain time granularity or the data might be aggregated at a specific level of any other dimension.

Let me use some examples to explain this concept.

Example 1

Headcount — Aggregation Across Time Granularity

Situation: Headcount is calculated on a monthly level by the data provider.

Concern: There is no way that the system can give headcount at a daily, weekly, quarterly or yearly level. Taking the sum of Headcount for all months of a year to present the yearly Headcount will be merely incorrect.
Headcount will only make sense at a Monthly Granularity.

Headcount for October 2019 is 300
Headcount for November 2019 is 340
Headcount for December 2019 is 440
Headcount for Quarter 4, 2019 CAN NOT be determined by adding up the above-listed headcount.

Example 2

Probability — Aggregation Across Dimension (City/Place)

Probability of WTC in New York, getting an earthquake in 2020 is 0.5%.
Probability of WTC in New York, getting a hurricane in 2020 is 0.6%.
The probability of WTC in New York getting an earthquake or a hurricane in 2020 CAN NOT be determined by adding up both the probabilities.

The Solution

As an Admin, I can define aggregation restrictions for a measure at a time period level and an attribute level.

Below is a screenshot showing what happens when users explicitly ask questions that do not make business sense.

Most BI platforms fail to capture this level of semantic knowledge, but, Cuddle can understand these details about KPIs and only presents answers or insights which make actual sense to a business user!