For business analysts, it is crucial to underline that analysis is done to support decision-making. Which dataset to use and what visuals to develop should all be determined by business needs,not the other way around. Therefore, the origin of business analysis is business ( or operational) issues.
Operational issues, such as ” why sales fell over the previous six months? “, are typically broad and vague, with few answers readily available. However, they, like lighthouses, steer our tour of analysis.The initial issue sets up the analysis goal, and the key to the solution is deconstruction.
Experienced analysts are good at decomposing abstract operational issues into specific problems from multiple perspectives to constantly verify and find answers. For example,the above sales issues can be decompsed into following problems :
- During the past six months, sales performance in different markets
- During the past six months, the sales of self-developed products compared with that of competitive brands
- During the past six months, the change in the number of customers
For senior business analysts, these problems could be expanded as follows by adding new angles of view.
- During the past six months, the composition of sales of new and old customers in different markets
- Since the launch of products, the sales of self-developed products compared with that of competitive brands
- During the past six months, the change in the number of customers with different order frequencies
Regardless of the complexity of the problem, we recommend analysts to graphically record the process of issue decomposition and problem verification. Every single business problem can be expressed by a view, and the dashboard is an interactive collection of these views.
1. Interpret the structure of problems
Business problems have a common feature. They are composed of three components: sample range, unit of analysis, and result of analysis. In terms of Tableau, these three components are filter, dimension, and measure. Each component is derived from a field or combination of fields in the data source. This similarity enables us to interpret the usual form of business problems and deal with various difficulties.
Let’s have a look at the following problems：
- In 2021, the average profit of each category and brand
- Top 50 customers(with sales), the last order date of each customer
- For orders containing furniture, the sales quantity and the joint sales ratio of each subcategory
These problems described business needs with standard structure. With the ‘comma’ and ‘preposition’ as the boundary, they are separated into three parts. The underline marked the sample range of the problem.Texts in gray ground are dimensions, which can be seen as the unit of analysis and determine the degree of aggegation. The result of the analysis, marked in bold, are the aggregated measures, which will be aggregated through specific methods such as average,maximum,count and summation.
*Please keep in mind:
- A problem must contain at least one measure,because measures are the results of analysis. By contrast, sample range and dimension can be omitted when representing “all” . For instance:
- (In company), the number of employees–– the sample range is “the whole company” by default, there is no dimension in this problem
- (In company), the sales by region–the sample range is “the whole company” by default
- Yesterday, the sales (of the company)— the sample range is “yesterday”, and the dimension is ” the whole company” by default
- Measures must be aggregated.Problems not involving aggragation, like “employee ID of each employee”, do not require analysis and can be answered by immediately “searching” the pertinent data list. In SQL, “search” and “analysis” form a broader concept – “query”; It refers to either looking up row level data (eg. SLECT * FROM TABLE) or aggregating fields by groups (eg. SELECT [region], SUM ([sale]) FROM table GROUP BY [region]).
2. Dimension, measure , filter and their relationships
The three components of the problem are closely related, as shown in Figure 3-1-1.
Figure 3-1-1 problem structure and relationships among components
- Dimension determines the number of analysis results (or the degree of aggregation),and measure is the value of analysis results. Briefly speaking, dimension is the (grouping) basis of measure. Similar relationships can also be found in Excel’s Pivot Table or SQL group by syntax.
- Filter limits the value of aggregated measure and may affect the number of values in dimension. Here are the examples:
- In the East region, the sum of sales of each province
- In the Northeast region, the sum of sales of each province “
For these two problems, the number of aggregated results are fully determined by the number of provinces in filtered region. Therefore, although they have the same structure, dimension and measure, the number and the value of results are different.
- In 2020, the sales of each month
- In 2021, the sales of each month
For the above two problems, the number of results will be the same because there are 12 months in each year, but the value of results will be different.
3. Technical concepts relate to each component
Each component of the problem structure correlates to a specific technical concept. As previously stated, the sample range corresponds to “filter”, and the result of analysis(measure) corresponds to “aggregation (AGG for short)”. For the unit of analysis (dimension), there is a new concept: “Level of Detail (LOD)”. This concept is a synonym for degree of aggregation, but is more commonly used in Tableau. As shown in Figure 3-1-2, each notion has its own set of knowledge, and their combination forms the foundation of advanced analysis.
Figure 3-1-2 problem components & related technical concepts
Now, we have a general understanding of business problem and related concepts. We advise newcomers to jot down business problems using the standard structure before analysis. In the following articles, we’ll go into great detail on how to transform the problem into appropriate visuals and how to carry out advanced analysis based on problems.