Answering business questions with data

One of the tasks of a data analyst is to help business stakeholders answer business questions using data.

Examples of business questions:

• Why is there a decrease in conversion rate in November? Is there a specific segment causing this decrease?
• In which city are there more delayed repayments of loans? Does it only happen in a specific month?
• Are there customers who prefer luxury products? What other products can be offered to these customers?

Answering these types of questions can help the business stakeholder decide on the actions to take in the future.

In the next article, I will describe the steps that an analyst needs to take in order to answer these questions.

Understanding the Business Question

In the first stage, before the analyst approaches the tables, he must understand what the business entity wants to understand from the data, what is the business question that is important to answer. An accurate definition of the problem is very important for the success of the analysis, and the analyst can help clarify the understanding of the problem so that he can answer it quantitatively in the following stages.

For example, when a business entity talks about the amount of details sold in the last quarter, the analyst needs to know whether he is referring to sales of all the company’s products or only products marketed by his department.

This process also allows for expectations coordination between the analyst and the business entity. Without this coordination, the business entity may have a certain idea in mind and in practice, he will receive a different result than what he expected to receive.

Locating data repositories

Visualization and Data Presentation

After constructing the appropriate model for the business question, the analyst will conduct investigations on the model to find answers to the questions. In order to present the answers later on, the analyst usually uses graphs. Analysts generally use graphs to display data because it is harder for the brain to compare quantitative data when working with tables.

In the market, there are many tools that can display data graphically, from Excel to advanced BI tools. It doesn’t matter which BI tool you choose, but the important thing is that the tool can produce a clear display and that the graphs are understandable at a glance with as few lines and symbols as possible.

The field of data presentation is a whole theory, but it is desirable that the analyst knows at least the basic principles before presenting the data.

You can read more about this field at Stephen Few blog – perceptual edge.

Sanity checks

The process of data analysis is a very complex process, and therefore it is reasonable to assume that errors will occur along the way in each of the stages. For example, an incorrect choice of fields in model construction will lead to illogical results, and an error in constructing the data model may distort the results. Therefore, in order to prevent these situations, the analyst will perform spellchecking at every stage.

In spellchecking, the analyst will return to the business entity and ask whether the data they received is logical in order to verify that there were no errors along the way. An example of such a question could be “Is it logical that the sales volume for the last month was 2 million dollars?” or “Is it logical that in March I see a sharp increase in the number of users in the application?”.

If the answers to the Sanity checks do not match the data that the analyst receives, he or she must go back to the previous stages and find where the error occurred. Sometimes in these checks, it is discovered that the data in the tables are incorrect and the DWH staff must arrange and fix them.

Data presentation to the business stakeholders – to tell the story of the data.

At the end of the analysis, the analyst will present the insights to the business factor. Usually, the results will be presented in a PowerPoint or a document that will arrange the graphs and data logically, showing the main insights and what caused them. Such a presentation can include business insights, actions to be taken, and assumptions for additional business questions.

When writing the insights, the analyst should pay attention not to be biased by the data analysis, and that the insights are not affected by intervening variables.

The way the graphs and data are presented should be in a straightforward and didactic way. The image of a teacher instructing an elementary school student is suitable here. The simpler the story is to understand and the more supported by findings, the easier it will be for the business factor to understand and use these insights.

This type of presentation is called “storytelling with data,” and the idea is to provide the business stakeholders with a pleasant and simple experience as possible.

I know the last lines may seem like they were taken from the world of marketing or psychology, but the reason for this is that there are business factors that are apprehensive about data analysis, and the easier the analyst makes it for them to present the findings, the less their anxiety will be.

To summarize

In the article, I described the steps in responding to a business question by an analyst, up to a full answer that tells the story of the data. The steps in the article are very detailed, and in practice, it is not always necessary to go through all the steps in order to answer every business question.

In some cases, it is possible to use an existing data model to shorten the process, and sometimes the answers are simple and already appear in the company’s KPI report. However, it is important to remember to always coordinate expectations with the business entity regarding the expected outcome and to always (always!) perform sanity checks during the data work.