Gathering initial requirements to kickstart analytical data modeling journey

5:30 min

Medium

In order to kick off an analytical data modeling project, it's of the utmost importance that you've investigated and documented all the details that need to be considered in the required analytical data model(s). You'll want to make sure you understand all relevant business processes and the kind of insights that will need to be supported by your analytical data model.

It it strongly recommended to document all this information in detail, that will be beneficial when designing the analytical data model as well as for future references, such as updates to the data model, onboarding new dashboard designers, and many more.

We recommend tackling this investigation in the following steps:

Identifying key business processes

Defining analytical requirements

Identifying key business processes

1. Engage with stakeholders

The first step in identifying key business processes is to engage with stakeholders across various departments. These stakeholders could include managers, data analysts, IT personnel, and end-users who will rely on the insights generated from the data model. Conduct interviews, workshops, or surveys to gather information about their needs and expectations.

2. Understand the business context

To create a data model that is truly aligned with business needs, it's critical to understand the overall business context. This includes knowing the company's strategic goals, key performance indicators (KPIs), and any current challenges or pain points. Ask questions like:

What are the core business functions?
Which processes are critical to achieving business goals?
Are there any regulatory or compliance requirements?

3. Map out core processes

Once you have gathered information from stakeholders, the next step is to map out the core business processes. Create detailed process flows that highlight the inputs, outputs, and dependencies of each process. This will help in understanding how data flows through the organization and where key data points are generated.

To identify key business processes, our example company "Stellar Sales" first engaged with stakeholders, including customer success managers, product managers, and the store owners themselves, to understand the strategic objectives and challenges faced by their clients. This engagement highlighted several frequently requested insights into the performance of different product categories, regional sales trends across planets, and average order sizes. The goal is to empower customers to optimize their sales strategies and improve their offerings.

After several engagements, Steller Sales mapped out the following core processes:

Product Listing Management: How products are listed and categorized across different planetary markets.
Order Processing: The workflow from order placement to fulfillment, including cross-planet logistics.
Customer Interaction: Tracking customer behavior, including browsing patterns, product preferences, and purchasing habits.

Defining analytical requirements

1. Determine the scope of analysis

Before diving into the technical aspects of data modeling, it’s essential to define the scope of analysis. What questions does the business need to answer? What level of detail is required? For example:

Is the focus on historical analysis, real-time insights, or predictive modeling?
What time frames of data are required: all data, limited to recent data only, etc.?
Are there specific dimensions (e.g. geography, holidays, etc.) that need to be included?
Which filtering capabilties are required? Which actionable insights could be beneficial to a user's workflow?

2. Identify data sources

Next, identify all potential data sources that could feed into your analytical model. These could include:

Data lakes or warehouses already in place within the organization.
Internal systems such as transactional database(s), ERP, CRM, etc.
External data sources like market trends, social media analytics, or third-party data providers.

3. Data quality assessment

Assess the quality of the data from these sources. High-quality data is critical for accurate insights. Consider factors such as:

Completeness: Is all necessary data available?
Accuracy: Is the data free from errors?
Consistency: Is the data standardized across sources?
Timeliness: Is the data up-to-date (i.e. available when it needs to be visualized)?

4. Define key metrics and dimensions

Work with stakeholders to define the key metrics and dimensions that will be used in the data model. This includes:

Metrics: Quantitative measurements that typically originate directly from business events. These are commonly used in data visualizations to aggregate upon (e.g. average the values, sum them together, retrieve the min/max value, etc.).
Dimensions: Categorical data used to slice metrics, often associated to business events but could also be part of generated business event (e.g. event datetime). These are commonly used to group or filter on in data visualizations.

Our example company "Stellar Sales" needs to define analytical requirements to provide valuable insights to its customers. These insights should help customers identify top-performing product categories, understand purchasing behaviors across planets, and optimize their inventory. Their customers want to be able to answer key questions such as:

Which product categories are performing best across different planetary markets?
What are the sales trends in specific regions or planets in the past weeks/months/years?
What is the average order size and how does it vary by region or product category?

The data originates from different sources (e.g. webshop, logistics, manufacturers, etc.), but for simplicity we'll assume most data is already synchronized and natively available in a single data source (e.g. a data lake, data warehouse, etc.).

Key Metrics and Dimensions

Stellar Sales, together with its stakeholders, defined the following key metrics and dimensions for their sales data:

Metrics: cost, revenue, amount of products, and order duration.
Dimensions: product info (type, category, etc.), customer demographics (age, gender, etc.), store (location, manager, etc.), and time (daily, weekly, monthly trends).

The logistics data will benefit from some of the same dimensions (often called a "conformed dimension"), but the metrics might include duration of shipping, time to prepare order for shipping, shipment size, etc. Similarly, customer interaction data would require the customer demographics dimension, while the metrics can be e.g. visit duration, number of events before ordering, number of searches, etc.

Gathering initial requirements is a critical step in the analytical data modeling journey. By thoroughly understanding business processes and defining analytical needs, you can lay a solid foundation for a data model that drives meaningful insights and supports strategic decision-making. Now that we know which requirements we need to keep into account, we'll want to make sure we evaluate and prioritize the business processes according to their impact and risk. More about this in our next article!