It’s no secret that the global ecommerce ecosystem has benefitted from consumers heavily relying on online shopping during the pandemic. Even as COVID-19 restrictions have loosened here in the U.S., we know that online shopping is not going away. In fact, companies are looking to capitalize on this change by carefully examining user data in new ways.
On the business side of things, more users want access to their systems to conduct ad hoc reporting. How are you to address these needs while keeping good data access control, performant systems, and making sure that the company KPIs are being calculated properly?
The answer is simple: you need quality Analytics Models. Let us review these different use cases and see how an analytics model can help us solve them.
A Use Case for Analytics Models
Let’s start with a scenario: your company has full control of your database processes. The data is flowing into a central warehouse. Your data management and reporting team can get the information they need across different departments and do consolidated reporting, but there is still a problem. With the new central data location, more requests are coming in faster than before.
This scenario sets the scene for how an Analytics Model can support, process, and produce validated data in a timely manner, with multiple end-users, and much more.
First and foremost, let’s define what we mean by an Analytics Model. Analytics Models are generally “in memory” models that are optimized for high-speed query performance, predefined analytics calculations, ease of data discovery, and reducing the time needed for non-IT personnel to be able to fully utilize the data.
Microsoft SQL Server Analysis Services is one platform that supports these features and is nicely included in the SQL Server license that you most likely already own. The different pieces that make up the Analytics Model are the following:
- Cleaned and validated data to feed the system (the warehouse)
- A data layout that defines the relationships and connections between the fact and dimension tables (modification of the warehouse structure)
- User-understandable column names setup in the model
- Restructure of data where appropriate to support analytics processing needs
- KPI definitions that set the rules on how values are calculated
- Calendar setup to allow easy date manipulation by end users
- Creation of multiple models to fit specific department needs
- Segmenting\data based on security needs where appropriate and making the data right sized for end-user groups
Let’s review how these items function within an Analytics Model.
Increased Data/Report Requests
The model created in analytics engines removes the need for report writers to have build joints and connections between the different sets of data. The model pre-creates and validates the connections between tables.
Report writers no longer need to spend time connecting up the various tables and building out complex calculations; they are already built in the engine. With these prebuilt items, report writers will spend their time laying out reports and building out visuals and not spending large amounts of their time setting up the data for each report.
The result of this is a large decrease in the amount of time needed to produce reports. With reduced time to produce and validate new reports, the number of requests that can be handled can increase.
Increased Number of Users Requesting Access to Data
Analytics Models make data easier to understand and use for non-technical personnel. The primary method for access for these users is through new tools like Power BI, which gives direct access to the model or using well-known tools like Excel, which directly interface with the model.
Additionally, specific models can be set up for various departments, thereby allowing them to access data that is relevant to their teams. These users can then access the information in easy-to-use tools that they are familiar with.
Access can also be granted to large data sets as desired. This allows for more data to be accessed by members without setting up smaller analytics models.
System performance will always be a concern when allowing more users with uncontrolled access to the system. In SQL warehouse configurations, users can create queries that can block access to other users and cause issues with server load running large queries that the system isn’t indexed or optimized for.
With an Analytics Model, unstructured queries that access many different parts of the system are what the models excel at handling.
Additionally, querying the data does not block other users from accessing their needs. The result is eliminating a large issue of system contention that can happen in warehouse environments.
Data Integrity/Proper Calculations
The final area where Analytics Models help improve over a warehouse is with data integrity. The models are fed with data from the warehouse, which is then cleaned and validated.
What the warehouse doesn’t handle well are complex calculations that cover different data scenarios. The Analytics Model fills this role. The calculations are built into the model and users just use the calculation to build out reports.
Each report writer does not need to recreate the different measures that are used in the reports, as they use the values set up in the engine. The result of this is the reports all use the same calculations for the data resulting in more accurate results. Without a system like this in place, each report would have calculations with very slight variations, which would provide different results to every user.
Analytics Models Improve Your Data Warehouse System
In summary, Analytics Models will enhance data warehouses, allowing for increased data usage and availability without lowering data quality. They are a great asset in most environments and can greatly enhance the usage and benefit of the information that is being gathered in the company data warehouse.
Have questions on how implementing an Analytics Model can help improve your business? Contact us today and we’ll have our Business Intelligence experts assist you with all of your data-driven needs.
