Online analytical processing, or OLAP, is a software capability used to create actionable business intelligence from a company’s available data by empowering analysts to navigate hierarchical relationships between categories and levels of detail in the data (known as dimensions). The power of OLAP is its ability to identify and anticipate trends—goals which are central to most business intelligence initiatives. It’s important that companies shopping for business intelligence (BI) tools be familiar with OLAP. Both end-to-end BI platforms and modern, self-service BI tools offer traditional OLAP or equivalent capabilities for multidimensional analysis.
In this guide we discuss:
As we’ve already suggested, the primary characteristic of OLAP is that it’s multidimensional. Dimensions of data include geographic categories (country, city, state), temporal categories (year, month, day), levels of aggregation (total sales, sales by store, sales by dept), etc.
You may have noticed that these dimensions can generally be grouped into conceptual hierarchies, and OLAP allows analysts to easily navigate between levels in these hierarchies to understand business problems. For instance, to understand why total sales plummeted in a given quarter, it may be necessary to drill down to a more detailed level: sales by store, and then compare that category with data on the types of products sold.
Generally, OLAP tools are used for historical analysis aimed at deriving insights about trends affecting the business, problems, opportunities for growth etc. In most contexts, a human user guides the analysis. This is opposed to operational analytics, or analytics aimed at processing data used in the business’s operations in real time or near real time.
Let’s take a closer look now at what distinguishes OLAP from operational analytics.
OLAP is frequently compared to OLTP, or online transactional processing. While OLTP handles the processing of data created in a business’s typical day-to-day operations, OLAP seeks to identify trends and help companies better prepare for the future.
Other differences between the two are highlighted in the chart below:
|Function||Monitors and records ongoing business transactions, such as purchases and sales.||Finds patterns that can help explain issues. Used to guide future plans and strategies. OLAP servers are commonly used in data mining and data warehousing operations.|
|Query types||Simple, standardized queries, such as: “How many units did Store 26 order last month?”||Complex, multidimensional queries, such as: “When Store 26 places a larger than average order, which other stores place larger than average orders the following month?”|
|Data source||Core business processes||Data gathered through OLTP|
|Database format||Relational; often presented in tabular form||Non-relational; comprised of data “cubes”|
|Read/write||Dataset is read/write and updated frequently.||Dataset is typically read-only to ensure it doesn’t get changed while analysis is underway.|
An OLTP database can be represented as a simple table or spreadsheet. This is easy to do because OLTP databases have a limited number of variables and the variables are directly related to one another.
This simple table shows average sales per day for each of a department store’s nine salespeople (A-I):
In the table below, we’ve added information from a third variable. It includes the same information as above: which employee sold how many products—but it now also shows which brands’ products were sold:
Both tables above represent the straightforward and limited nature of the types of datasets used in OLTP. They provide simple, clear transactional information—and little else. They may not be fancy, but few businesses operate without them.
Now, imagine the department store from the example has six branches. Each branch maintains its own sales records in a separate OLTP database, similar to those shown above.
The department store’s head office then “stacks” the individual transactional databases from the six branches into one single dataset. This creates a data “cube”—the format used in OLAP systems—also called multidimensional cubes or hypercubes.
We’ll continue with this example as we discuss the common functionality of OLAP.
The data cube above looks impressive, but astute readers may have noticed: despite the added dimension, it can still be read like an ordinary table or spreadsheet. Arranging datasets into cubes only makes it possible to begin the analysis.
The actual processing is where the value of OLAP lies. It relies on three common functionalities, all made possible by the flexibility of the non-relational database.
Due to the unique way OLAP arranges data, it offers benefits that other methods—those using relational databases—simply cannot. The wide variety of variables allowed and the ability to slice and dice them any which way gives companies new opportunities to find value in their existing company data.
Many use OLAP systems for predictive analytics, often for the purposes of forecasting and problem solving:
Predictive analytics tool from Halo showing forecast
The department store from the example might use predictive analytics to determine the ideal number of salespeople per branch for each day of the week. The OLAP tool tries several options and then “predicts” how the number of salespeople might impact sales. The store can then weigh this information against the cost of staffing and determine the ideal number of employees for per day.
Another benefit of OLAP is that it can uncover patterns and relationships that have not been previously considered. This is useful for problem solving. Let’s say the department store above is having a problem with sales of a particular brand’s products, but only in two of its six stores. OLAP analysis could reveal the root cause as an inexperienced manager who works in both of the problem stores.
We're able to offer this service to buyers for free, because software vendors pay us on a "pay-per-lead" basis. Buyers get great advice. Sellers get great referrals.