Jeff Erickson | Tech Content Strategist | June 22, 2023
Most of us regularly interact with real-time analytics, even if we’re unaware of it. This type of analytics works in the background to help pick the route for a driver dropping off our package, scan for hints of fraud on credit card purchases, and help utilities keep the lights on with proactive maintenance on power-generating equipment.
And while not every business needs to act on data in milliseconds, real-time analytics continues to improve, making the technology—and the business advantages it brings—possible for many more organizations. This is shifting how we think about analytics; instead of just looking back at what happened and how to improve next time, real-time analytics is a moment-to-moment operational decision tool.
Real-time analytics takes data the moment it’s generated—whether by a website click, a social media comment, a transaction, or a sensor—and flows it into a system for analysis and immediate action. Real-time analytics in some business processes operates in milliseconds to pull data from multiple sources and feed it into a system where it’s organized and analyzed—and then either acted on by an automated system or communicated to people in graphs, text, or voice. It’s how ticket sellers adjust prices based on demand, how an airline updates a flight status, or how a bank pings you immediately when there’s a charge that its artificial intelligence algorithm doesn’t like.
Real-time data analytics systems may draw from structured and unstructured data sources. Structured data comes in predictable and consistent formats from sources such as business applications. Unstructured data, sometimes called big data, requires additional processing and comes from sources that include social media sites, text documents, and videos. Data analytics systems can combine these two data source types for a richer analysis and then present findings in ways that people can easily understand and act on.
The technologies that enable real-time analytics include databases and data lakes, machine learning (ML) algorithms, data integration tools, programming languages, data science notebooks, and various open source projects. Combined with machine learning, real-time analytics systems can do more than aid decision-making in the moment—they can also look for trends, bottlenecks, or business opportunities hidden in operational data.
Real-time analytics is one flavor of data analytics, and it’s gaining popularity with savvy digital businesses. It’s an extension of traditional data analytics and uses many of the same skill sets. Traditional analytics, often called batch analytics, is a slower process in which large quantities of stored data are prepared and then sent to an analytics platform to generate graphs or charts in a dashboard. The data could be hours, days, weeks, or even months old and is used to paint a picture of what happened in the past. This was, and remains, a key resource to help guide future decision-making.
In contrast to traditional data analytics, real-time analytics is about what’s happening now. Instead of storing data and then periodically moving it into an analytics system using a complex technical process called extract, transform, and load (ETL), real-time analytics immediately pushes data into the system for analysis and action—often just milliseconds after it’s created. It’s easy to understand why this can sometimes be called streaming analytics.
Many organizations are switching from batch processing to real-time processing and from request-driven architectures to event-driven architectures that enable more automation.
Many data management architectures can support real-time analytics, but one that’s gaining popularity due to its simplicity is called in-database analytics. This lets analysts execute analytics where the data is stored rather than taking the extra, time-consuming step to ETL large data sets to a separate analytics database. Analysts at Forrester have dubbed this in-database analytics model a “translytical” platform—combining transactional and analytical functions—and it can make it easier to maintain data integrity and do analytics at scale.
At retail stores predicting demand, marketing agencies accelerating targeting decisions in milliseconds, and many other organizations, people are finding the in-the-moment insights of real-time analytics a valuable tool for making decisions or automating actions.
Real-time analytics gives businesses the information they need to act in the moment, whether that’s to change driving routes, react to a manufacturing problem, change a marketing campaign, or update a supply chain partner.
Real-time insights about a customer’s order or service request provide a smoother, more personalized customer experience.
Businesses can adjust prices, change offerings, or update product availability in real time to improve efficiency and revenue in ways that a less digitally adept competitor can’t.
Real-time analytics can help marketers identify trends as they unfold. Using analytics that combine diverse factors such as sales and social media sentiment, the technology can adjust messages or even suggest product changes to capitalize on the trend before the competition.
Creating the integrated, scalable data infrastructure needed for real-time analytics has typically required planning, expertise, and funds. One key factor behind many of the challenges to real-time analytics is assembling an architecture that’s both powerful and efficient enough to let data collection, integration, and analysis happen in real time. However, complex architectures can lead to downtime and headaches for engineers and possibly lower adoption if the service is unreliable. Below are three steps to help overcome the challenges.
One of the first challenges to implementing real-time analytics is accounting for all the data sources involved. For example, a retail application draws data from product suppliers and feeds it to financial accounting software and customer service applications. The right sources for a real-time analytics initiative might be inside or outside the business and include structured or unstructured data. IT teams can use many tools to locate and catalog data sources.
Once a team has identified data sources, the data must be integrated into a stream of data that can be used by the analytics system. This step often requires an integration platform that provides the APIs and prebuilt connectors needed to ingest data from multiple sources.
Because real-time analytics draws from data sources that change based on business activity, data volumes can be unpredictable. The compute resources assigned to real-time analytics must be either provisioned for the highest possible use case or built on a cloud service that can scale up and down to meet changing needs.
Both structured data and unstructured data can be used in a real-time analytics system. In fact, combining the two for analysis, to quickly paint a clearer picture for the business, is what makes many real-time systems so valuable. These two types of data are different in just the way their names imply: Structured data comes in consistent, predictable formats from sources such as business applications, which makes it easier to put in a relational database. Unstructured data lacks predictable formatting; it’s pulled from sources such as social media feeds, customer comment forms, text documents, or videos and then formatted for use in the real-time analytics system.
|Data Type||Definition||Key Differentiator||Example|
|Structured Data||Data that’s organized into a clearly defined format||Easy to sort, track, classify, and put in a relational database||Sales results, survey responses, customers’ addresses or purchase history|
|Unstructured Data||Data that doesn’t follow a predetermined format||Difficult to fit into a relational database||Email text, social media posts, audio, videos|
A real-time data analytics process will lean on the quality of an organization’s overall data management practices. Enterprise data management software should include the ability to scale quickly, integrate data from many sources, ensure data quality and strong governance, and, of course, prioritize data security. Below are best practices to consider.
First ask the question: Who is this real-time analytics engine for? It’s unlikely to apply companywide, so you need to assess if it will be used by an entire department or just select users within it. Having a crisp, focused set of goals will help with this assessment. Sorting this out will lead to what data sources inside and outside the company you will need to access. Another question to ask in this process: Would you be more ambitious with those goals if you had more or better data?
Keep to a minimum the number of times data must move or go through an ETL process. ETL processes can create latency and can increase data security and compliance risks as data moves between data stores. A current trend is to use in-database analytics, where data processing is performed within a transactional database to avoid moving large data sets to a separate analytics database.
Even a medium-sized company averages 20 paid SaaS products in use, according to a recent survey. Add that to on-premises software and other third-party or unstructured data sources, and you have a lot of choices. Identify those that your real-time analytics initiative will require.
Different machine learning models reveal different kinds of insights based on how they look at data. ML models can be trained for regression or classification tasks, anomaly detection, or other purposes. Beyond getting real-time insights, machine learning can help detect trends, make faster decisions, and automate actions or recommendations.
The right data tools can help you compose a real-time analytics system. If you use ETL processes, you will need tools to extract data, clean and transform the data sets, and flow them into the appropriate systems.
There are two ways to think about monitoring the performance of your real-time analytics. One is purely human—to establish relationships with people in the business who can report on how it’s working on the ground. Is the factory floor running smoother, or are customers getting the automated information they need? The second way is to monitor your data processes to identify negative trends and bottlenecks and be able to react.
A real-time analytics system can have many data sources and dependencies. When a change in the business environment brings a change to one of those inputs, make sure your real-time analytics system and the employees who use it have a way to note the issue and a process to fix it.
Brazil-based Tetris.co shows how a business can benefit from giving decision-makers direct access to real-time analytics. The company brings together data from several media sources into a MySQL database and uses real-time analytics to understand how advertising investments perform. The company achieved the speed their software requires by moving to MySQL HeatWave, where they could run transactions and real-time analytics workloads directly from a MySQL database, eliminating the need for data movement and integration with a separate analytics database. The high-performance system helped front-line analysts understand trends faster and improve marketing results by shifting investments out of underperforming advertising platforms and into higher-performing channels.
Many skills and tools can help build a real-time analytics system that produces results for your organization. They include tools for data modeling, data quality, and data visualization. A good place to start is to consider your current software and skills. For example, an organization that uses MySQL Database for transactions might simply opt for a cloud-based version that offers in-database analytics plus in-database machine learning, eliminating the need to ETL data to separate analytics and ML systems.
If your organization needs the advantages of real-time analytics, MySQL HeatWave offers a powerful solution that lets you simplify your architecture. MySQL HeatWave is a fully managed database service, powered by the integrated HeatWave in-memory query accelerator. It’s the only cloud database service that combines transactions, analytics, and machine learning services into one MySQL Database, delivering real-time analytics without the complexity, latency, risks, and cost of ETL duplication.
MySQL HeatWave Lakehouse lets users natively process and query hundreds of terabytes of data in the object store—in a variety of file formats, such as CSV, Parquet, and Aurora/Redshift backups. With MySQL HeatWave Lakehouse, MySQL HeatWave provides one service for transaction processing, real-time analytics across data warehouses and data lakes, and machine learning—without ETL across cloud services.
What is an example of real-time analytics?
There are many examples of real-time analytics in business. One company, FANCOMI, aims to become the world’s largest performance marketing advertising network that lets advertisers pay when their desired marketing outcome is achieved, instead of the traditional way, when the ads are placed. It’s using real-time analytics to monitor and measure the impact of 20,000 advertisements to 2.6 million agencies and media websites 24 hours a day.
Why do businesses need real-time analytics?
Digital systems, including Internet of Things sensors, social media sites and apps, and online retail, combined with behind-the-scenes systems such as CRM, ERP, and human capital management (HCM), are generating data in unprecedented amounts. Businesses that can rapidly make sense of that operational data deluge to see changes in their businesses and respond with the right decisions will beat the competition.
How does real-time analytics improve decision-making?
Real-time analytics uses data the moment it’s created, when it’s most relevant. Organizations that don’t use real-time analytics may make important decisions based on data that’s already stale by the time it’s available for analysis.
Learn how to build machine learning models, easily migrate to MySQL HeatWave, or explore other MySQL HeatWave topics of interest.