We live in a real-time society where people want things immediately. The impulse for instant gratification is powerful, and the current on-demand economy reflects this. To keep pace with consumer expectations, businesses are increasingly automating processes with the help of machine-to-machine communication, the internet of ihings (IoT), artificial intelligence, and machine learning—resulting in more real-time transactions.
As the economy continues to be fueled by the need for immediacy, there’s been a surge in organizations across industries working to develop a new generation of applications that leverage the power of real-time decisioning. In this two-part series, I’ll discuss use cases for real-time decisioning that showcase real-world examples of businesses making the most of their high-velocity data.
Enhancing customer experience with hyperpersonalization
One of the biggest use cases driving the development and interest in machine learning is personalization. Customers are more likely to build a lasting relationship with a brand when they are able to make personal connection with a product or service. While the idea of providing a personalized experience is not new, success in doing so has become more difficult as the window of time to deliver relevant, personalized content continues to shorten.
Consider an online news publication with hundreds of thousands of daily readers. By ingesting and processing users’ behavioral data, the outlet can alter each user’s experience, storing various pieces of user event data to make changes to different pieces of the site. For example, real-time decisioning and automation exposes users with ad blockers to ensure alternative UI components are used to help engage the visitor on the site. Another example is providing personalized recommendations to users, empowering them to curate their own version of the site to include topics that they are most interested in. A hyperpersonalized system learns a reader’s behavior based on a combination of preferences like topic interests and alerts coupled with learned behavior through pageviews, click rates, location history, shared social media information, and additional web activity to identify suggested articles.
Taking personalization a step further, publications are using real-time analytics to implement headline A/B testing, where headlines are posted and tested in real time before the system selects the one with the highest response rate for future readers. News cycles today move quickly—an article can become old news in a matter of minutes, so optimizing page views in real time is critical to keep readers engaged.
Leveraging machine learning for ad fraud prevention
According to recent study, more merchants are focused on acceptance than prevention when it comes to fraudulent swipes. While this approach generates increased revenue in the short term by processing all transactions, it leads to longer-term losses for fraud repayment as well as a hit to brand reputation, the losses of which can be even more detrimental. On the other hand, being overly cautious results in false positives, which not only reduces revenue in the short term, it can drive customers away.
The cost of traditional fraud detection methods is rapidly increasing, and fraudsters are constantly innovating to stay ahead of countermeasures, compounding the issue. Making correct decisions to accept or deny a transaction, and reducing false positives, is crucial for both attracting and keeping customers and merchants. For in-transaction processing, the standards are high. The system must be capable of processing thousands of card swipes, network functions virtualization (NFV) taps, and online payments per second, with a time budget of only milliseconds, and strict consistency requirements—losing data due to nodes going down is not acceptable.
For example, when a person swipes his or her card to purchase a new shirt, a database immediately runs hundreds of input variables such as location, time of day, recent purchases, and previous transactions at that retailer through complex logic to determine whether or not to decline the transaction, all within milliseconds.
Increasing player retention and ARPPU
Gaming is the undisputed king of mobile applications. Just ask anyone under the age of 20 if he or she plays Fortnite. In 2017 alone, 80 percent of app spending was generated by mobile games, more than double that of PC gaming and more than triple that of console gaming, yet the overall average revenue per paying user (ARPPU) was only $7 compared to the top 16 percent of mobile games, which had an ARPPU of $50. Retention rates tell the same story—industry experts consider a strong retention benchmark to be at least 15 percent for Day 7 retention (meaning the number of users returning on exactly the seventh day after install), yet in 2017 most games hovered around 4 percent.
The average length of a player’s game session is only five minutes, so increasing engagement and driving long-term retention is highly dependent on creating an experience that is specific to a player’s ability and then providing timely, relevant, in-app offers relevant to that specific player’s needs in the moment it matters. To successfully achieve this, developers must have the ability to implement an adaptive gameplay, where the difficulty can be fine-tuned on a per player basis, ensuring that each session is sufficiently challenging without being boring or frustrating. From there, in-app offers can be optimized to provide the right offer in the exact moment of need, such as when a player is stuck on a level or has failed a number of times in a row. Players at that point in play will be more willing to purchase lives or watch an ad to gain a competitive advantage.
Real-time decisioning has a place in almost every transaction and interaction we have with data and technology today. In part two, I’ll explore where this level of immediacy comes into play with the internet of things (IoT), communication service providers and financial trading.
This article is published as part of the IDG Contributor Network. Want to Join?