Automating Monitoring for Proactive Revenue Protection

Featured article by Avi Avital, VP Customer Success at  Anodot

As online commerce has exploded in the last decade through micro-transactions, single purchases, and subscription plans, the need to monitor revenue-related metrics is greater than ever.

While this isn’t a groundbreaking assertion and many companies agree on the importance of revenue monitoring, the challenge is that revenue streams are highly complex, fragmented, and dynamic, which makes them notoriously difficult to monitor.

Revenue monitoring includes  tracking billions of  events each and every day across products, segments, payment providers, and devices in real time. This includes but is not limited to changes in advertising performance, payment gateways going down, and shifts in purchasing behavior

These days, it’s also common to have more than one business model, so whether you include micro-transactions, subscriptions, partners, or affiliates in your revenue ecosystem, small incidents and anomalies can have a massive impact on your bottom line.

Traditionally, revenue monitoring has been done with static BI dashboards, which involved setting manual thresholds for alerts, updating them manually as new anomalies occurred, or simply summing up and reviewing performance at the end of each month.

The issue that many companies have realized is that these traditional monitoring techniques are simply too reactive for today’s fast paced, data-driven business world. The impact on revenue is too great to leave it up to these manual monitoring, analysis, and detection methods.

Instead, as we’ll discuss in this article, many forward-thinking companies have embraced AI and machine learning for revenue monitoring due to its numerous advantages including scale, granularity, and accuracy.

What is Revenue Monitoring?

In today’s digital world, you probably already know that running a business, regardless of which type of business model you use, involves tracking and monitoring a huge number of metrics and KPIs.

This will often include customer experience metrics such as monitoring logins and active users, as well as monitoring website performance metrics such as bounce rate, exit rate, time on site, and so on.

Revenue monitoring follows a similar process, although it is arguably more important than any of these performance metrics, and involves tracking the metrics and KPIs that influence a company’s overall revenue. This includes tracking metrics such as conversion rate, churn rate, average order value, and many others.

The challenging part about revenue monitoring is that each one of these metrics isn’t machine generated—it’s created by human behaviour, and as a result is incredibly complex, dynamic, and volatile.

In the past, companies have tried to feed these metrics into monitoring solutions built for machine-generated data such as IT and APM monitoring solutions. Similar to traditional BI tools, this often involved setting static thresholds, and then having the DevOps team manually adjust these as new events and incidents arose.

Not only does this approach pull valuable human capital from your technical team, but as many companies have realized, these systems simply don’t have the accuracy, granularity, or scalability that’s required for revenue data. In addition, since each of these metrics and KPIs has an impact on revenue, it’s simply too costly to leave anomaly detection up to manual thresholds.

To solve the unique challenges posed by this human-generated data, companies have realized that using AI and machine learning is the only way to effectively monitor revenue-impactful events and stay ahead of the competition.

AI & Machine Learning for Revenue Monitoring

The fact is that each one of the revenue-related metrics that needs to be monitored each day is constantly changing its normal behavior. This can be due to things like seasonality, changes in consumer behavior, or any number of other reasons.

In order to deal with this ever-changing complexity, one of the primary benefits of unsupervised learning is that it’s completely autonomous. It can learn the unique behavior of each individual metric, autonomously adjust to new norms of behavior, and also find correlations between various metrics and events.

In particular, an AI-based revenue solution has a few main advantages over traditional monitoring techniques. These include:

– The solution can learn and monitor each revenue stream on its own

– AI can monitor both success metrics (i.e., purchases, subscribers, etc.), the events leading up to the successful action (i.e., the entire purchase funnel), as well as ratios between events such as the conversion rate

– When issues inevitably arise, AI can correlate most metrics and events in real time so that you have the shortest possible time to resolution. For example, incidents can be correlated with individual cost metrics so you know exactly what needs to be resolved.

While unsupervised learning is entirely autonomous, the fact that revenue-related events often require context, it’s best to pair it with a team of humans for optimal revenue monitoring.

Now that we’ve discussed a few of the advantages of AI, let’s turn our attention to how it actually works in practice by reviewing the revenue monitoring logic:

– Unsupervised Learning: Firstly, unsupervised learning is a branch of machine learning that takes in unlabeled and unstructured data and can uncover patterns, structure, and insights from the data

– Real-time monitoring of 100% of the data: Next, the monitoring solution is applied to each and every data point, as anomalies can occur across various metrics. For example, the solution can monitor and correlate events for each individual product, location, and device.

– Seasonality Detection: Since revenue-related metrics are often human-generated, the AI must detect changes in normal behavior, such as seasonality, in order to prevent alert storms from false positives and false negatives.

– Anomaly Scoring: Finally, the AI scores each anomaly by significance so that the team is only alerted of the most significant events.

In short, by detecting events and performing a root cause analysis, your team can find and resolve incidents before they significantly impact revenue.

Revenue Monitoring Use Cases: How Companies are Staying Ahead of the Curve

Now that we’ve reviewed the high-level logic of an AI-based revenue monitoring solution, let’s review several real-world use cases of how companies are using it to stay ahead of the competition. In particular, we’ll look at several use cases in two common business models: subscription-based monitoring and monitoring micro-transactions.

Subscription-Based Monitoring

New Subscriber Monitoring: New subscribers are the lifeblood of subscription-based businesses, which means that monitoring for spikes and drops in registrations is essential. For many businesses this means monitoring events between countless locations, languages, and devices — something that is particularly well suited for an AI-based system:

– Churn Rate Monitoring: Equally important for subscription businesses is monitoring how many customers you’re losing over time, otherwise known as your churn rate. Monitoring for anomalies in churn can not only help you prevent revenue loss, but it can also help you identify issues that will significantly improve the customer experience.

Micro-Transaction Monitoring

– Add-to-Cart Monitoring: One of the best leading indicators of revenue is add-to-cart data. In the image below, we can see the AI is monitoring for significant spikes and drops for each traffic source, device, country, and product. Each anomaly is scored so your team knows exactly what to focus on.

– Conversion Rate Monitoring: Conversion rate is another high-value metric that can be improved with AI-based monitoring. If there’s a sudden drop in conversion rate, this can mean something has broken in your checkout process. By pairing real time alerts with a root-cause analysis, issues related to conversion rate can be detected and resolved much faster, ultimately leading to recovering a significant amount of potential lost revenue.

– Payment Gateway Monitoring: With so many payment gateways on the market today, it’s likely that your company uses several different providers. Regardless of which one you use, an AI can monitor its API and can catch drops in payments from a specific provider in near real-time. For example, we can see in the image below that the solution is monitoring and correlated events between both the payment provider and a spike in API errors:

Summary: AI for Revenue Monitoring

As we’ve discussed, the volume of complex data that’s generated each and every day presents new challenges for effectively monitoring revenue.

Since revenue-related metrics and KPIs are driven by human behavior, its dynamic and volatile nature requires a solution that is monitoring 100% of the data and constantly adapting to changes in its normal behavior

While companies have traditionally relied on BI platforms and IT monitoring solutions, these systems are too static and require too much manual labor to be effective in practice.

Instead, since monitoring revenue has such a significant impact on the bottom line, many companies have turned to AI and machine learning in order to stay ahead of the competition.

Author Bio

Avi Avital, VP Customer Success at  Anodot

Avi has managed the technology and business operations of global organizations for more than a decade. As VP Customer Success, he leverages his experience building large-scale analytics and AI systems at PayPal and DHL, to lead Anodot’s global CS team. Avi’s unique strategic and creative approach, coupled with his experience and passion for making a difference, help him deliver high value to customers, employees and businesses.