In this first post of our new series on overlooked incidents and AI analytics, we’ll discuss how intelligent monitoring of your metrics can open the door when opportunity knocks…before it walks away.
“The main thing is to keep the main thing the main thing”
– Stephen Covey
Or at least that’s the intent when starting off. While it’s technically true that revenue and cost are only two key performance indicators (KPIs), it’s also true that both of those metrics are in fact aggregates – or more accurately, summaries. At such a high level of aggregation, there’s zero actionable insights because your team can’t directly alter either of those two metrics, but rather the myriad of little decisions which can tune the performance of the complex engine which is your business.
Without access to much more granular data, you’re forced to make specific actions with incredibly vague information. The solution, of course, is to monitor a much higher number of much more specific metrics. This approach, however, can turn into an embarrassment of riches if your monitoring solution can’t keep up all the signals that will now need monitoring.
Metrics monitoring can easily become an overwhelming task
There’s an example we like to use which really drives home this problem of scale: Let’s say you’re an analyst at an ecommerce company and at the moment you’re monitoring only two KPIs – the number of purchases, and revenue. You’re a smaller e-tailer, with only 50 products total spread out over 10 categories, and you’re focused exclusively on the United States and want state-level data. As a company that does business on the web, you’ve already been bitten by platform and version-specific problems, and thus you want statistics from eight operating systems (four major Windows versions, 2 desktop Mac OSs, and one version each for both iOS and Android) to help you identify and fix future problems much faster.
So how many total metrics will you be monitoring? The answer is:
2 X 50 X 10 X 50 X 8 = 400,000
As you get more granular, the number of permutations increases rapidly; in the off chance that Puerto Rico becomes the 51st state – that number will increase by 8,000.
400,000 metrics is far too many for manual metrics monitoring via traditional BI dashboards, alerts and teams of data scientists and analysts to be practical. Fortunately, AI-powered automated machine learning solutions are able to take human eyeballs off the dashboards because these real-time analytics solutions are able to accurately and automatically detect the anomalies in all that time series data. Those anomalies are the real signals in your data upon which you need to act.
Accurate and real-time anomaly detection, coupled with the ability to correlate related anomalies across multiple data sources, is ushering in a new time in business intelligence when no data-driven organization should be surprised by an unexpected business incident ever again. The anomalies in the data which point to the opportunities in the market can now be found. Think of it this way: if you monitor everything, you can detect anything, especially events you didn’t know you had to look for. This is the real power of AI analytics.
Absolutely no metric is overlooked
Perhaps for a better perspective, let’s discuss a real-life example of how machine learning-based anomaly detection can help a business gain actionable business insights.
When a celebrity endorses a product on Instagram, the free positive buzz can really drive up sales, but only if the reaction is in time. A large apparel conglomerate learned that the hard way when their BI team discovered the endorsement…two days later. If they discovered the sharp uptick in sales for that product and the rapidly dwindling inventory of that product in one of their regional warehouses in real time, they could have capitalized on the opportunity by increasing the price or replenishing the inventory to keep the customer demand fed.
Now, that same Fortune 500 heavyweight is an Anodot customer and hot opportunities like that don’t slip by anymore. According to the Data Analytics Director, “With Anodot, we get real time alerts for sales spikes…or when an impending snow storm causes a decline in in-store purchases in the Midwest.” Switching to much more effective metrics monitoring has obviously added to their bottom line.
When you have a data scientist in the cloud, any user can easily and automatically gain actionable business insights. Whether it’s a celebrity endorsement or signs of premature equipment failure from a swarm of IoT sensor devices, AI analytics tools are data agnostic and can find the signal hidden in your time series data. With these tools, you can extract not only actionable insights, but ultimately increased revenue from the business incidents discovered in your data.
And isn’t that supposed to be the main thing?