Michael is the director of e-commerce for a leading retailer in the UK. He generally starts his days off by checking his key e-commerce metrics, looking over revenue charts, keeping a close eye on conversion rates simply to analyse them and make sure everything is ok. But what happens when something isn’t ok? What kind of steps does Michael have to take? Let’s take a look.
It’s another day at the office and Michael spots something. Conversion rates are down in comparison to last week. In a hurry, Michael decides to email his marketing director Kyle:
“Kyle, conversion rates are down, what going on?” Kyle writes back,
“Not sure, I am looking into it”
Kyle gets back to Michael a few hours later with a reply stating:
“There is a problem with checkout form submission for Apple iPhone users, you need to talk to IT”
Michael forwards the email to IT only to receive a reply stating “We’re looking into it”
The following day, IT gets back to Michael to let him know the issue is resolved.
It’s important to take a step back and understand why this is inefficient. As with the case above you can see that most of the time was spent diagnosing the problem, emailing back and forth only to fix the issue the following business day. Michael was relieved they sorted out the issue yet upset about the revenue lost in the meantime. Could have Michael solved this issue more quickly? Let’s look at the same scenario had Michael used Millimetric’s anomaly detection software.
A Different Type of Day
Another typical day, Michael wakes up to an email in his inbox titled “Critical Alert: E-commerce Conversion Rate 27% drop”, Michael clicks on the anomaly and is redirected to Millimetric’s website where he clicks on “root cause” to see the source of the issue. He sees the mobile device related breakdown of conversion rates and sees “iPhone down 27%” he quickly shares this anomaly directly with IT, bypassing Marketing because he now knows it’s an IT issue. IT replies “That was a quick fix, thanks for the insight”
By simply using Millimetric Michael could have bypassed a whole day worth of lost iPhone-sourced revenue.
This is just one of the thousands of aspects where Millimetric’s anomaly detection diagnosed the problem, triaged it and delivered an actionable insight within minutes!
Great story, but what is Anomaly Detection?
An anomaly, or outlier, is defined as something that does not fit the expected pattern. In other words, an anomaly is a deviation from usual data characteristics. An example of this could be an unexpected increase in advertising costs or a sudden drop in sales. At its core anomaly detection is about sorting data into groups of regular and irregular data. A data point that falls in the irregular data group is called an anomaly or outlier.
How Can Millimetric Improve Your business?
Millimetric automatically notifies you if an anomaly has been detected within the provided data. Simply add a data source (e.g. Google Analytics, Facebook Ads etc.) and our Machine Learning Algorithm starts to understand the normal, by analyzing data from the past. Then Millimetric detects unexpected changes, incidents or problems and reports the must-see data on a daily basis.
The ability to quickly find and resolve business incidents enables you to prevent revenue-loss and keep costs at acceptable levels. But, its not only about resolving issues that arise, Millimetric also provide positive anomalies. The idea is to create levers within your company to help you lessen the bad and increase the good.