Anyone can get more Return on Ad Spend (ROAS) and lower customer acquisition costs by unlocking the power of Anomaly Detection and Root Cause analysis
It’s easy for the modern marketer to get completely overloaded with data.
For marketers, this can feel paralysing – especially when the data is coming from so many different places. Spending hours digging through it to understand marketing performance is not only less than appealing as an activity, it is often impossible because of the time it takes and the other priorities marketers have to juggle.
According to a survey from Gartner, 54% of marketing leaders say Analytics hasn’t had the influence they expected. Research also found that 31% of the marketers face the challenge of “too much data to analyse” when optimizing ad performance.
Machine learning can take the legwork out of the process and ensure you identify issues and opportunities immediately with
Machine learning flips this situation on its head. It removes a big chunk of the legwork, giving you back time to spend figuring out what the issues are and finding opportunities immediately. In other words, anomaly detection and root cause analysis for marketing allow you to focus on what’s next rather than looking in the rear-view mirror. Hindsight isn’t always the most helpful tool when you need to focus on the future.
How marketers generally manage data without ML
Marketers use analytics to speedily identify how their campaigns have impacted their KPIs, and where they need to improve.
Let’s say that as a marketer, I would like to use marketing analytics to answer some questions like “On which channels should I spend more/less to increase ROAS?” or “Which brand should I focus on to reduce cost per lead compared to last week/month/year?”.
Now, I could use a traditional BI tool and get some dashboards built in order to give me some insights.
We can see here a significant increase on the 13th of August.
But dashboards are “passive”. Meaning I would need to keep an eye on them and spend more time analysing the data if we want to explain why a drop or increase occurred on a specific metric.
And that’s where anomaly detection for marketing comes in.
How Anomaly Detection and Root Cause Analysis can transform your process
If I use anomaly detection, I can automate the process of identifying unusual behaviour in my metrics. So, I don’t have to manually check reports, I’ll just be notified when there is an anomaly. I wouldn’t need to stare at the dashboard to know that there was an issue.
An alert notifies me directly that sales are unusually high on 13th of August through Anomaly Detection.
The natural second step is to get deeper in the data.
I need to get more details about the anomalies and drill down to analyse the root cause of the issue. This is where it really changes the game:
- I can view how the KPI has changed from one period to another.
A simple day to day comparison is showing me how big the difference was compared to last week
- I can understand where in the customer journey the performance change has occurred – is it a marketing issue or opportunity, a site issue or a payment issue?
Root cause analysis is showing me what led to the increase in sale: spending more as well as a better click through rate are the main drivers behind the good performance
- I can drill down further to see additional dimensional splits of the data to pinpoint a problem or opportunity
The combinations of Brand=’Hugo Bross’ and Channel=’Paid Search’ had a significant impact on the increase in sales.
- I can drill again to see the individual changes and the impact on my goal
We can drill in the metrics to know how they are involved in the increase and see that there is still room to improvement with the average order value.
All the information I get back using Machine Learning for Marketing can then help my company to take the right decisions in order to reach every marketer’s target: reduce customer acquisition cost and increase ROAS.
With this information at my fingertips, I can choose to double down on areas that are generating a great return; or, I might find certain channels and campaigns don’t perform, and I’ll stop activities there and redirect my budget.
Let the machine do the tedious work
Be honest with yourself – how much do you really need dashboards?
We’ve seen that based on a simple example, with traditional BI tools you need to build multiple dashboards with different dimensions and granularity. And you have to