In my previous post Marketing Analytics Part I - Good Metrics, I talked about how businesses are increasingly turning to data and analytics to gain deeper insights into their strategies and outcomes.
In this post, I will continue with the theme of Marketing Analytics. As a companion to the Good Metrics post, I will consider the benefits and frustrations that having dedicated marketing analytics capabilities can bring.
As the Marketing industry continues to evolve, businesses are increasingly turning to data and analytics to gain deeper insights into their strategies and outcomes. Marketing analytics involves the systematic analysis of data from various marketing channels to understand performance, optimize campaigns, and drive informed decision-making.
Although marketing analytics covers a broad range of areas, this blog post will be the first of 2 posts. This first post will discuss the importance of marketing metrics in marketing analytics and focus on how to pick good metrics and avoid bad or inappropriate ones. The second post will consider the benefits – as well as the frustrations – that building or having marketing analytics capabilities can bring.
I always think of the gold rush in the 19th century when I think of data mining. Just as these miners were looking to strike it rich in gold, data miners are hoping to strike it rich in insights. Data miners are trying to find that "golden nugget" of data or information that will help boost their business profits or take them from a struggling start up to a million or billion dollar going concern.
Data mining has been around for a long time but with the improvements in computing power and in the era of big data, it is a key component within data science. It is also used interchangeably with the term Knowledge Discovery. There are really two big areas that data mining focuses on: descriptive modeling and predictive modeling.
Attribution modeling - or the allocation of a conversion event to a customer interaction - is important to determine how a marketing activity is performing. The challenge is that customers often have multiple interactions prior to conversion. Combine that with the many different ways you could attribute conversion and you end up with a fairly complex situation. So, do you just throw your hands up in the air and give up? I recommend not. Just take a deep breath, grab a cup of coffee or tea, and finish reading this post.
But before I get into the details on the different types of attribution models, I should mention there is no one model that is considered the best. Each model has its pros and cons, depending on the situation and business need. But there are ways you can compare them. I came across this method of comparison from Klaviyo where they suggested that different attribution models can be regarded on a continuum based on simplicity vs precision.
The following chart uses the idea of Klaviyo’s Simplicity-Precision continuum to show where some of these most common attribution models fall. Below the chart is a table (clickable image) that provides the descriptions and illustrations of how each attribution model works along with their pros and cons.
Are you drowning in a sea of acronyms—KPI, OKR, ROI, CPC, ROAS, MRR? It's a cacophony of letters, a metrics meltdown. How do you make sense of it all? Let me help you navigate two categories of metrics - KPIs and OKRs.
Let's start at the beginning, shall we? Key Performance Indicators (KPIs) are the metrics that monitor the performance of specific aspects of your business. Imagine them as the dashboard of your car, telling you how fast you're going or how much fuel you have left. KPIs measure the tangible outcomes like sales, customer satisfaction, and productivity. Want to know how many products you sold last month? Check the KPIs. Curious about customer satisfaction? The KPIs will have that information too.
But KPIs are not a one-size-fits-all solution. Sales, marketing, customer support — each department has its own set of vital signs. It's about tailoring the KPIs to your business DNA, ensuring they're the true pulse of your operations.
Now, enter Objectives and Key Results (OKRs), the cool, trendy cousin. OKRs are ambitious, goal-oriented frameworks designed to drive organizational alignment and success. OKRs are the dreamers, the visionaries. They set the stage for the big picture, outlining ambitious objectives that push the boundaries of what's possible. The key results, like the milestones on a road trip, help you track progress toward those lofty goals.
An Objective (O) might be to grow from $75 million to $100 million dollars in revenue. A Key Result (KR) for that particular revenue objective might be quantitative: improve customer retention by 10% while another KR might be qualitative such as: create a roadmap of Sales and Marketing strategies to increase revenue.
In my last two posts, I introduced measures of central tendency and the normal distribution. Why? Mostly because they are important statistical concepts but also because they lead nicely into the topic of statistical testing.
Statistical testing is very important because it provides a level of confidence providing that results from experiments are not due to chance or luck (good or bad). There are at least 4 key steps to perform in statistical testing. I say at least because for some of these you could easily break them down into smaller steps. Oftentimes, these are implied or done so seamlessly that it doesn’t seem like they are steps at all but trust me, they are. Each step will be discussed in turn but they are:
Just a note here, I will not be going into detail behind the math of these. My goal is to explain the concepts behind statistical testing and provide two examples of their use.
As I mentioned in my previous post on measures of central tendency, this is part 2 of a 3-part post on some basic statistical concepts.
If you’ve ever been graded on a curve or have heard the term “bell curve”, then you know a bit about the normal distribution. This post will focus on the concept of the normal distribution and some very basic elements around it. Although this concept might seem fairly straightforward, there is a lot of detail and math that goes into it, which I will not be going into here. Also, the normal distribution is only one of MANY distributions, which I also won’t cover.
I have to say, there is no shortage of data being provided during the COVID-19 pandemic. Every day in Alberta at 3:30pm our Chief Medical Officer provides updates, including the number of new and recovered cases. There are also Kaggle competitions trying to get people to build data models to better understand this virus and predict its spread. Not to mention all the research going into its causes and creating a vaccine. And then there is my current favorite site – updated daily – and it includes geospatial information (including an interactive map), age group, and gender breakdowns.
Data dashboards are probably one of the most common analytics tools used by companies and interestingly, they are not unlike your vehicle’s dashboard. Think about that for a minute. In your vehicle, you have a lot of important pieces of information directly in front of you that help you make decisions.
I’ve recently been working on getting my Data Science Certificate from NAIT, which is a local polytechnic school. In one of our classes our instructor gave us the Titanic competition from Kaggle to experience what an end-to-end machine learning project is like. It has been frustrating, challenging, sometimes daunting, but yet super helpful. Even now, a week after the class I am still working on my model trying to get the accuracy higher and improve my ranking on the Leaderboard. And still wondering how on earth so many people got 100% accuracy?