What do machine Learning, data science, predictive analytics and MARKETING have to do with each other? A lot actually.
There is constant discussion about how all these forms of advanced analytics are shaping both the present and future of marketing; there is a reason for this. Successful marketing depends on the presence of relevant data about consumers and their response to marketing campaigns.
Therefore, marketing can’t remain unaffected by advances in data and analytics technologies and many marketers recognize this.
Still, many of the terms associated with this "data revolution" sound like buzzwords to many marketers.
- Should you be treating them as such? Or is it wiser to start exploring this space for any emerging opportunities?
- If so, which are the most relevant parts in the "data analytics universe" that are a good fit for marketing?
- Finally, once you discover them, what's the best way to apply these techniques in your organization for your own marketing use cases?
Topics of advanced analytics are often discussed in very technical terms. In this post, I‘ll present an overview of the top analytics techniques with a focus on the needs of marketers. I‘ll walk you through some specific examples to help you gain some intuition on how these techniques work in practice. We'll then discuss how data-driven marketers can integrate advanced marketing analytics into their work.
Advanced Analytics in a Marketing Context
To support marketing decisions and answer marketing-related questions, marketers work with various types of analysis.
Basic marketing analysis includes the calculation and monitoring of ratios. Examples include:
- conversion rate
- cost per acquisition
- cost per click
These metrics are frequently combined with segmenting traffic sources and filtering customer types as needed, then visualizing the relevant data.
This style of analysis helps to quickly answer questions and produce useful insights. In many cases this is sufficient; however, some questions require more complex logic to be answered, which is not possible to do by eyeballing a graph or manipulating a pivot table.
Advanced marketing analytics, on the other hand, can help marketers to automate and optimize an increasing number of marketing decisions and processes. They are based on statistical techniques and machine learning models. These are essentially mathematical recipes that can be applied to marketing data in order to spot the patterns and relationships within it.
Many of these advanced analytics techniques focus on predicting the future as opposed to looking at what happened in the past in order to describe or summarize that information. If humans were to undertake these tasks, it would either be impossible to carry them out from start to finish, or it would take a prohibitively long time.
Step 1: Clean up your data
The best investment before moving into advanced analytics is to carefully prepare the ground for it. You'll want to make sure your basic reporting needs are well taken care of. Having a solid automated data and reporting pipeline in place will free up resources, reduce human errors, and improve data quality.
The quantity and diversity of data also play a key role. The reason for this is that most of the advanced analytics techniques perform significantly better in the presence of larger volumes of granular data collected from a variety of sources. Remember that the outcomes you‘ll see from your advanced analytics deployments will be only as good as the data you provide.
Improvado helps organizations accelerate their transition to the most advanced forms of analytics
Need help to prepare for your transition to advanced analytics?
Improvado can help with all of these aspects of your preparation before you dive into advanced marketing analytics, from automating your marketing reports to collecting and storing granular level data.
An overview of the Top Advanced Marketing Analytics techniques
Advanced marketing analytics can be applied across all stages of marketing.
This section covers some of the most widely used techniques. Including:
- Customer Lifetime Value
- Marketing Attribution
- Conversion Prediction
- Anomaly Detection
Customer lifetime value
Marketing to the wrong prospects can be a very costly endeavor. You can use the conversion prediction techniques discussed above to produce a list of users who are likely to turn into customers, but how can you determine who is going to be the most valuable?
Marketers know that a very large percentage of customers —even if they are satisfied with a product or service— simply don't return. Those who do might later churn. Typically, it is a mere fraction of customers who stay faithful to the brand and even fewer who become real brand champions. This minority group is easily the most valuable. The Pareto principle (80/20 rule) is a good fit for this scenario: 20% of customers bring in 80% of the value.
Using customer lifetime value techniques, you are in a position to predict in advance the expected lifetime value of a customer based on a short history of transactions. Knowing this, you can minimize spending on unprofitable customers, optimize acquisition channels, and seek to reactivate customers with high likelihood to be profitable.
Are my marketing campaigns effective? How much should I spend on each of them? These are classic questions marketing needs to answer. As customer journeys become more complex and involve multiple touchpoints, attribution is becoming more relevant.
There are many variations of this technique, some involve simple business rules such as attributing all credit to the first or last click. Others involve more complex calculations and probabilistic approaches. What would have happened if marketing channel x was not available? How much would conversion be impacted? The idea is to explore multiple scenarios like this for all channels. Based on the results, credit can be assigned to each of the marketing channels representing the relative importance of each one.
Multi-channel attribution is suitable in an online setting where businesses keep track of metrics such as clicks, conversions, and click paths. In organizations that work with more traditional marketing media, an alternative technique used is Marketing Mix Modeling (MMM). It is based on a well-researched statistical technique, regression analysis and works with what-if scenarios. For example, what will happen to revenue if TV spend was to increase by x %? Having these answers on hand can inform the decision for budget allocations when you plan for future campaigns.
The concept of clustering fits naturally into marketing. It's all about helping marketers to segment prospects and customers in natural ways. Content, campaigns, and offers can then be created separately for each segment. One could come up with heuristic rules to generate various customer groups: "If the customer is a millennial, show content type A. If a baby boomer, show content type B."
This may have been a reasonable way to do things in the past, but today, having access to the right data, there are more options. With clustering, you can group customers intelligently by taking into account an arbitrary number of customer characteristics. These groups are naturally formed based on the calculated mathematical distance between the different characteristics. Customers with similar scores will be grouped together.
These characteristics, or features, can include age, income, spend, time since last purchase, etc. The same technique can be effective in clustering keywords, e.g., based on their organic ranking, the number of competitors, and opportunity score. Similarly, it's also possible to cluster products, campaigns, ads, and so on.
With conversion rates typically in single-digit percentages, conversion prediction is in no way a simple task. It's a lot like looking for needles in a haystack. To have a higher chance of succeeding, you‘ll need a lot of historical data about user behavior. Based on this information, these early behaviors associated with a future conversion can be identified.
Once users with behaviors that signal high likelihood to convert are found, you can prioritize and target them appropriately. Moreover, this technique is also good at discovering the factors that have the strongest impact on conversion. It might include a combination of gender, location, type of device used, or any other combination of relevant dimensions depending on the site and its users.
Marketing is becoming a real-time activity with high volumes of data involved. Display and search campaigns, for example, can participate in thousands of programmatic auctions every day and contain a high number of ad groups and keywords, each with its own conversion rate, spend, ROI, etc. A wide array of numbers that are constantly changing.
Most of the time, the change remains within the bounds of natural variation, but this is not always the case. As a marketer you need to be constantly alert so that as soon as something stops performing as expected, you are there to promptly take corrective action.
Anomaly detection uses statistics and machine learning to alert marketers as soon as key metrics such as conversion rate, revenue, and traffic deviate too much from what is expected and accepted. This technique sees all these numbers as statistical time series and knows how to spot seasonal and weekly patterns as well as how to avoid triggering false alarms. This way, if you have outliers in your data or segments that under- or overperform, these can be swiftly flagged.
Forecasting is everywhere: financial markets, economic indices, corporate sales, etc. So, it's no surprise that this technique can also forecast online traffic, conversion, revenue, and other metrics marketers care about. Similar to anomaly detection, forecasting uses historical data to predict trends. This, however, is not always possible and, as a result, forecasts are sometimes not accurate.
In order to make interpretation of predicted results more flexible, forecasting techniques provide bands within which forecasted data can range with given probabilities. If uncertainty is properly accounted for, forecasting can be used as a technique to help you better adjust your future campaigns and targets.
How to Implement Advanced Marketing Analytics
As businesses grow and mature, the need for more data-informed and automated decision-making increases. Advanced analytics is a way to achieve that goal and gain a competitive advantage.
In the past, it was hard to implement advanced marketing analytics without understanding the science behind it or being able to code a custom solution. The alternative was to pay immense license fees to vendors, which typically only large enterprises could afford.
The playing field now is more level with a variety of tools and tech, both commercial and open-source, that can make this space more accessible. Though, it’s not all that simple and easy. There is always the risk of wasting time and resources by deploying advanced techniques to provide sophisticated answers…to the wrong questions. Therefore, developing some intuition around these concepts is critical before being able to effectively apply them in practice. I hope that this article provided you with some insight and helped you to gain this intuition.