I have a confession to make: Marketing analytics isn't my strong suit. I've always been a marketer driven by creativity, not data, but a venture into Facebook ads made me realize that analyzing data isn’t so bad.
Nowadays we have several marketing analytics tools and data analysis guides that make the task so intuitive that even a data-phobe like me can become a professional marketing analyst. Here’s a guide on how to analyze marketing data for those who have zero knowledge of stats or programming.
Step One: Define your goals/KPIs
Usually, in marketing data, we have a very clear goal of what we want to achieve. Ultimately, it's to increase ROI (return on investment), but there are many smaller steps in the equation such as achieving higher click-through rates, newsletter sign-ups, followers, Google rankings, product conversion rates etc.
We need to define which variables we want to maximize.
We also need to define which variables we need to minimize. Almost always this will be cost.
We’ll be using these variables later on - so keep note of them!
Step Two: Explore Your Data
Look through your spreadsheet.
Look at each column.
Think: Which variables will have an influence on our KPIs? Which ones can we control?
Step Three: Determine How Each Variable Influences Your ROI
Polymer’s Auto Insights tool is designed specifically for marketers to be able to find the best combination of factors that influence their ROI.
The beautiful thing is, it doesn’t require any programming knowledge or advanced knowledge of stats. Using the drag and drop interface, anyone can become a data pro in minutes. Here’s how:
1. Firstly, head over to the Auto-Explainer tool within Polymer.
2. Input your “goal” into the “Metric to Maximize” field. This should be metrics like ROI, conversions, sign-ups etc.
3. Change the operation to “average.” (Most of the time, you want to use average. Sometimes, you want the SUM. The best way to learn which one to use is through experimenting). Luckily in this case, it doesn't matter whether we use SUM or AVERAGE.
Now, here’s where the magic happens:
Polymer will spit out a summary of the data for you - showing all the variables that influence your goal and order them from most important to least important.
The variables at the top have the biggest influence on your goal and demand special attention.
The ones at the bottom have the least influence on your goal metric, but that doesn’t mean you should ignore it. Least influence could still mean some influence, just not as much as the top variables.
Here’s an example:
Using the sample Google Ads data within Polymer Search, we want to see which demographics, landing pages and ad-spend strategies are responsible for the highest ROI.
There is no “ROI” column in the data, so in order to do that, we’ll need to put in “conversion value” into the “metric to maximize” field, and “cost” into the “metric to minimize” field. This will divide conversion value by cost, giving us ROI:
Here’s the report that Polymer auto-generates:
Notice how “country user” is at the top, followed by “date” and “search keyword?”
This means “country user” has the highest impact on ROI, followed by “date” and “search keyword.”
This is where you need to look through each one individually by clicking “See Details” for each one. This will give you a full list of countries, dates, search keywords etc. to analyze.
Let’s see which countries are producing the highest ROI and which ones are performing poorly.
The things you should pay attention to are the ROI column and the # of Results.
I would sort the ROI column from largest to smallest.
At the top of the list, we can see that India is doing very well - with an ROI of 46! The number below (+587%) gives us context on where this country stands. It means India is performing 587% above average.
Now looking at the number of results, we only see 1 result for this country. This could mean that it’s an outlier. In this example, it’s not actually an outlier because the data is divided into ad groups rather than individual users. The ad group targeting India has over $10,000 ad spend, making it a decent sample size.
Next: we can look at the worst performing countries which are: Mexico, Spain and Costa Rica. These countries are performing over 50% worse than average!
This data provides us guidance on the next steps. We either remove these countries from the ad spend, or we try to diagnose the problem and figure out “why.”
Step Four: Find the Top Performing Combinations
Now that we know how to analyze the effect of individual variables on ROI, it’s time to find the best combinations of variables.
For instance, it’s useful to know that a certain age group prefers our product, but it’s better to narrow down our target audience into something like this:
But how do you find the top performing combinations? The answer is exactly the same as above, but by also utilizing the ‘breakdown by segments’ feature.
If I wanted demographic information, I would input “gender, age, country and device” into the “breakdown by segments” field. Like this:
This will give us the top performing combinations of whatever factors you want to see!
Again, I’d sort the information by ROI from largest to smallest.
Which variables should you include in the “breakdown by segments” field? It depends on the dataset and you’ll have to use your “marketer’s intuition” for this part. There’s no right or wrong answer.
It’s often a good idea to start with the demographic information like I did. Start simple, then build upon that knowledge.
The next question I would have about the data is: What’s the relationship between “search keyword” and “gender?” Do certain genders convert better with certain keywords?
With marketing data, there are many instances where correlation =/= causation.
The only way to see if there's a causal relationship between variables is to do further testing by running more ads. Make sure to use a control group!
Not all correlations are worth testing, so it'll be up to you to put on your thinking caps and decide which correlations are worth testing.
The biggest mistake people do when testing is not controlling all other factors. Example:
They find targeting the US results in higher ROI than targeting the UK, but in reality, the ads targeting the UK are only targeting the poorer cities where people have less to spend, whereas the US ads are targeting rich cities.
Be sure to control for other variables when testing and avoid these mistakes.
If you have any other questions on how to maximize your results with marketing, feel free to email us at: email@example.com