1. Audience segmentation
Text and sentiment analysis can be applied to almost any channel, but there are specific channels they are more aligned with. Audience segmentation, on the other hand, applies to all channels at one point or another.
2. Audience segmentation basics
There is a saying in marketing: "one-size-fits-all advertising delivers one-size-fits-all results."
Marketers can tailor advertising via audience segmentation. Segmentation essentially means identifying groups within a larger audience to deliver tailored marketing.
Marketers strive to create relevant messages for each segment, and making a message relevant is much easier with segments (versus the entire customer base.)
Marketing Analysts are key to the segmentation process and use analysis to create data-driven segments where possible.
3. Segmentation data types
There are many data sources to draw from for audience segmentation purposes.
One common source is demographic data, which includes attributes like age, location, and gender.
Psychographic data enables segments by customer lifestyle, opinions, and values like "enjoys working out."
Behavioral data is a very rich source of data for audience segmentation. Analysts can separate customers that engage with ads, or people that abandoned the shopping cart during checkout, for example.
A common pitfall is designing segments that are too granular to be actionable for marketers. A segment that is age 36-point-5, lives in a specific postal code, and has tried to purchase in the last 3 days, is too specific for marketers to use in most cases.
4. Cluster analysis
Let's say that we want to create segments based on a marketing KPI, such as conversion rate. Conversion rate is the ratio of customers who complete a goal out of all customers.
The problem here is that we don't know which attributes explain the conversion rate behavior.
In this scenario, we should use a classification modeling approach called cluster analysis. Cluster models optimize based on how similar data is within the same group (called a cluster), and how different data is between groups.
Cluster models will automatically group data into segments, but analysts must determine how many segments are needed.
After this step, we can look at common attributes of each cluster. Did everyone in one cluster engage with LinkedIn or business podcast ads (indicating interest in products for business use)? Did everyone in another cluster use Snapchat or visit product pages for families? This could indicate a cluster aligning with personal products. The cluster attributes can then be applied to marketer tools!
5. Audience targeting
Once segments are defined, marketers can re-purpose segment definitions in their campaign planning and execution.
Remember how DSPs (Demand Side Platforms) and DMPs (Data Management Platforms) manage segment definitions across marketing channels?
Having standardized audience segments in those tools enables tactics like retargeting and look-alike modeling.
6. Retargeting and look-alike modeling
Retargeting shows ads to target audiences based on behavioral data. Every time we get an email asking, "Did you forget something in your cart?", that is a form of retargeting.
Look-alike models use characteristics of segments to show ads to larger segments with similar characteristics. One example could be seeing Adidas ads after purchasing Nike in the past. Analysts can compare the performance of these tactics to see if they outperform normal ads for the same segment!
7. Let's practice!
Now we can practice audience segmentation for ourselves!