Personalized marketing, including the targeted address of individual customer groups, are issues that concern everyone in marketing day in and day out. However, the question is frequently raised as to what criteria should actually be applied to form useful segments? Or are there already some segments that cannot be monetized as expected. In this article, you will learn about the key points of segmentation and what you should pay attention to if you want to find your very own profitable segments for your business case.
Basically: Your most valuable segments are as individual as your business model and your company history. A segment that works well in case A might be useless in case B. In other words, ultimately you have to discover and define your own rules, which will help you to create an actual uplift. But no worries, there are some general approaches that can help you achieve this.
1. Segmentation by Activity
One approach, generally applicable to all online business models – regardless if portals or online shops – is user activity. In other words, when users last had contact with your site or brand. It is crucial to first of all consider the contact points in the individual channels – for example, when a user last opened an email, the last time they visited the website and the last time they saw an ad. Why is this important?
Especially in the area of email marketing, it often happens that customers stop reading newsletters, but still continue to regularly check out the online store. This may be, for example, because they are already familiar with the shop and know exactly what they need. They are not really interested in any additional promotions. Or the user has just generally decided that they want to deal with fewer advertising emails. The list of possible reasons is endless. As a result, it often happens that from the perspective of the newsletter, such customers are classified as inactive and are sent “we miss you” style emails. The customer will be left wondering what this is all about, because they were only recently on the site and maybe even bought something.
Therefore, a decisive point of contact for classifying activity is invariably, the last visit to the site. This is subsequently followed by recent interactions in a variety of other channels, which can be used to reactivate users in individual channels when they are no longer active on the website. In the above example of newsletter subscribers who no longer read the newsletter, you could use the website and an email campaign to offer users the opportunity to compile the content of the newsletter themselves. By this we mean that they could select the categories that interest them anyway and then perhaps subsequently adjust which information they want to receive. For example, new products or special offers. You can try and get back those totally inactive users, who have not been on your site for a long time, with attractive offers via channels such as Facebook, email or display ads. The marketing message which is to be applied here, must be tested.
The final challenge for segmentation by activity is that of defining when a user is considered to be inactive. Again, there are no universally applicable guidelines. Because, due to the products offered, the average visit frequency of an online food retailer is potentially much higher than that of an online furniture shop. Therefore, you need to determine meaningful criteria individually in the context of the behavior of your user universe. That means, what is the cycle in which the most active 30% of your users come to your site. And how the 30% most inactive users behave with regard to visit frequency. And in what context the 40% of average users come to the site. How well do you know their behavior? Are you able to classify how active a particular user is, and when there is a need for action, based on the “last page visited X days ago” metric?
2. Segmentation According to RFM Score
A second, promising segmentation is the one according to RFM scoring. I am sure you are familiar with the recency-frequency-monetary ratio model. You can see when your customer has made the most recent purchase, how often they have made purchases during their customer lifetime, and what sales or margin they have thereby generated.
In Customer Relationship Management (CRM), it is recommended that you basically work with margins. Because only these provide information on actual customer value, the so-called “Customer Lifetime Value”, an increase of which is, or should be, the primary target in CRM. These three factors are considered in relationship to one other and ultimately result in a score between 1 and 5, which tells you how good or how valuable the respective customer is.
Once you have categorized your customers according to an appropriate scoring model, there are two options: Segmentation by total score, i.e. the RFM ratio, or by individual scores per factor. Which version makes sense and when, will depend on the objective of the respective campaign or brand message. If, for example, it is a “sale” campaign, segmentation by overall score is appropriate. Your best customers who have a score of 4 or 5 can be given somewhat earlier access to the Sale category and you can advertise it in an exclusive pre-sale campaign. You can grant your customers with lower scores of 1, 2 and 3 access a few days later and you can then advertise unprecedented discounts of up to XX%. If however, your campaign is generally designed to develop not so great customers into better ones, segmentation in terms of individual scores makes more sense, as a poor overall score may be due to a variety of reasons.
A few examples: Customer group A buys often, has also made recent purchases, but exclusively from marginal categories or on sale. Customer group B has a good margin, has made a purchase quite recently, but so far has only purchased once. Customer group C also has a good margin, has purchased often, but their last purchase was a long time ago.
Each of the three customer groups calls for a different approach: Group A should, for example, receive a discount coupon for a very high-margin category, which cannot be used for special offers. This group, for example, can be approached for previously unknown product ranges. Group B should be given an incentive for their next purchase, which ideally should have a very short time limit. In some cases, a very high incentive may be worthwhile. It is often the second purchase that is the hurdle to cross to becoming a regular customer or at least to make a third or fourth purchase. Finally, Group C still needs to be reactivated. Here it is necessary to analyze in advance, the possible reasons for their discontinuing purchasing. Was there a problem with the service? Then an apology campaign might be helpful. If you have no idea of what the reason may be, what often helps is a survey at the first resumed contact with a follow-up incentive to purchase.
3. Segmentation by “Predictive Segments”
If you are a bit more advanced in the segmentation process, and have already learned how to specifically convert individual customer groups to a higher CLV, you can use past user behavior to make predictions on future behavior, including that of new users. This is often a very productive method of customer segmentation.
To implement this, take a customer group that has been successfully tested in the past, and analyze it with respect to its initial purchase. If this does not provide sufficient information for an identifying pattern, you can also add second and third purchases to your review. This usually results in a commonality, such as the purchase from a particular product category or even the repeated purchase of a particular product. So now, you can check all newer and recent customers, who have made their first purchase according to these criteria and add a corresponding flag to the respective customer data records. All flagged customers can now be sent the already successfully tested campaigns.
Of course, in addition to these three segmentation approaches there are countless others that can be applied and tested. It is worth mentioning segmentation by interest – either by directly querying customers or derived from their purchasing history. In some cases, segmentation by age or gender can also be useful. It is however, not enough, to adapt colors and images to allegedly gender-specific differences. The success of male/female segmentation lies in a completely different approach. Just think about your friends and your partner, and make yourself aware of how men and women “tick” differently in so many things. To find out how to ideally promote your product to either men or women, it may be useful to develop personas and, building on this, find brand messages and buzz words that correspond to these individual personas.
About the author: Olga Walter is a CRM specialist who works as a freelancer or consultant to a variety of companies. She is the author of the book CRM for Online Stores published last year by mitp publishers. Up until the beginning of 2015, she successfully established CRM at windeln.de. In addition, until recently, she taught CRM, email marketing and e-commerce as a private lecturer at the Baden-Württemberg Cooperative State University.