Customer segmentation can be a simple or a complex process. It may be undertaken manually or using predictive algorithms, and can come in many shapes and forms. The assumption underlying all segmentation modeling, however, is that within your customer database you will have groups of customers that share similar attributes, behaviors and buying habits. If you can understand how individuals within a particular segment are interacting with your products and services, then you can start to fine-tune your marketing to ensure that you can deliver what they’re looking for, and therefore achieve better engagement.
Traditionally, once you have mined your data for those common attributes and grouped similar customers together, you will define each segment with a name that represents it. When organizations do not have enough data points to highlight any discernible differences between individuals, they may turn to 3rd party data enhancements to append additional data to the customer records, or else the brand may survey the database to collect those additional data points.
For example, if you poll your customers to understand their favorite food from a selection of answers, you may be able to link ‘Pizza lovers’ to a particular gender, age-group or location, as well as certain buying attributes. It’s odd to think that food preference could determine how someone might buy, but then, at a Big Data conference BlueVenn attended seven years ago, a relatively unknown firm called Cambridge Analytica asserted that they could psychometrically profile and segment Facebook users to 92% accuracy, based on the types of photos they liked.
We now know, of course, that their segmentation methods were used for political, as well as retail purposes, and none too ethically, but nonetheless Cambridge Analytica has demonstrated how powerful segmentation can be when optimally deployed.
So, the key to effective and accurate segmentation is to:
- Have lots of accurate and complete data (the more and cleaner the better).
- Invest in fostering the skills, technologies and knowledge base needed to work with data and segmentation strategies.
Provided you have a reliable Single Customer View, composed of clean data and containing the details of all information held on the customer within the business’s data repositories, then you’re in good shape. According to our ‘Seven-Stage Customer Data Maturity Model’, the first stage is to amass data and create a Single Customer View, while the stage two involves deploying analytics to gain insights, which will help you to determine your optimal customer segmentation model(s). Any segmentation strategy will yield poor results if these first two fundamental stages are skipped. (Download your own copy of ‘The Seven-Stage Customer Data Maturity Model’ here.)
What types of data can you use for customer segmentation?
Demographic data is information collected about common characteristics of the general population, and within your database, basic segmentation of customers with similar demographic attributes could help you to determine how likely a certain group is to buy a specific product or service. Demographic data will help you to identify key market segments and understand your ICP (Ideal Customer Profile). This data might fall into categories like:
That sort of information is commonly collected from customers, or can be sourced from third party data pools such as visa application forms, national census records, or passport applications. Demographic groups are easily determined and a form of segmentation in themselves, so they commonly form the basis for more advanced segmentation models. However, the relevance of this type of data will depend on the product on offer. For example, a fizzy beverage could be of interest to any age, while an alcoholic one will certainly only be targeted at those legally of age.
This type of data is less factual and more to do with a person’s feelings, motivations and choices, but as marketers we have tons of it at our disposal, which is lucky because behaviors have infinite possibilities. A long history of behavioral data is therefore needed to generate meaningful insights. Behavioral data might include things like:
- Whether a customer has previously bought your product.
- How they interact with your channels.
- How regularly they buy from you.
- At what intervals they typically buy from you.
- Whether they have displayed loyalty towards your brand.
This information will help you to categorize customers so that they get an offer message when they’re likely to buy again, show them ads for products that others who show the same behaviors commonly buy, or send basic product information to new customers only, not returning ones.
In its most basic form, this is to do with where people live. However, it can be used to segment audiences according to the terrain and geographical makeup of their local area, not to mention the climate. Thus, you can not only send a customer coupons valid in their local store, but also target them with offers for products to keep them warm and dry when it’s wintry or rainy outside, promote beach stuff in summer if they live near the sea or hiking gear if they live near the mountains, or promote gym gear if they are city dwellers.
A particular use case for geographic segmentation that BlueVenn has been seeing a lot recently with retail customers is to identify lockdown areas, where stores must close due to COVID-19, then use this information to provide relevant offers with online-exclusive deals or promote the Click & Collect service.
This type of data used to be more important in B2B, as it showed the technology stack held by businesses. Now, of course, it reveals which devices are used by which customers. It will help you to categorize your customers in the most appropriate way – for example, to develop a business app for the platform most customers are on, ensure messages appear in the most appropriate format for popular email clients, and perfect payment processes for commonly used devices. It can also be valuable to companies that sell complementary products, for example phone cases, laptop chargers or printer repair services.
Psychographic data is a window into your customers’ passions, opinions and priorities, and is typically collected to determine a person’s emotional connection with your product. It’s possibly the most powerful customer-centric form of data when it comes to hitting the right tone with a customer, but it is also the most complex information to gather and interpret. It is typically uncovered through surveys, but it can be costly or time consuming to obtain information about a person’s:
- Values, e.g. community pride.
- Attitudes, e.g. political leanings, devotion to causes such as the environment.
- Interests, e.g. favorite books/movies/sports.
- Activities, e.g. specialty cooking, playing an instrument.
These are things worth knowing, because they are invaluable when it comes to message customization and the conveyance of brand values. For example, knowing the environment is important to a person will help a travel company to pitch the right sort of holiday with the right emphasis, and perhaps alter its practices (for example by becoming carbon neutral) if customer opinion so dictates.
However, segmentation doesn’t depend solely on single data types. Introducing segmentation models into the mix can help you to achieve even more granular segmentation, leading to even finer targeting of suitable customers.
Common segmentation models and methods
RFV (recency, frequency, value) segmentation
RFV segmentation combines behavioral and financial factors, such as how recently a customer has bought a product and how regularly they typically buy, with hard facts regarding how much was spent. Knowing all of this will help a brand to:
- Understand which of their customers are most loyal and reward them accordingly.
- Reactivate lapsed customers, by first identifying and then incentivizing them to purchase.
- Identify new customers and welcome them appropriately, perhaps with free delivery or money off.
- Encourage upsell and cross-sell by showing those who regularly buy low-priced products ads for premium or complementary products.
Customer value segmentation
This may be based partly upon, or lead to a program of, RFV segmentation. A customer’s value is strictly determined by their cumulative spend, so a LTV (Life Time Value) scale can be used to determine eligibility for offers, loyalty rewards, promotions and other special campaigns. Ultimately, segmenting your customers into high and low value brackets can help to identify how high-value customers are commonly finding you and how you can skew cheaper acquisition strategies towards the lower value customers.
Customer status segmentation
Customer status segmentation looks at whether customers are active or have lapsed, with reference to the RFV model and whether expected purchases have happened in the expected timeline or not. For example, if insurance is renewed annually and the renewal date passes without a purchase, or if a person typically buys face cream roughly every three months but hasn’t done so for six months, it’s time to look at wooing them back.
This classification stands on the shoulders of demographic data to start siphoning customers into segments according to likely life stage or embedded ideals (with reference to supporting data such as products bought, spending patterns etc.). These can be used to target customers with messages on their level about products they might enjoy.
Life stage segmentation
This is obviously an overlapping segment, but focusing less on values and more on what their requirements might be, based on what is happening in their life. For example, if someone has bought a lot of baby clothes suddenly, it makes sense to start pushing products like car seats or mark them for targeting with toddler clothes in a couple of years’ time. Or, if they suddenly buy a bed, a couch and a television table, to start targeting them with homeware messages.
This type of segmentation is especially important to companies that market internationally, as it pays to be aware of what the season will be in a customers’ location, to avoid targeting Brits with beach products and Australians with woolen wear in December. Holidays such as Christmas also fall within this bracket, enabling shopping sprees to be anticipated and encouraged.
AI/ML driven segmentation
AI and Machine Learning aren’t segments, but enable vast datasets to be evaluated by looking at historical data linked to your customers. They then do the heavy lifting to find common attributes and build segments in record time, while helping to make them more granular and impactful than ever. This can be done by identifying lookalike customers to swell segments, or new ones to help market products to finely defined brackets of persona, and personalizing communications at scale to improve engagement. It may then be possible to use Next Best Action, or other predictive analytics methods, to provide optimal customer experiences.
To find out about the segments and models mentioned in greater depth, watch our “Customer segmentation: key models every successful marketer is using” webinar today.