What Is User Modeling and How to Use It to Sell More Products2020-05-25T18:23:27.000Z 2020-05-25T18:23:27.000Z How can you use modern technologies and user modeling to sell more products and become even more relevant to your customers?
Every business needs customers to thrive regardless it’s a B2C or B2B business model. Defining the business’ target audience is one of the first steps towards finding customers - you need to know who you would like to reach with your products or services and how your business needs to be adapted for the market. User modeling is one of the techniques how to do that and in this article, we’ll delve into details such as:
- The definition of user modeling & its importance
- How user modeling works
- User modeling and machine learning technology
- Examples of modeling in action
Any kind of industry can benefit from understanding what makes their clients tick and what challenges they face, so let’s learn how to use the innovative technologies for our benefit.
What is User Modeling?
User Modeling is a more technical approach to defining one’s target audience and adapting the systems to the audience’s specific needs. It is the subdivision of human-computer interaction that is used to create an understanding of a user.
Why Is User Modeling Important?
Why would one need to adapt the entire system to the users when we can make them as we like? Well, you need to deliver relevant content to your potential customers so they would become your actual customers.
Data Sources & Methods of User Modeling
In order to create a good user model for your business, you need to define the information sources that you would require. Start with asking the questions of what goals you are trying to achieve and what kind of information might be useful for reaching this goal.
There are many user modeling approaches, but here are three that are the most relevant information sources (subjective opinion of the article author):
- Relevance feedback. You can find out what people are looking for, and you can also find out how they are responding to the queries, the feedback. Therefore, you can adapt your system in such a way that the best results (most relevant to your users) will be ranked on top as a result.
- Mobile environment user modeling. With the way things are changing and going more mobile, you need to adapt the system for the mobile environment to see if the user behavior changes under different circumstances. The results of this research help you to build a better mobile user interface and therefore simplify and streamline your user’s journey.
- Demographic user modeling. This is one of the most traditional ways to create a portrait of a potential user. Demographics research statistics relate to education, religion, age, gender, etc. While you cannot discriminate people according to various things, but at the same time, different people groups respond differently to different stimuli. For example, a young girl would respond to an ad differently than a 60-year-old lady. Or, people of the active dating age (18 - 35 years old) tend to have broader social connections rather than people after 35 years old, who prefer small, closed (and, often, same-gender) social circles. Knowing these things, you can subtly tailor the user experience to your visitors.
What do you do with the information that you’ve got? That’s the next decision that you have to make and there are several approaches for this as well.
- Static User Modeling. As you can guess from the name, once you’ve gathered the information, it remains static and unmodified. These types of data sets are useful for services that don’t need customization on-the-fly. For example, it can be a blog platform that does not need constant adjustments.
- Dynamic User Modeling. In this approach, the information is gradually updated. The data can either be updated fully or some parts of it that are more subject to change. Recommendation engines can use this type of user modeling approach.
- Highly-adaptive User Modeling. Some businesses require extreme customization on the fly in order to show the most relevant and precise data to your customers. For example, think of your experience of shopping on Amazon or AliExpress. Once you’ve visited a product page, you’re then seeing the products that are related to what you’ve seen.
- Stereotypical User Modeling. This type of user modeling is created from the generalized version of static profiles, i.e. you make assumptions about the users based on larger chunks that are united by common characteristics. It’s the opposite of the highly-adaptive approach. Stereotypes are useful when you don’t need or don’t have access to personal information, for example when you need to comply with General Data Protection Regulation standards.
As a result, you get your user profile modeling that helps to make the content of your product or service relatable (and therefore more converting).
In order to simplify the data gathering process (especially if you have a lot of incoming information), machine learning algorithms can be of help. User modeling uses these methods of ML:
- Supervised classification: the algorithm uses a labeled dataset for training to be able to classify the new incoming samples. It also can detect anomalies in the data (for example, when instead of a phone number - numerical data - there are words, as well as more complicated cases)
- Supervised regression: the algorithm looks for relationships and values between different elements and this is used as the foundation for the model.
- Unsupervised clustering: the advantage of this algorithm is that it is trained on the go by looking for patterns and clustering similar information.
- Random forest method: multiple decision trees help to assign different data to relevant segments.
Read more about machine learning here.
How Does User Modeling Work?
The way the user models are created is quite straightforward:
- The information comes from either from the user herself (for example, when she registers for a user account or signs up for a newsletter, etc.) or from monitoring the user behavior on the website or inside the app.
- Once gathered, the data is ready to be processed. Technically, you can do it manually or with basic instruments, but if there is a large amount of data, it’s much easier and cost-efficient to set up proper machine learning algorithms that will study the raw data and gather insights from it. What kind of info can you gather?
- Personal information (name, gender, age, job title)
- Contact details (physical and electronic address, phone number)
- Acquisition channel (where did your user come from - referrals, social media, advertising, etc.)
- User behavior (what type of content does this user prefer, how often s/he shares information on social media, what content increases the session time on the website/app?)
- The lifetime value of a specific user, etc.
All of this information is compiled to create a comprehensive user profile.
IMPORTANT: Due to the General Data Protection Regulation (GDPR), be extremely careful with collecting, processing, and storing personal & contact information. Our advice is to limit its collection unless it is absolutely necessary for your operations. If you do need it, then make sure you have a policy regarding this data as well as proper data protection processes in place.
4 Real Examples of User Modeling
Social Networks & Search Engines
User models are created in order to adapt the content the person sees, for example, on her / his Facebook wall. Sometimes it can even feel a bit creepy when you feel like you’ve JUST talked about something and the next minute you see an ad for this very good or service on your wall. While this might be a bit unsettling, most of the time it’s just what is called the Baader-Meinhof Phenomenon.
However, modern social networks learn from each given user’s behavior to tailor the information to your personal interests and make the social network more attractive to you as a result. For example, if you’ve scrolled through an ad for renting a house in the woods, and then came back to it to check out the information for more details, this would tell the machine learning algorithms that for some reason, a house in the woods was interesting to you. Sometimes it also becomes too proactive, for example, when I was preparing for a friend’s baby shower and shopped for diapers. For weeks after that, I was bombarded with information about diapers - types, discounts on them, and what ecological alternatives can there be to them.
Who uses user modeling for their operations?
- Amazon shows relevant products to you based on the items you’ve checked out in the past. It also studies the correlation between goods and shows you “You might also like” or “People have also purchased this...”
- Google also tailors the search results to your previous behavior (if you’re logged in) or basic information about you that it can get (if, for example, you’re using Incognito mode in Google Chrome.)
- Netflix uses behavioral data to offer you the videos that you might like to watch based on your previous choices.
Product Management & Improvement
Before you launch your product, you usually have an understanding of who your target audience should be. Once the product or service is launched, it’s important to continue learning about your audience to make sure they continue to like what you’re offering.
How can that be done? User modeling comes to help here as well. By learning how your current clients use the product or service, you can add more details to your initial target audience assumptions and tweak the features (or add new) that would make your offer stand apart from the competition.
What questions can you ask?
- Who uses my product/service, why, and when?
- What additional problems of the user does my product/service solve?
- What are the unexpected flaws that arose during active usage?
As a result, you get a view of a bigger picture that helps you to create strategic and tactic plans for your product/service improvement that would increase customer retention. It also can help you understand the market even better and detect interesting insights about the people who are your customers.
Digital Marketing & AdTech
User modeling to digital marketing is like water to fish. In order to spend your budget efficiently and reach good results, you need to understand the people you’re targeting with the ads and other marketing efforts and activities.
In order to create powerful marketing campaigns that would resonate with the people, learn from the data available to you. The resulting user models are like an infinite loop:
- They help you to reach the audience initially
- Then they help to learn more information about the user profiles and improve them for the future.
E-commerce & Retail
Most people love shopping. Even those people who say they don’t like shopping, usually have something they can shop for hours. For example, I don’t like shopping for clothes, but I can spend quite a lot of time at bookstores or sauce aisles at a grocery store.
How can someone enhance my shopping experience? One of the ways would be to see what I like the most and offer me more of it. For example, if I like books and the online bookstore’s system have noticed that I do my purchases most of the time when there is a discount available or some kind of another special offer, one of the ways to attract my attention would be to offer a discount on a book that I was researching before. (Amazon does this to me continuously and I should know better by now, but I still get caught on this hook.)
The purpose of the user models in retail and ecommerce especially is to maximize the desire to make a purchase. Observed behavior patterns for chosen user segments help in this case. Technically, you don’t even need any personal information about the user in order to target them with information about new products or services. When you know what people like, chances are you have what they want.
What do you do about it?
User modeling is a great way to maximize the results of your business in the most efficient way. As you can see, this technology can be applied in pretty much any industry, and machine learning algorithms help to process, analyze, and store large amounts of data, freeing you to glean insights and implement them into your business.