Customer Persona Building with Telco Network Data
Introduction
Knowing your customers (KYC) is very critical to all the businesses today, may it be a conglomerate which operates across continents or may it be a coffee shop at the corner of your home street. It’s vital to find out the actual personas of the clienteles and they might not be what you assume they are!
Personas help brands to differentiate their products across customer segments and to provide a product or service for a group of genuine buyers. It strengthens the long-run relationship of the existing customer to the offerings and attracts leads that are most likely to convert in the near future.
Carefully crafted personas can provide significant traits of the consumers and offer very specific insights into individuals and their problems & needs.
The most important thing to build personas is DATA because it should not be stereotypes of people or guesswork and it should certainly represent the real people from real data!
Let’s have a look at the personas that can be crafted using the Data of a Telecommunication service provider.
Generally what type of data a telecommunication service provider maintains?
- Customer data (CRM)
2. Call/SMS/Data records
3. Network data
4. Transaction (payments & reload) data
5. Service /package add-on history data
6. Device data
Using these data, it’s possible to identify variant personas that help to analyze to increase profitability by optimizing network usage & services and enhancing customer experience.
Customer Segmentation based on Usage Trends & Consumption Patterns
One thing to keep in mind is the personas keep constantly growing according to the needs and challenges and require constant updates. Let’s have a look at those personas that can be derived based on Telecommunication data.
1. Demographics
Age / Gender / Marital status / Parents / Income level / Designation / Occupation /Educational background / Spending habits (budget-premium-luxury)
2. Geographics
Residing location / State / Region/ Province / Town / City / Village/ GPS coordinates / Vacation destinations
3. Work and Office
Office location / Working hours / Commuting time / Commuting routes / Office Community
4. Preference on Communication
Language preference /Communication channels / Communication timings
5. Digital activity platforms
Blogs that read /Online Shopping platforms / Social Media Platforms /Video- Movie Platforms / Chat — Messaging platforms / Music Online access / Digital adoption / Professional carrier related platforms (LinkedIn etc.)
6. Offline hangout platforms
Meeting friends / Restaurants / Coffee Shops / Vacations / Beauty salons /Supermarkets /fashion outlets
7. Type of Influencers they make — which segment they can influence / social media networks /social networks (roles they play as gatekeepers /followers) /what subject they post and on what they influence others
It’s crucial to identify the telecommunication data sources that can be utilized to derive the personas. The following chart should help for mapping the source with each persona category.
Journey Analytics using Subsegments
Personas have a key role to play in the increasingly important discipline of customer journey mapping — it is the story of a customer’s experience.
It is possible to derive different subsegments based on customer personas and map those in different phases of the customer life cycle with the product/service. This essentially helps to easily segment users in a journey, creating different paths to match their activities and enhance the value derived from the customer with the product journeys.
It can be illustrated as follows.
Applications of Personas
There are few interesting applications of personas in the advance analytics domain.
1. Identify unique clusters in the potential target base.
This helps to narrow the target segment and increase the responsiveness of the targeted base leaving more room to achieve the objective set at targeting a specific segment with relatively lower marketing costs. Further, it is possible to create different ML models for each identified unique cluster-based on prominent persona characteristics.
2. Use the derived new personas as a direct input attribute for machine learning models to achieve higher model performance metrics.
This way the feature engineering and prioritization can occur in more elevated way as the identity and behavioral personas helps to add value through new feature creation.
3. Vital for end targeting and campaign planning with customized offerings & rewards
In any business today, modern consumers expect the ability to customize anything and everything to fit their individual preferences for which the derived personas come in handy. The personalized offers increase the user relevancy for the product and thereby increasing the conversion rate. In the long run, this can build loyalty toward the brands. Here the business gets the opportunity of knowing the customer better to push services that truly will be enjoyed by the customer.
How to Construct Personas
Step 1: Defining the Persona (Scoping)
The different persona names can mean different things for different people.
For example, if we want to derive and label a ‘business user’ and a ‘retail user’ for a using telecommunication data set, it’s essential to list down the characteristics of both users and understand what type of user with what characteristics your business needs to identify to generate value.
For example, when we consider telecommunication data sets, a user who keeps a separate SIM from his personal SIM and talks on entirely two different social network groups, can indicate a sense of business SIM and a personal SIM. When people have multi-sims it is important for your business to target the personal SIM for value-added data services and data bundle offers as the user might not have total discretion to bring add-on facilities to his business SIM.
Step 2 : Brainstorming / Defining Logics & Boundaries
The logic and boundaries for a selected person can be defined based on available data sources and here it can be creative to logically match and derive customer behavioral patterns that relate to the persona you are looking for. In order to increase the richness of data, you can think of ways of gathering data from external sources and conducting research on your own customers in the format of personal interviews, informal discussions, surveys and social media platforms.
Defining precise parameters which characterize each segment can be a good way to start with. But this might be after an exploratory analysis of relevant data sources.
For example to the same persona of deriving a ‘business user’, it’s possible to think of the underlying logic, as in;
>What are the popular business user apps? such as in Teams in Mobile, Outlook, Authentication Apps
>What kind of Avg Data usage can be expected? Is it safe to use the logic of higher than the average in daytime consumption?
>Can it be a mix of Wi-Fi and direct Mobile usage?
>Daily routine commuter? Can this be traced from Locality info?
>What if the real user is working from Home?
Another decision that require criteria to consider would be the ‘the time window’. This is based on what level of event repetitiveness is needed to label a person for a certain behavior without a reasonable doubt. In other words, it’s the minimum threshold for recursive behavior. Say that you want to capture a high level of credit card users as one of the personas. So based on the distribution of credit card transaction records, a logical threshold to categorize a user as a frequent user must be devised.
Step 3: Hypothesis Testing
As highlighted in the start of this article it’s important not to derive who you think a user should be but focus on the actual behaviors of users from real data.
For example, a hypothesis that needs to be validated might be; In the case of Dual Sim usage, can one SIM always be a Business SIM?
Step 4: Constructing personas
At this stage, you will require all your tech resources of data engineers and data scientists to apply the logical arguments correctly and label the users according to the persona you are looking for.
Step 5: Validation and Buy-In
This involves sample sense checks and may be contacting a couple of real customers to validate the accuracy of logic and labels and to get the business buy-in.
The retention period & expiry of behavioral personas also needs to be agreed upon at this stage. This mainly addresses the question of; from which time intervals is it required to check the same behavior of a person to ensure previously identified behavior is still in existence.
Step 6: Process automation
It’s possible to bring the validated logic to data pipelines and schedule everything for automatically labeling the users according to the personas moving forward.
Step 7: Application
This is where the business realizes the value of this entire process and as mentioned above in the application section, the personas can be made use of, for the business.
Targeting Customer Enrichment Levels
There are two ways of focusing on discovering new personas.
One is targeting to uncover more knowledge of existing customers that you already know of certain personas. This way the enrichment level of existing known customers is enhanced. Say that we have found a particular women group in the network who are very keen on ordering food from a popular food chain. We can try to discover more personas on this same group of people such as their banking partner preferences by interviewing or surveying.
The other one is focusing on discovering different personas in a variety of groups enabling coverage across many people as possible.
Therefore, once you know one persona of a customer you can explore and discover more personas of that same customer. So, when you get more time series data with different varieties and velocities, there’s more opportunity to discover the real person who is as close as to the actual behaviour. Especially with unstructured data sources (such as comments on social media platforms and other online platforms) and big data streams the opportunity of achieving high enrichment levels are far more than you can imagine. This totally depends on the number of gathered personas, the accuracy and how close the logic can map the actual situations and the business priority.
Summary
Amidst of heavy competition in the marketplace, building buyer personas is a vital business decision today. This can be useful to craft the right product, enhance service quality and for result-oriented low-cost marketing efforts. Understanding who your customers are imperative to help a business maximize the advertising return on investmentand content effectiveness. There are many different behavioral and identity characteristics that can acquire from real data and decent applications to use for businesses. Constructing personas requires attention and the business to adhere to a good process if the business plans to take critical decisions and invest with credibility in the marketing campaigns based on those.
Written By Nadeesha Ekanayake, Senior Data Scientist.
