The data you need to achieve your business goal

In a 24/7 digital ecosystem, data is generated at every point. Companies are recognizing that data is an integral part of goal-setting and other business decisions, and are using data in various ways to increase the efficiency of their business. Perhaps your company wants to start collecting data to achieve your organizational goals but does not know what data to collect and how to do it? Or maybe you are just wondering if it is worth collecting data with no specific objective because you think it might be useful in the future?

Let’s emphasize one thing at the beginning – storing data “just in case” is not allowed by the GDPR regulations. However, the situation changes if you specify the purpose of their collection and processing. In that case, you can start storing them as long as you have the consent of your users or customers.

It is also worth paying attention to the quality of the data collected, so that valuable conclusions can be drawn from it in the future. It is best to collect data in an organized way (with appropriate identifiers) and to check the achievable accuracy and completeness – we do not recommend storing data where it is less than 50-70%.

In the next few paragraphs, we will examine what types of data you need to achieve exemplary business goals.

Business Objectives for Data Collection

Here are some common business goals and the process of data collection to achieve them:

Objective 1: Assess the risk of fraud

Fraud management is a major challenge in industries such as insurance or banking and finance. Acquiring a new customer in this case is inextricably linked to the process of assessing fraud risks. In order to adequately assess them, it is necessary to acquire certain types of data.

Here are some examples of data that provide the basis for successful customer risk assessment models:

  • Identifiers recorded during the application (form)
    • Name, surname, PESEL
    • Identity document – number and type (ID card, passport, residence card, etc.) possibly expiration date, date of issue
    • NIP, REGON
  • Addresses: main, correspondence, and delivery addresses (based on addresses, the data can be enriched with additional location data, e.g., on the number and demographic characteristics of its residents)
  • Contact information (email, phone)

Additionally, for the WWW channel:

  • Identifiers associated with the device from which the application is made
    • IP
    • Cookie
    • Fingerprint of the device
    • User agent string (UAString)
  • Browsertype (e.g. Tor)
  • Form input method (keyboard presses, paste from clipboard, button presses, mouse movements, etc.).  
  • Location sensor information taken from the device on which the application is submitted (e.g., accelerometer 
  • Identity confirmation method (e.g., KIR, subscriber account login, courier, etc.). 

Objective 2: To assess the risk of default

The financial or telecommunications industry faces the challenge of customers failing to pay their debts (such as phone bills or installments for hardware purchases). In order to minimize such risks and increase the ROI of the business, it is necessary to collect relevant customer data and make appropriate assessments based on it.

Some examples of data types for Telcom:

For customers with seniority <3 months

  • Customer data (demographics, place of residence)   
  • Data with economic information 
  • BIG/BIK/claims exchanges 
  • ZONK (including the history of customer inquiries from other operators)   
  • Product data (including the level of subsidization)   
  • Channel (WWW/POS) data: location of sales site 

For customers with seniority> 3 months

  • Payment history (paid invoices, completed recharges)  
  • Account structure (number of sim card sheld by market/brand), life cycle (acquisition, retention, migration), status (active/suspended/reactivated)  
  • Usage data (date, sms_in, sms_out, voice, roaming, etc.).   
  • History of reminder actions (reminders, BPW, etc…).  
  • Customer-initiated actions on the associated account (via web or app), e.g. billing checks, tariff plan changes, purchase of additional packages

Objective 3: Forecast sales and the impact of promotions on sales

Companies are building predictive models to prioritize sales and marketing strategies. They need different types of data to get a satisfactory return on investment in marketing activities (ROMI).

  • Internal data 
    • Sales – customer, product, quantity, list price, the transaction price 
    • Applicable promotions – to whom targeted, duration, parameters of the advertisement (depending on the type of promotion it may be discount, reach, cost of the promotion, etc.). 
  • External data (publicly available reliable sources of information updated regularly)
    • Market data (sales of competitors, their prices, volumes, promotions carried out)
    • Consumermood, PMI index  
    • Currency exchange rates (especially if the products/services in our category are imported) 
    • Weather data 

Objective 4: Product recommendations

Retailers and ecommerce companies want to increase cart value and user retention. They use customer behavior data to make personalized product recommendations.

Examples of customer interactions that collecting can be useful for building a recommendation system:

  • Viewing a product 
  • Buying/adding to the cart 
  • Liking a product 
  • Reviewing/evaluating the product 
  • Search parameters (if the site has a search engine for products by their characteristics) 

It is also worth describing each product with its features (e.g., with apartments, it would be the number of rooms, square footage, location, availability of a balcony, etc.) to enrich the data collected and increase the effectiveness of recommendations.

Objective 5: Targeting marketing activities

Marketers are constantly trying to optimize their operations to achieve the best results and beat the competition. They use historical data to increase their success rate.

For example, if the goal of a marketing campaign is to re-sell product X, then we need data to identify customers who responded positively to the last campaign and bought product X. Where such data is unavailable, it can be collected by implementing a pilot marketing campaign on a small scale.

Behavioral data of customers, such as:

  • Customer interactions with the product catalog (described in the point above)  
  • Customer seniority 
  • Timeliness of payment of invoices/payment history 
  • Customer’s interactions with the BOK (Customer Service Office) etc. 

Application data collected during customer registration, such as: 

  • Date of birth (age)  
  • Residential address, mailing address 

Objective 6: Process automation (RPA)

RPA – Robotic Process Automation is a type of solution that allows automated mass decision-making, usually simple but tedious. The first step is to identify the decision points in your process that consume the most time and cost. For example, in the process of analyzing a complaint request, the first step is to assess its validity or assign it to a problem category.

The easiest way to automate is to use ML techniques to build an algorithm that will mimic the actions of employees. So you can collect all decisions with their results from the last period (e.g. 12 months) to feed the ML model. In the case of a complaint, the decision could be to accept or reject it. This information will be the target variable for your model.

Next, we need:

  • The content of the complaint notification 
  • Description of the product about which the complaint  
  • Information about the date and price of its sale and other relevant warranty conditions 
  • All data on purchases and integration with the customer in the period before the filing of the complaint 

This data may often need to be transformed and enriched, such as the content of a request for analysis by NLP methods. 

Thus, the data collected is combined and analyzed using the ML model. If the prediction results are satisfactory, we can use the model to make automated decisions. 

Develop your data collection strategy with Algolytics

The examples described above are some of the most common goals set by companies in terms of data collection and use. However, they do not exhaust all possibilities. So if your company has not yet defined its goals or data collection strategy, Algolytics can help you get started. Our team specializes in developing artificial intelligence and machine learning solutions and has successfully implemented them in financial institutions as well as telecommunications, e-commerce and logistics companies such as Bonprix, DHL, Play and Alior Bank SA.

Optimize your data collection by replacing ad hoc processes with a structured approach. Algolytics can implement sound data strategies with a customized solution tailored to your business goals. With our expertise, you will achieve a successful implementation of contemporary machine learning models and applications. Contact us to discuss your needs and learn more about how Algolytics can make your business more powerful.

Pin It on Pinterest