The short answer to this is yes! You do have data for machine learning.
Using modern machine learning techniques, value can be extracted from data in all forms.
Every computer system that you use within your organization is storing data behind the scenes in a database. Enterprise Resource Planning systems (ERPs), Customer Resource Management (CRM), and software-as-a-service solutions, such as Workday or Salesforce are all storing backend data that can be accessed and used for machine learning.
One particularly valuable insight that can be extracted from various systems is a 360-customer view. This can be accomplished using fuzzy matching techniques, such as n-grams, across databases that don’t use the same unique keys (or any unique keys) to identify customers. The 360-customer view allows organizations to know every touch-point a customer has with your organization, which can drastically enhance the ability of an organization to forecast and remediate customers who are planning to churn.
Many organizations don’t think of PDFs, resumes, sales emails, or call transcripts as data but insights can be extracted from these nonconventional sources. Customer service call transcripts, for example, can be used to optimize an organization’s sales. Think of it this way: if a customer has a negative experience on the phone with customer service at 10 am, you don’t want your sales representative calling that customer to sell them on additional products/services at 11 am because that customer isn’t in the right mindset to consider purchasing other items. A different phone call, perhaps to try to rectify the poor customer service experience earlier in the day, but be more adequate, especially if this customer has been identified as a key customer by your organization. These are the types of insights/optimizations that are possible from nonconventional data with machine learning.
Another organization may use machine learning to parse resume employee PDFs so they can include characteristics about each employee in their HR system and use that information to validate pay bands at each level. This can help an organization, especially a large organization, perform the internal research necessary to ensure they are complying with equal pay laws.
Publicly Available Data
Many organizations choose to enhance their organizational data with publicly available data. For example, a fire station may want to use machine learning to optimize their staffing. A machine learning model for this specific scenario may pull in publicly available data on the forecasted weather, including the number of days since the last rainfall and the amount of rainfall, wind advisories, temperature, snowfall, and any other factors that may impact the propensity for fires, traffic accidents, or EMS responses.
Not sure if you have the right data? Contact me for a free consultation.