The term “data” is often described as the gold of the 21st century or sometimes the “new oil” though with the collapse of the price of oil, this name has fallen away. But isn’t this statement reasonable? What data do you need for machine learning or AI projects?
What data do you need for AI Projects: Quality over Quantity
The fact is, we can make data say what we want, but how we interpret data is the key factor. However, any correct data interpretation must rely on quality data collection. This means not only having the right data collection and storage tools but it also means have strong internal data governance and collection procedures.
What type of Data do you need for Machine learning projects?
Beyond this obvious requirement, we must distinguish between two main categories of data to collect for an AI project, administrative data and usage data. The former is necessary for the company to manage clients. These are, for example, the customer’s name, address, telephone number, credit card number, etc. But this doesn’t say much about him. Whether a prospect has a “Gmail” or a “Yahoo” email account doesn’t reveal much.
On the other hand, collecting a client’s usage data – linked to the core business of the company, if possible, is strategic. If the company sells cars, knowing that a client has an ecological approach to his driving, that he drives 20,000 miles/year, that he owns a detached house with a garage (therefore with an electrical outlet) and that he has two toddlers – this is all worth gold! Armed with this usage data, a company can develop a truly personalized marketing approach that significantly improves how relevant its customer-directed messages are, and therefore the conversion rate to purchases as well.
How is usage data collected? Among other ways, with “cobots” (collaborative robots) which we talk about in great detail, covering their use and functioning, in our book, “The Rise of the Cobot“, that is available on Amazon or at our management committee conferences.
