"Data is not the solution to everything: it's not a strategy in itself!" Francois Xavier Pierrel, Chief Data Officer at JCDecaux
Following the Data Marketing and Big Data World shows, here are some key points to keep in mind if you're looking to use your data on marketing projects, but don't know where to start.
TIP #1 - GET STARTED!
"Learning on the go is possible, even as a Chief Data Officer" says Francois Xavier Pierrel, Chief Data Officer at JCDecaux. It may seem basic, but if the subject interests you or members of your team, learn about it. Read articles, meet peers who have launched projects, or start from a problem you would like to solve.
The technical aspect of data storage can seem daunting, but clearing up the subject makes it easier to get started.
TIP #2 - CHOOSE YOUR USE-CASES
As stated in the beginning of this article, data is NOT a strategy. It must serve a purpose, support decision making.
What used to be the prerogative of Business Intelligence, digital transformation and Big Data has brought it to many more people in the business. Employees have to make decisions with more data at their disposal, but at the same time they have to use it more often while involving more employees.
Marketing has seen, more than any other sectors, an exponential growth in the amount of data processed and an increasingly detailed collection of it online. Giancarlo Miluccio, CDO at L'Oréal, identifies up to 26,000 micro-conversion events tracked on a single e-commerce customer journey!
During his conference at DataMarketing, Cory Chaplin (Lead Data Manager, M6) outlined 3 levels of data to guide decision making.
ANALYTICS OR DESCRIPTIVE ANALYSIS FOR INSPIRATION
Readable, accessible self service (acmes simple and autonomous) data (SQL, DataViz, collaborative)
STATISTICAL INPUT FOR STRATEGIC DECISION MAKING
Data platform (ETL pipeline) + notebook (Jupyter, Panda, Spark...) + a data scientist working closely with the business department with great statistical expertise and good explaining skills
PREDICTIVE MACHINE LEARNING
Data scientist ML + production affinity) + ML engineer (data engineer + math skills)
There is an assessment grid for choosing which level you will need, originally provided by Cassie Korsykov, Head of Decision Intelligence at Google.
According to C.Chaplin, it is essential for the first 2 levels to have a good knowledge of SQL in order to understand how a database is structured and to facilitate queries and data visualisation. A better understanding of this structure helps to grasp the questions that the available data can return.
For those who do not have the time to learn, or for those who want to save time once they have followed a training programme, there is the askR.ai data-assistant.
TIP #3 - SET UP YOUR DATA TEAM
You will have to find a team, because the complexity of the subject will quickly require a lot of skills.
Among the recruitment criteria, consider your team's collaborative and explaining skills. Don't forget that the use-cases you have in mind will need to be explained to the business department, who will have to approve or not of the proposed solutions!
You can start with a minimum of 3 persons. Data and digital transformation is good, but a transformation grounded in reality is better. For Simon Blaquière, Director in charge of Data and Customer Care at Generali, says "it's important to create data tools while including the business and IT department in the process".
Here are the profiles of a data team you need to hire, in this specific order!
Decision-maker
Data Engineer
"Core" Analyst Experience
Analyst Statistician
AI/ML Engineer
Data Scientist 1, then 2
Data Scientist Manager
Quality Manager
Expert Researcher
TIP #4: GET TO KNOW THE MANAGEMENT DEPARTMENT
Your data team won't just be able to rely on you or the decision-maker to drive use-cases and get feedback on them. It's going to have to involve the management department. Both because they are the ones who will make the decisions, and they need to learn how to use these new tools, but also because they are the ones who know the company’s strategy.
Severine Marquay, AI Director at Orange, insists on the central role of operational staff in the training of machines. They will be able to offer the necessary inputs for ML, but also determine the final consistency of the data results put forward!
In short, setting up a data team involves simple steps. We hope that the feedback from these professionals will help you fine tune your strategy.
