What does it mean to be data-driven?
Many companies claim to be data-driven. In this article, we will dig into what this concept means and see what steps are needed to achieve it.
Data driven : Definition
There are several trends that define what it means to be data-driven. But what these trends have in common is that we will talk about the maturity of the company in terms of data. A data-driven company is therefore a company that has put data at all levels and will make all its decisions using data.
We will define data maturity at several levels. A company at level 1 will be data-driven at the basic level while a company at level 5 will be 100% data-driven. Here are the 5 levels of data maturity that a company can have. There is a hierarchy between these levels:
- Reporting: Ability to visualize data
- Analyze: Ability to go deeper into the data to understand why a phenomenon occurs
- Optimize: Optimize business processes by providing knowledge through data
- Empower: Empowering employees by providing the tools and knowledge to perform analysis
- Innovate: Use data to innovate products and transform the organization
Reporting is the first step in a data-driven enterprise. This step will lay the foundation for the future of a data-driven enterprise.
At this point, it is clear that we want to build a data-driven organization: Excel and Google sheets must disappear. The proliferation of spreadsheets creates more confusion than anything else. The first step is to implement reporting tools.
We will therefore have to set up the first data retrieval in a more process-oriented way. We will therefore choose what data we collect, how we collect it and where we store it. The goal will be to build dashboards, called by all, to identify the first trends.
To create dashboards, we need to determine the main metrics we need, called KPIs. These KPIs must be relevant and have value. This is what will drive management to invest in data.
Start instilling a data culture
At this stage, if the decisions taken are largely based on intuition, some interlocutors will not be very sensitive to the figures given in the reports. Some will even be skeptical of the metrics presented and afraid to move to an organization where rationality prevails where previously it was the person with the highest salary or the highest hierarchical level.
It is also important at this point to clearly define the questions you want to answer. It's a task that seems simple, but in fact requires spending a lot of time: in general, in data, it's important to define the problem well so that you can choose the best tools to answer it. And at this stage of maturity, this will be especially useful to define the right metrics.
Once we have built the first dashboards, we will want to go further in the analysis. We will probably need to collect more data to make more relevant analyses.
At this point, the data collected starts to come from different sources. The question of data quality will start to arise. Therefore, you have to start implementing tools to monitor the quality of the data.
A "professionalization" of data collection
The difficulty brought by having new sources means that data collection will now have to be more processed. The data infrastructure will be better defined. The way to integrate data, to store it will be standardized. In addition, data scientists will have to look for more tools to improve their analysis.
At the same time, dashboards are becoming more and more sophisticated.
Start to spread a data culture in the company
At this point, some employees who are not part of the data team - the employees most interested in data - can start learning basic data skills. Management is beginning to be convinced that a data-driven culture needs to be instilled throughout the company but does not yet have all the tools, knowledge and methods to do so.
The success in the reporting part must serve as a driving force to convince the rest of the organization, especially those who do not have this culture and for whom "it has always worked like that" and who do not want to change their way of working.
Where reporting only focuses on the past, we use analytics to begin to understand the past and derive actions to predict how the future will unfold.
At this stage, we will start to optimize business processes with the help of data.
Collecting data again and again
It's going to be important to continue collecting data at this point. We're starting to get to the point where we can help personalize the experience. So what's going to be important is to collect behavior and purchase data to be able to offer this personalization.
Moreover, to choose or not to collect new data, we must now systematically ask ourselves the question of its use. We will no longer try to only bring knowledge with these data, but rather try to deduce direct actions from these data.
An evolution of skills
Where before there was only a need for data engineering and data mining - to retrieve data and analyze it - the new needs require new skills, especially in machine learning. With this, we will be able to use past data to make models to better predict the future. We are moving fromanalysis to prediction. The people who have the analysis skills and those who have the prediction skills are generally not the same. You also have to be technically prepared to be able to make these predictions.
We will also try to get closer and closer to real time.
Get the data out of the data team
At this point, the question arises as to how to ensure that data is no longer a matter for the data team alone, but that it concerns everyone's daily life. Data must then become an integral component of the company's strategy.
Now that we know how to visualize data, how to use it to make analyses and predictions, and that we have convinced management and some employees that data should be at the heart of the company's processes and decisions, we might want to say that this is it, the company is totally data-driven. But no, it is actually at this point that we need to accelerate. Starting by collecting more and more data from external sources and analyzing it, using it to improve predictive models.
The question now is: how do we get everyone in the company to use the data? Several elements:
- Technically, be able to easily open up the data so that it can be better disseminated in the company
- Train all employees to use basic data analysis tools, so that they can do what we call self-service business intelligence. Basic questions should no longer come to the data team, people must have the tools to answer them themselves.
- Check that the tools produced by the data team can be used by everyone in the company. And then, check that everyone is actually using them
- Find the right skills to align people with business jobs and people with technical jobs.
At this point, everyone in the company should be making most of their decisions based on data. Most people should now be optimistic about using it in their daily lives.
It is when we are at this stage that a company is totally data-driven. Having optimized everything in the company gives it a competitive advantage over its competitors. Data has become the company's most important asset, so much so that in some cases, the company's business model can be transformed around data.
If we take the example of a Google, the data will be essential to address the most relevant ads to users. Google will therefore be able to sell its advertising at a higher price because it will bring more revenue to those who pay for it.
At this stage, everything goes through data, we will use this data to innovate in the right way
New data sources
At this stage, we are still collecting new data sources. This can be for example unstructured data: image, sound, videos, ...
New skills will have to emerge. To innovate, we need people with an entrepreneurial mindset. To innovate, we need to be at the level of, or even ahead of, academic research. It will therefore be relevant to recruit researchers with a PhD in Data Science or Computer Science.
To go into new business, data will be at the heart of the decision. Buying a competitor, going into a new market? Data will allow us to answer this question.
Automate part of the data team
Now that a large part of the company has been automated by the data team - so that everyone works on their core business - why not automate the data team? The point is to spend as little time as possible on the repetitive and interesting tasks and focus on the most important task. Remember, we talked about this at the beginning: define correctly the problems you want to solve by asking the right questions.