On a day-to-day basis, when we connect to a website, when we buy, we look for a Wi-Fi network…. and in any business activity, data is generated that is stored and accumulated “without criteria”, with the vague idea that, in the future, it will be useful.
However, it is important to be clear that the mere accumulation of data is not information. What converts the data into information are the criteria applied in its architecture (uniformity, structure, metadata…) and to generate it, it is convenient that we have clear business objectives, in order to create a database with criteria that facilitate data fusion with different sources of interest and the subsequent interpretation of the information, to better serve the aforementioned objectives. Because if we don’t contextualize it, what does a 40, 50 or a 5 mean?
Basic keys to create a database
For this, we must take into account some basic good practices to create one:
- Structure: A database must be organized and structured according to our needs. The first step to create a database is to be clear about our objectives and our business, from here we can design a data model that corresponds to our business model and that meets our needs.
- Cleaning: To be effective and useful, a database must be “clean”, that means having controlled the variables that we use and that they are in the same format.
For example, if when creating a database we allow only adults to register in it, we must put barriers from the beginning, but it may happen that we inherit old databases and that these controls were not there; For this reason, when we merge we must verify that everyone was born after a certain year and otherwise label that value as “invalid”. - Necessary information: What information do we need from our users? To create a database we must have identified those variables that are necessary for the analysis, control and monitoring of the client, variables that must be refined avoiding the occupation of unnecessary space. All this will be easier if, previously, we have defined the structure.
- Standardization: If we have different data sources, and we want to merge them, we must carry out a prior homogenization work, following a single criterion to save the same variable or types of variables. As an example, if we have different dates saved, it is best to always save them in the same format, and if we are working with clients from different origins, we must set a time, for example that of the server, and maintain it for all types of users, and when temporal analyzes are carried out, we will have to change it to know in what temporal space we are speaking.
- Indicators: Before creating a database we must ask ourselves what data we want to report and go down until we reach the best way to store it. What identifiers we need to position ourselves in an evolutionary way and how we can optimize the information so as not to have a large store of meaningless data, will be a differential feature a posteriori. Since we should not save indicators as such in our database, but these must be calculated in the period we need, otherwise we will fill our base with summations and calculations that will lower efficiency. As for example the difference between saving the date of birth and the age at the time of registration of a user.
If we follow these steps when thinking about and working with our data warehouse, we will have the smoothest path for further analysis and the creation of dashboards that will make it easier for us to monitor our business.
Stay informed of the most relevant news on our news channel
Facebook Comments
Disqus