the data mining, for data mining in French, is also known as data mining and data mining. The technique involves using automation processes to extract actionable insights from a large volume of unorganized data. Data mining thus enables behavior to be understood, a model to be drawn from it and strategic actions to be implemented on this basis.
> Download this free kit and do in-depth market research. ” align=”middle”/>
In business, data mining allows for example to act on the reduction of costs, on the resolution of a problem, on the increase in turnover or on the optimization of a service. The origins of the technique go back a long way: as early as 1662, an English demographer analyzed mortality data in London with a view to anticipating the appearance of bubonic plague.
Today, data mining adapts to the context of big data. The traditional system of statistics and probabilities is applied on a large scale using new digital technologies: more powerful tools for an exponential volume of data.
What is Data Mining?
Data mining uses human and technological resources to process a considerable volume of data, in order to bring out patterns, trends and correlations that do not appear obvious in view of the mass of complex information available. Data mining, supervised by the data scientist as part of the data management strategy, involves 5 steps.
- Defining an objective: this first step is the role of strategic decision-makers. Example: the manager of the marketing department is considering data mining with the aim of increasing the loyalty rate of customers of the e-commerce site.
- Collect data: this is the starting point of data mining. As part of a digital strategy, data is collected via numerous points of contact: during the user journey on the website or during subscriber interactions on social networks. In the previous example: the marketing department collects data useful for segmenting customers by profile, as well as data on purchasing behavior on the site. The sales database thus records for each profile the dates of purchases, the amount of the basket as well as the references of the items ordered.
- Prepare the data: the data is organized and stored using a data warehouse type tool. In addition, at this stage, the necessary corrections are made to ensure the quality of the data. This includes identifying duplicates, and eliminating unrepresentative data. In the previous example: if the amount of data collected for the “male under 20” profile is insufficient, the profile is excluded from the analysis.
- Model using an artificial intelligence tool: the machine automatically analyzes the data available. The cross-referencing of information makes it possible to highlight “patterns”. In the previous example: the modeling shows that the “thirty-something woman” profile makes purchases of women’s clothing on Wednesday mornings, for an average basket amounting to €80.
- Deploy strategic actions: decision-makers identify strategic actions, which are then implemented by operational teams. In the example: on Wednesday morning, the marketing team sends 30-something customers a newsletter with a promotional offer valid from €80 spent on the site, in order to build their loyalty. To generate additional sales, the marketing department may decide that the promotion applies exclusively to men’s items.
Why do data mining?
Data mining is applicable in many fields. Examples:
- Data mining is mainly used in the fields of consumption analysis and customer relations. The mining of data related to consumer behavior makes it possible to optimize the commercial offer and the customer experience, to gain in efficiency in the marketing strategy and to improve the brand image of the company.
Example: Disney World provides guests with a MagicBand that tracks their journey through the park; the massive data thus collected makes it possible to improve the customer experience: if the bracelet registers low attendance for the parade, the park communicates more about the event; data mining can also make it possible here to implement strategic actions to direct visitors more favorably towards souvenir shops, in order to generate additional sales.
Another example of the success of data mining: Netflix’s advanced data mining system allows the company to offer personalized suggestions to improve the user experience as well as the company’s innovative brand image.
- In criminology, data mining consists of identifying and exploiting data related to crimes. The objective: to model the profiles and behaviors of criminals to, ultimately, facilitate the identification of perpetrators of crimes on the one hand, prevent risks on the other.
- In banking, data mining is used to score customers and classify them according to their level of risk. In this way, the credit institution is able to adapt its commercial policy in a secure manner. The bank, for example, requires additional guarantees to grant a loan to a risky client. Banking data mining is also useful for fraud detection.
- In mail-order sales: mail-order companies use data mining to identify the profile of this type of consumer, so as to focus their marketing and sales actions on this target clientele, to ultimately optimize their costs.
What are the main data mining methods?
Descriptive methods make it possible to organize, simplify and understand information from data sources. This data is indeed available, but it is drowned in the volume: data mining optimizes the data to offer a clear vision. Among the descriptive methods, also called unsupervised techniques:
- Classification, also known as clustering and segmentation: this involves creating subsets, each grouping together a packet of data similar to each other, and different from the data of the other subsets. Example: the company wants to promote a childcare product for sale to young mothers; the classification makes it possible to segment the company’s clientele according to the attributes of gender, age and family situation, in order to effectively target its marketing strategy.
- Association, also known as affinity or sequence analysis: this involves highlighting how one event leads to another, in order to deduce behavioral trends. Example: the analysis of the shopping cart shows that the man who buys a shirt buys the same shirt in another color; therefore, suggesting other colors on the shirt’s product page improves the customer experience as well as the average basket amount.
Predictive methods aim to extrapolate measured data, to anticipate target variables. Illustration: the bank collected the data of its customers at risk; when a new customer enters the database, he is considered a risky customer if he shares the same data. Predictive methods, also called supervised techniques, include regression methods, decision trees and neural networks in data mining.
Example of a predictive decision tree as part of a product development strategy:
- The company wants to improve its offer: this is the starting point of the tree from which 2 distinct main branches start.
- The first main branch represents the hypothesis of the development of a new product. The development cost is €10,000. From there, 2 secondary branches represent the gain estimates based on existing data: the first branch displays a high estimate at €100,000, the second makes a low estimate at €50,000.
- The second main branch represents the alternative hypothesis of an update of the existing product. The cost is €5,000. The 2 secondary branches of gain estimates show receipts of €70,000 and €60,000 respectively.
Using this data mining method, the company calculates its profitability: developing a new product allows it to generate €65,000 in revenue, compared to €60,000 for updating the existing product. The company makes the decision to develop a new product, to optimize its costs and profits.
For further, download this free market research kit for study your competitors and identify the profile of your potential customers.