Data Science

Thanks to Data Science you can extract information from all data coming from business management software such as ERP ( enterprise resource planning ), CRM ( customer relationship management ) or SCM (supply chain management ) and the recent entry of industry 4.0, both small and large companies generate a huge amount of data that, once analyzed, can provide great value. The competitive advantage it confers is of such magnitude, that those companies that do not adapt to the digital transformation will find very difficult to survive.

Nowadays only the 0,5% of all data generated by companies is analyzed according to a MIT survey.

Thus, through a Data Science project, a company can increase its revenues thanks to product recommendations and customization of the purchase process with customer segmentation or reduce your costs through more efficient purchasing planning thanks to demand prediction, optimized routes, personnel management or predictive maintenance.

In our  Study Cases you have some real cases that we have done with innovative companies.

What is Data Science?

More detailed, Data Science combines computational, mathematical and statistical methods with the knowledge of a sector or industry, to extract from the standard data that will allow to predict or automate processes. This enables companies to reduce costs, increase sales and, in certain cases, create new products.

Data Science makes intensive use of Artificial Intelligence to provide solutions to problems of various kinds. For example, it is possible to use classic Machine Learning algorithms to predict the demand for a specific product, or in more recent Deep Learning techniques to automate complex cognitive tasks such as image classification or autonomous vehicle driving.

Data Science Applications

Data Science is a cross solution suitable to any type of business with multiple departments.

Demand forecasting
Depending on the sector, the cost of the supply used can have a considerable weight on the final price of the product. Reducing the cost of the material without affecting the quality of the product can be achieved by buying at the right time the needed quantity. Moreover, a good demand forecast allows better management of the production chain avoiding unexpected stops due to stock breakage and better storage management by supply excess.
Recommendation system
This application is essential for e-commerce since analyzing the data of user behavior in the web and the level of product satisfaction, makes it viable to include a recommendation system with new or complementary products. In this way the user has alternatives without leaving the page and other products of potential interest.
Predictive and prospective maintenance
The usage of the machines in the production chains continuously, uninterruptedly and sometimes to the maximum of their capacity, can lead to unexpected failures generating the associated costs, repair and, in the specific case of the food sector, cleaning of them and loss of raw material. The continuous and uninterrupted usage of the machines in the production chain, as well as, at times, working to its maximum capacity, can lead to unexpected failures generating along with the costs, repairing and, in the specific case of the food sector, cleaning of them and loss of raw material. Mistral can analyze the historical data, both of sensors and of operation, of the machinery used in the production chain and relate them to the failures detected. Using Data Science techniques, algorithms can predict failures with a predetermined time in advance or, estimate the remaining lifetime in order to be able to preset technical stops that affect the dynamics of the production chain with the lowest cost possible.
Customer clustering
Thanks to customer segmentation you can group your clientele according to common characteristics and offer products that best suit their needs with personalized offers. This application is of special interest for e-commerce since everything that happens on the internet is registered you may in this be able to analyze these variables, such as web behavior, or purchase history can help you identify your most profitable customers (and take actions to improve their loyalty), and the least (to attract them again).
Shrinkage Reduction
Shrinkage in material or products generate additional cost to many enterprises. These production errors are mainly done by failures in machines that affect the quality of the product. Thanks to Data Science, you will be able to prevent that and solve it before it happens.
Prediction of product delivery delays
As customer we are very exigent with product deliveries. Having a bad experience can affect to our company's image and is more likely to lose that client. In addition, delays may come along with penalty payments if you are a supplier or over costs in customer care. By analyzing the machine status, production level and staff available, you will be able to identify possible delays and take actions in order to avoid future problems.

Data Science project process

 

Business Questions

Our customer may manifest a need, whether a specific one to their own company or a more general one common to a sector, usually with its goal in reducing costs, increasing revenues or creating new products.

Know the Business

Mistral collects as much information as possible to answer these Business Questions and identify controlling factors and bottlenecks.

Data Collection and Exploration

Once the business processes are known and the control factors are identified, we translate this information into variables and data. Then an exploratory analysis is done to value its quality.

Mathematical Model

We apply Machine Learning methodologies to develop models that compete to see which one is the best. We start with simpler models so we can offer a faster response to customer needs and he/she can identify benefits as soon as possible.

Implementation

When we have a minimum viable model, it is introduced into the company's control system to learn continuously, with a scalable implementation and capable of training itself as it has more data.

ROI Evaluation

In the last phase, the revenue on investment (ROI) is evaluated, since the best mathematical model does not have to be the one with the greatest benefits. If ROI is lower than 0 we would return to the mathematical model phase to improve it.

Once the minimum viable model is implemented, the cycle continues with the development of more complex mathematical models that will replace the one previously installed if it is considered that it can reduce costs or increase profits. In addition, as more data or new variables are obtained, the goodness of fit index and ROI will be reevaluated each time.

In this phase Mistral monitors the real effectiveness of the algorithms developed by implementing dashboards (integrated or not in the control system), visualizing interactive and real-time evolution of the most significant parameters. In the following link you have more information about the Life cycle of a Data Science Project (Spanish).

Get value from data