Our customer was facing the need to carefully predict the Operations and Maintenance (O&M) expenditures of the Wind Farms for the following years due to a rising cost of unscheduled maintenance.


Development of a system that dynamically estimates the reliability of different wind turbine generator (WTG) components throughout its useful life and predicts failure rate of the wind turbines.

The factors that significantly influence the reliability have been identified, such as WTG manufacturers, drive train configuration, power class, site and terrain class and wind characterization.



Prediction models take into account the technology diversity and geographical spread of the wind farms, allowing for the O&M budgeting of the Wind Farms associated to preventive and corrective actions (due to expected repairs and maintenance policies in use).

Related Case Studies



  With the increasing diversity of the products commercialised by the various industries, such as food retail or manufacturing industry, inventory management has taken a core role for an organization’s success. Only an efficient and accurate inventory management allows a company to reduce its amount of stock – thus minimising the necessary investment and the risk of obsolescence and/or deterioration (spoilage) – without jeopardizing the customer service level.   To better clarify the impact of inventory management in these dimensions, it is necessary to explain the diverse relations of cause-consequence in this area, using food retail as an example to illustrate them. It is also fundamental to describe the most present concepts in inventory management, namely, the time intervals used in this area (lead time and review time) and the safety stock.   Stock-out and overstocking consequences In general, inventory management tries to find the optimal quantity of stock for an efficient process. If the stock in a store diminishes, there is a greater risk of generating a stock-out, which means the store will have no stock, which leads to a loss of sales. Besides, even if a stock-out doesn’t occur, the product may have a stock level low enough so that there’s lack of visibility in the presentation space. This lack of notoriety can equally induce a loss of sales. On the other hand, if the stock in store increases, for the inverse motives, there’s potential for a sales growth. However, there’s also a greater risk of spoilage, since the higher level of stock in the store creates a higher probability of the products expiring before being bought. These products cause a financial loss to the company since they will no longer be sold.   Review and lead time in inventory management Although there are several inventory management methods, all of them have in common the definition of the moments in which stock is ordered, as well as the quantity to be ordered. Related to the moments in which the stock review and order (if one is necessary) are executed, the time interval that measures two consecutive stock reviews is called review time. After placing an order with a supplier, he takes a certain amount of time to deliver the order. This time interval is called lead time. To ensure that the store doesn’t incur in a stock-out but also doesn’t have stock excess, the ideal would be to guarantee that the quantity received in a certain order is equal to the one to be consumed in the period until the arrival of the next order. To achieve it, the store should, in the ordering moment, compare its stock level with the one corresponding to the protection period (review time + lead time), and then order the difference between the two stock levels. The stock level corresponding to the demand in the protection period is called cycle stock. As one can deduce, to know the stock level which will be consumed during a review time and a lead time, it would be necessary to know the demand in this period. Given that this quantity is not known a priori, it is necessary to use demand forecasts to better estimate this value. These forecasts are always associated with a forecast error, this is, the real demand may be lower or higher than the one predicted. If the demand is higher than the forecast, the risk of stock-out is aggravated. To preclude this hypothesis, an extra stock quantity is used in the store to reduce the probability of stock-out due to forecasting errors. This stock is called safety stock. The size of the safety stock depends, naturally, on the forecasting error associated with a certain product. The larger the error, the greater the safety stock necessary to assure a certain customer service level. Other constraints of the replenishment process may lead to higher safety stocks (ex: low supplier service level). So it can be said that the inventory management theme, on the one hand, is structured in a company’s management, since it can have a huge impact on its success potential, and, on the other hand, is covered with a vast complexity, either by the necessity of different data (lead time of upstream processes, review time, presentation stocks, …), as well as by the necessary integration with demand forecasting. Given the clear difficulty of implementing an inventory management method in a company, the big challenge in this area is to be able to improve simultaneously the three key performance indicators: stock level, spoilage and customer service level.


  Warehouses are a critical part of a seamless supply chain. The current focus on assigning value-added activities to warehouses, in order to reduce downstream labour costs, asserts that warehouses need to be designed in detail and carefully thought.   The significant amount of processes happening simultaneously and the interdependencies between them imply that it is crucial to make accurate decisions when implementing any changes. When studying a process improvement, a layout change or the introduction of a modern technology in the warehouse operation, the critical question that arises is the following: Will this change affect negatively the warehouse’s performance?   What are the benefits of having a more balanced operation? A warehouse can only operate at peak efficiency if all its processes are balanced as a whole and are communicating without any major disruptions. The existing discontinuances can either be due to suppliers that are delivering too early or too late, to a bottleneck in the warehouse operation or to the difficulty in finding something specific in the warehouse. Considering the example of an unbalanced reception, if the warehouse’s suppliers are not allocated to the best periods of time, it might lead to the need of having more operators and more buffer space in the reception area. Also, an unbalanced reception could imply that downstream processes, such as sorting or shipping, would need more operators to work at its maximum capacity. Simulating warehouse flow can help determining bottlenecks in the operation and the hidden cost of an unbalanced warehouse operation.   Where should products be to improve sorting efficiency? The current demand for more products with shorter lifecycles and for increased differentiation is a challenge to warehouse operations. This evolution leads to more investigations on the placement of the different products in the warehouse, to ensure the reduction of the amount of time spent in picking activities, the most labor-intensive warehouse operation. Simulation is a proven methodology to test different location assignment positions and to verify the impact of changes in the amount of time and resources needed in a warehouse.   Can an innovative technology or layout improve warehouse performance? A crucial part of managing a warehouse is the need to improve its performance and to make complex decisions among diverse options on how to achieve the proposed improvements. Could the change of a warehouse’s layout lead to an increase in the congestion between operators? Could the adoption of a new automation process, which would double the overall processing rate of the warehouse, be worthy, without an increase in the reception rate? Would that lead to any savings or would it just make the warehouse more unbalanced? All these different scenarios can be properly simulated in detail, evaluating the trade-off between each measure and weighing the distinct opportunities holistically.   Minimising the risk of warehouse changes Usually, investing in warehouse changes requires big investments that can either derive from new equipment, from outsourcing or moving the operation while the change is occurring, from overtime compensation to the workers, among others. In worst case scenarios, these changes can have consequences which may affect the performance of the whole supply chain and put it at risk. It is highly uneconomical and extremely risky to test different scenarios in a real warehouse, enhancing the importance of virtual reality. Simulation presents itself as a new world of dynamism towards decision making and demonstrates an endless growth potential in companies.  


  Increasing loyalty by executing customer retention and acquisition promotional campaigns has been one of the main marketing objectives for the last decades. While marketing and commercial teams work hard to deliver the best value to customers, the brand-customer relationship is becoming more dynamic and multi-layered with customers having increased competitor awareness and preferring to use proximity stores to cherry-pick shopping baskets.   Whether on retail, banking or any service companies, the strategical planning of loyalty targeting and promotional activity is becoming a very complex task which, if not built upon solid profitability evidence, quickly turns into a  resource-usage and margin-diminisher managerial problem. While the promotional strategy is typically enforced on a top-down matter, the customer response complexity and dynamicity urge a bottom-up approach that manages to ensure one-to-one ROI focused on real customer needs and basket growth opportunities. A broad analytical vision of the customer behavior history and future response predictions are the most valuable assets towards ensuring the promotional value proposition is suited to customers’ needs and ready to drive long-term profitability.   Embracing the dynamic customer’s lifecycle A typical promotional strategy is usually defined over some sort of low-dimension customer lifecycle segmentation, more often than not based on the core RFM variables (Recency, Frequency, and Monetary value). However, both customers and companies are engaging in a multitude of ways, in which different products and channels are tailored to target specific customer needs, such as convenience, large-basket destination stores and online retailing. In this era, the increasing customer price awareness along with the emerging promotional make loyalty a much more dynamic and hard to acquire concept. While RFM models have very valuable predictive power and provide a simple but effective way of segmenting customers, they fail to pick up on the necessary consumer behavior detail to fully optimize the marketing relationship. One could dive even deeper and find such different behaviors within the same customer, perhaps across different product categories. Multi-product companies typically face very different competitors across their assortment range, and customer retention and churn measurements should also include this diversity dimension. Otherwise, the company may never extract the full potential from its customers and may be losing more money than they even realize. Broad lifecycle segmentation is no longer enough to respond to the evolution of customer loyalty, as companies need to figure out the driving needs for customer shopping and understand their true potential value and, ultimately, profitability. The pursuit of one-to-one ROI Figuring out the customer level ROI is all about the understanding potential for customer improvement and the relationship between customer investment and expected probability of growth activation. Analyzing customer potential is two-fold. It involves computing detailed customer-need specific lifecycle segmentations and also mining association rules to estimate them for customers with insufficient transactional data. Customer level information is obviously extremely rich on providing information for this value. The next step is to get answers on how to acquire that value potential. Past promotional campaigns and extremely frequent A/B testing should be deployed to accurately analyze the efficiency of each promotional vehicle, communication channel and promoted products on converting customers’ potential value into loyal recurrent transactions. Developing these predictive models enables companies to avoid the most common mistakes of promotional planning: Focusing on quantity instead of quality: Having too many promotions will, at most, maintain customer’s cost perception while eroding most chances of sustainable profitability; Ignoring customer data on the promotional definition: Suppliers are the main source of promotional funding, but the value of such mass promotions are limited regarding the impact of customer growth and acquisition. When strategic guidelines are spread down from general segmentations, customers may end up getting targeted for promotions they don’t need, on channels they don’t respond to, and with a disproportionate value regarding their potential brand engagement. The only way to avoid this issue is to strategize not on the promotion and channel usage, but on how different customers should be targeted. Before deciding on what to offer to each customer, companies should strategically define: Which segments to invest and disinvest in? What is the target ROI to decide on whether customers should be contacted? What impact will the available campaigns have on customers’ shopping patterns? How much will each customer segment, if targeted, increase market share or profitability?   Then, by compiling all the type of discounts and promotions that may be included in campaigns, companies should be able to look at the expected response for each pair of customer-promotion and figure out the expected customer response in shopping behavior.   Overcoming organizational challenges Effectively planning promotional activity across the company, making sure both customer needs and supplier funding opportunities are being considered, poses three main challenges to large companies: Managing and navigation huge amounts of data; Building a robust promotional planning process that brings all stakeholders’ needs and capabilities on board; Ensuring the customer-level promotional portfolio is correctly planned instead of having promotional silos that may have cross-promotional effects and thus erode profitability.  There is no point in devising the ultimate promotional campaign for the customers if the commercial teams can’t fund it and there is also no point in engaging in non-profitable promotional activities that contribute little towards achieving the targeted customer value proposition. Overcoming these difficulties requires developing all the processes and frameworks that enable all stakeholders to work together during promotional planning stages. Having the ability to plan, control and evaluate mass and customer-specific promotional activities is the only way to ensure marketing investments have sufficient returns and to avoid them turning into a margin-consuming spiral activity. Only then can companies pursue purposeful customer marketing strategies and develop promotional proposals that leverage customer knowledge into sustainable growth.


  Everyone likes a good deal especially when they could be part of it.  In the past years, promotions have been continually growing in sales and consumer behavior influence. However, the demand variability imposed promotes stress in the whole supply chain, especially concerning the expected demand of each product – promotional forecast. Being able to accurately predict promotional demand involves many decisions where typically analytical models delivered superior results and unbiased insights. Yet, there are several challenges placed every time that deserves to be considered regarding this process.   Shrinkage vs service level Although promotions face an increasing impact on the consumer, one pivotal fact is the need to have products available on the shelf to the end customer. If too much product is placed for a promotion it could result in shrinkage or negative margins due to product natural life-cycle. Subsequently, this might induce an over-dimensional operation with higher risk and capital invested. As opposite, an out-of-stock situation will impact customer satisfaction and loyalty. For manufacturers, out-of-stock could be even more damaging and lead to a competitive disadvantage with a loss of brand equity and loyalty. As promotional impact increases the more imperative will be to have a balanced situation and by using a more rigorous forecast less stock will be needed to fulfill possible forecast-demand deviation and, consequently, decreasing risk.   Promotional factors Predicting promotions is considerably more complex than predicting non-promoted products. The number of factors influencing demand change within promotions and, for similar promotional conditions, the data available is considerably less or non-existent. Marketing campaigns, price, store display, geographic location, brand, gifts, and many other features consistently impact how clients react to promotions. Nowadays, there are several models and methodologies to tackle this problem, however, it is necessary to have the analytical expertise to parametrize and understand the black box that not always deliver the expected outcome. The probability of ending up at an overfitted situation with meaningless results is higher, but when correctly applied is possible to take advantages increasing forecast accuracy and, especially, promotional knowledge. With an advanced understanding of each feature impact, promotional plans have more information to meet the expected results and be in line with organization strategy.   Products interaction New products introduction and promotional assortment add even more complexity into this process. With an increasing product diversification, the differences with similar products are narrower and, in many cases, irrelevant for client requirements and satisfaction. This creates an intense network of product interactions where one promotion has an impact on others promoted and non-promoted products. The introduction of new products enhances the density of the problem by requiring a forecast without historical data to support the analysis. On this subject, product attributes and similar products have an important role in order to prepare and get a sustainable forecast. By looking for attributes instead of individual products it is possible to get close to future demand and products interactions - leveraging the available information. Promotional forecasting is a process ready for improvement and there are plenty of options to achieve it. With good processes and technology, organizations can continue to exploit the benefits of promotion without tarnishing it by using inaccurate forecasts. This consequently leads to better promotions plans with better information, which helps further forecast to be more accurate. What follows is a significant cost reduction and a synchronized supply chain. Besides, having a good methodology could be the core point to leverage the business and take advantage over the competition.