About customer
Kostanay Minerals is an export-oriented company, with the majority of its revenue coming from the sale of products in US dollars. Fluctuations in the USD to KZT exchange rate can significantly impact the company's profits. To enhance the stability and predictability of financial outcomes, Kostanay Minerals requires the development of a predictive model for the USD to KZT exchange rate using Microsoft Azure Machine Learning.
Customer
Kostanay Minerals JSC is a mining company specializing in the extraction of chrysotile fiber. The company is the world’s leading manufacturer and exporter of chrysotile fiber, one of the most essential elements used in the global industry.
Current state
Kostanay Minerals required Exchange Rate Forecasting within predictive analytics on macroeconomic business indicators based on internal database and external sources. In the past a data analyst had to use a wide variety of tools, including Python code and other models.
Key objectives
-
To create a predictive model for the USD to KZT exchange rate, which will be used to forecast future exchange rates and incorporate this forecast into the company's reports using Power BI.
-
To train the internal data analytics team at Kostanay Minerals in the use of Microsoft Azure Machine Learning and build internal expertise on the use of AI machine learning capabilities within Power BI Premium.
Challenges and solutions
Approach
Awara IT's approach consisted in deep cooperation between the customer's specialists and partner's specialists, DIY - do it yourself approach consisted in a large number of trainings (thanks to MCI workshops and other Microsoft support), as well as joint work on the project, so that all the work was done by the customer's own hands with simultaneous consultation and assistance of Awara IT specialists, which gave the effect not only in fast and safe implementation of the solution, but also allowed the customer to understand the solution's working scheme and further independently maintain and develop it.
We suggested to build the model based on historical data of the USD to KZT exchange rate and other relevant financial data. Machine learning algorithms used to train the model on this data, enabling it to predict future exchange rates.
Technologies
Before:
-
Rigid custom solution. All Power BI capabilities have not been used.
After:
-
Solid automated solution for business forecasting incl. Azure ML regression model (tenge/USD prediction) with PowerBI integration
Technologies: Azure ML Studio, Azure Blob storage, Power BI Premium, Azure Entra ID, Azure Advisor, Azure Secure Endpoint. The final solution works better than the tools that Tableau and 1C could offer.
Work Plan:
-
Development of the mathematical forecasting model.
-
Selection of factors influencing the KZT exchange rate.
-
Evaluation of the reliability of the chosen model based on the division of historical data into training and testing datasets.
-
Publication of the completed model for regular use.
Implementation
Eventually an Azure pay as you go subscription was deployed internally, MFA security was configured, and cost management budgets were configured. Within the configured Azure subscriptions, Azure ML Studio was deployed, where workspace was deployed as well, training was provided to the customer's data analysts on the use of Html, regression model and Time series analysis.
Based on the resulting models, the best machine model was analyzed, and the best machine model was deployed as a secure endpoint that is used directly from Power BI to calculate analytical parameters for the customer's company.
Results
Project results
In the past, a data analyst had to use a wide variety of tools, including python code and other models to make predictions within a few weeks, then manually load those predictions into Power BI and compile them into analytic reports for business forecasting. Machine model calculations are loaded automatically into Power BI itself, in fact, the analyst can simply enter data into Power BI by forecast or by actual change in values and predictive analytics is calculated automatically, i.e., he sees reports in Power BI without the need to use any additional tools. This reduces the analyst's workload, allows for much faster forecast creation and more forecast variants for more accurate analysis, which ultimately helps to make better business decisions and generally increase the efficiency of the business.
Result for the customer: reduction of forecast preparation time from a week to 1 hour, reduction of analysts' workload, possibility to create large versions of the forecast due to the development of internal expertise centers.
The project data is a trade secret, but according to industry benchmarks, such solutions help to reduce IT costs: $35,000 compared to organizing your own on-premises solution for computational and analytical models.
«Through the implementation of this project, Kostanay Minerals will significantly enhance the forecasting capabilities of the organization. This will assist the company in making more informed decisions, minimizing financial risks, improving competitiveness, and enhancing the professional skills of the internal data analytics team through training in Microsoft Azure Machine Learning» - Oleg Vladimirovich Leskin, the Director of Digital Technologies and Strategic Development , Kostanay Minerals.
This case required understanding the specifics of local regulation, and that's big win for Awara IT team because now we know everything about the process and that means that we are ready to scale it to going for dozens of customers without risks.
Awara IT expertise in this Solution Area
-
Exceptional collaboration processes & demonstrated customer obsession.
-
Excellence in PowerBI implementation.
-
Deep expertise in different industries incl. manufacturing.
Over the past 12 months, Awara IT has undertaken several exciting projects worldwide, Including the construction of a solid consolidated corporate data management platform based on Microsoft Data Platform technologies & a PowerBI-based analytical system at the wholesale retailer and Microsoft Fabric implementation in large bank.
Crucially, our affiliation with the Microsoft AI Cloud Partner Program, participation in the ISV Success Program, Solutions Partner designation in Data & AI (Azure), and certified employees (i.e., Microsoft Certified: Power BI Data Analyst Associate) reinforce our capabilities.