Adamed Pharma looks to the future with advanced analytics and data visualisation

Ensuring access to therapy for millions of patients

Adamed Pharma is a Polish research and patent-based pharmaceutical and biotechnology company with 100% Polish capital. Founded in 1986, it currently employs almost 2,700 people and has 2 production plants in Poland and one in Vietnam.

The pillars of the company's development are foreign expansion and investment in increasing the production of medicines in Poland and in innovation. For more than 20 years, the company has been carrying out its own innovative R&D, for which it has allocated USD 520 million since 2001. Adamed is currently implementing projects in three therapeutic areas: oncology, diabetology and neuropsychiatry.

The company's intellectual property is protected by over 200 patents in most countries of the world and its portfolio includes over 500 products. Adamed Pharma produces more than three billion tablets annually, which are sold in more than 70 markets worldwide. It ensures universal access to therapy for millions of patients in Poland and many other countries.

The challenge of accurate demand prediction

The ability to effectively forecast demand across multiple product lines is a high priority for all large manufacturers such as Adamed Pharma. This is a significant and complex challenge due to the various factors that influence the demand for pharmaceutical products.

Marketing spend, competitor activity, seasonality, and individual product and brand attributes all need to be taken into account. For example, seasonality plays a crucial role in forecasting the demand for flu medicines, but is much less important for indigestion medication.

Each forecast is key to effectively planning production schedules, calculating lead times and organising distribution.

Historically, Adamed Pharma has relied on subject matter experts to produce its 18-month forecasts for its various product lines. However, this approach presented a number of challenges. The human expertise and insight required was labour-intensive and time-consuming, especially given the large number of unique products that needed to be considered.

In addition, the process lacked transparency, as it was not immediately clear how the data was feeding into forecasting decisions, which was a major cause for concern. Furthermore, if a subject matter expert left the company, their knowledge would also be lost, creating further difficulties. Taken together, these issues resulted in an inefficient process with limited data visibility and a significant risk of forecasting errors.

The road to unveiling actionable insights

With the aim of modernising and optimising its ability to forecast and plan effectively, Adamed Pharma turned to SoftwareOne. As a long-standing Microsoft partner, SoftwareOne is well placed to recommend and implement the most effective solution.

Adamed Pharma knew that automating activities would help it reduce friction, eliminate time-consuming manual processes and minimise long lead times. The company also wanted to leverage predictive analytics to reduce the margin of error in its forecasts. Based on these requirements, SoftwareOne designed and developed a fully extensible platform built on Azure Data Services and Power BI that would enable Adamed Pharma to generate automated forecasts to optimise its supply chain and marketing decision making.

Improving forecasting accuracy with Azure-based architecture

In order to build an accurate forecasting model, large amounts of data (both internal and external) from different parts of the business had to be consolidated. This included marketing spend, customer data from CRM, budget data and competitor information, as well as the attributes of different medicines. Azure Data Factory was used to connect to, process and unify data from multiple disparate sources. Each of these data sets was consolidated within Azure SQL.

SoftwareOne then used this data set to build and train a forecasting model using the Azure Machine Learning Service. Numerous models were tested for each product line, including logistic regression, random forest, and gradient boosted decision trees. The model was programmed to generate 18-month forecasts, which would be updated each month if any of the contributing factors changed. Once the team was satisfied with the model’s performance, they deployed it as an Azure web service, making it accessible to both Power BI and third party applications. To ensure that the forecastings remain as accurate as possible, the model is retrained on a monthly basis using new data.

Enabling self-service reporting and forecasting with Power BI

With out-of-the-box connectivity to Azure SQL, SoftwareOne was able to bring all of Adamed's business data directly into Power BI. This enabled Adamed Pharma to leverage the seamless integration of Azure Machine Learning within Power BI to enrich the data with product forecasts, all with a few clicks. Because all data transformations in Power BI are recorded sequentially, Power BI calls the Azure Machine Learning web service every time data is refreshed, with no setup or user intervention required.

Once the data was cleansed and enriched, Power BI dashboards could be built on top of it. SoftwareOne created a series of reports that addressed business issues for different types of personas, such as board members, department heads and sales representatives.

The dashboards provided analysts with a validation report, allowing them to compare the output of the Azure Machine Learning model with the previous system and actual data. The significantly improved performance of the new model gave the analysts confidence. They also gained the ability to track monthly trends and behaviours across different markets, brands and SKUs, further increasing their confidence in Adamed's new forecasting system.

The new Power BI reports enabled product managers to monitor and forecast product demand over the next 18 months. In addition, users were now able to drill down into each of Adamed's individual brands and examine corresponding products produced by competitors. This helped product managers answer key business questions, such as:

• Which products are most likely to become market leaders?
• Based on competitor momentum, how and where should we prioritise marketing spend?
• How much of a particular product should we produce in the next 3 months?