In the pharmaceutical sector, forecasting is one of the areas that receives the closest attention. It underpins production planning and the optimisation of the supply chain. Forecasting is central because the actual availability of medicines on the market depends directly on its accuracy, which in turn shapes a company's ability to meet demand and set strategic objectives.

A crucial element of this process is the forecast hierarchy, which makes it possible to break data down into different levels of aggregation. This methodology, also known as forecasting hierarchy, allows data to be analysed through a tree structure in which each level represents a different degree of aggregation by time, geography or product. Each level therefore corresponds to a different degree of detail, offering both a granular and an overall view of the forecast.

In pharma, the forecast hierarchy helps manage the complexity of data coming from different sources: the supply chain, sales, and market research. This method not only supports collaboration between country managers, regional managers and account managers, but also makes cumulative forecast values visible at controlling-account level, improving transparency and decision-making efficiency.

Using machine learning techniques and advanced analytics tools, companies can manage and analyse large volumes of historical and scenario data, simulate a range of market events, improve forecast accuracy, reduce uncertainty and optimise resources.

This approach makes it possible to account for variables such as product hierarchy, patient population growth, research innovation, emerging competitor products and government policy, delivering more accurate and reliable forecasts. Adopting time series forecasting models within a hierarchical structure also allows future scenarios to be simulated and market shifts to be met promptly, strengthening the decision-making capacity and competitiveness of pharmaceutical companies.

The three dimensions of forecasting

Demand forecasts in the pharmaceutical sector rest on three main dimensions: materiality, geography and time. Together they form the basis of the forecast hierarchy and make it possible to build a detailed, precise picture of future demand.

• Materiality: forecasts can be produced at different levels of detail, such as SKU, product, segment or brand. They can also be measured using different metrics, including units, value and active ingredient.

• Geography: demand can be forecast against a range of geographical criteria, such as country, region, market, channel, customer segment or warehouse.

• Time: forecasts can be produced across different time horizons, from daily to weekly, monthly, quarterly or annual.

Alongside the three main dimensions, another crucial aspect is the forecast horizon: how many periods ahead you need to forecast. This can vary considerably depending on business needs, from a single week to two years or more. In pharma, the forecast horizon matters particularly for supply chain management and resource planning, because it allows the business to adapt quickly to market shifts and research innovation.

Scaling forecasts up or down

The forecast hierarchy is a crucial element in pharma, where forecast accuracy can significantly affect the supply chain and product availability. Forecasts can be built at different levels of aggregation, such as store, day and product, or month, country and product family.

Three main techniques can be used to optimise this process:

• Bottom-up: aggregate forecasts are built by summing the granular levels below. For example, adding together the forecasts for each warehouse in a country to reach a national forecast.

• Top-down: detailed forecasts are built by disaggregating high-level forecasts using simple rules or historical weighted averages. For example, turning a monthly forecast into a daily one using a flat split, or basing it on the historical revenue of retail outlets.

• Middle-out: this intermediate technique uses a single forecast to generate both a more aggregated view and a more detailed one. For example, starting from a country-level forecast, summing it by geographical region and distributing it by retail outlet.

Adopting hierarchical forecasting models in pharma does more than improve forecast accuracy: it also enables more efficient resource management and a better response to market dynamics. Using machine learning algorithms and advanced platforms (for example Profiter's artificial intelligence solutions for the pharmaceutical sector), companies can analyse time series forecasting and simulate complex scenarios, gaining greater responsiveness and adaptability.