o9 AI/ML Forecasting Capabilities with a Large Consumer Health Enterprise
Summary video of the Demand Planning DPP we performed with the consumer health division of a leading pharma company.
The objective of this proof-of-concept (POC) was to assess whether a new digital demand planning solution can enable the use of multiple data sources to generate forecasts, enable planning via exceptions in a simple user interface, and ultimately drive improved forecast performance through a standardized solution.
The POC was successful in demonstrating that o9 can meet all of these objectives. o9 uses a multi-variant segmentation approach to analyze historical sales patterns, the type of product, stage of life cycle, and other factors to generate forecasts. The solution can also accommodate variations in business models and data availability, and it enables cross-functional collaboration and exception-based planning.
As a result of the POC, o9 was able to improve forecast accuracy by 10%. The solution also helped to improve the efficiency of the demand planning process by automating Excel dashboards and KPIs.
Overall, the POC showed that o9 is a valuable tool for businesses that need to improve their demand planning process. The solution is easy to use, flexible, and accurate. It can also help to improve the efficiency of the demand planning process, freeing up time for demand planners to focus on other tasks.
The objective of the POC was to assess whether a new digital demand planning solution can enable the use of multiple data sources to generate forecasts, enable planning via exceptions in a simple user inter ways to ultimately drive improved forecast performance through a standardized solution.
Approach taken to generate the machine learning based forecast.
For the forecasting approach, o9 does multi variant segmentation, analyzing historical sales patterns, the type of product, stage of life cycle, and more. It detects the level of algorithm identifies the drivers, tunes the algorithms and carries out a final forecast generation.
All variations of business models, data availability, and different horizons were accommodated in the same platform. For level determination of algorithms, o9 is not constrained by the provided hierarchy, which may not always represent the optimal grouping to apply the algorithms.
Instead, we analyze the end customer demand centric attributes of the product based on structured and unstructured data to understand substitutability, cannibalization, and grouping potential.
This is done by leveraging the customer's master data and publicly available information.
Know that both of the products with Sensidine have similar purposes, strengths, and or action.
As shown here, o9 could break apart or club together the children with a particular category or brand based on attribute analytics.
For drivers, o9 is able to import customer data and augment it with external sources, for example, holiday and weather data or macroeconomic factors. It can also act as a collaborative repository for maintaining all the event information, all in one platform. To illustrate the performance and root cause analysis approach, let us look at one example. It is seen that o0 is showing better accuracy in all cases. However, there is room for improvement when it comes to sensodyne co pays. This is a highly promoted brand and the model couldn't capture these deviations due to a lack of promotions information.
How can we do better? With better data, quality, availability.
As shown here, the accuracy potential goes up substantially with the right driver inputs.
The good news is o9 can help organizations in identifying, collecting, cleansing and maintaining a variety of internal and external driver data elements in the platform. In addition to applying advanced ML models, O9's platform also enables cross functional planning process, collaboration, exception boarding, and process orchestration.
Let's take a look at O9's workflow for demand planners.
The planning star hinges on an exception based paradigm, which is why users start with a dashboard with key insights to be reviewed or acted upon, and top learnings to be baked in future cycles.
The forecast value add dashboard helps users understand where the planner forecast is doing better than the system and vice versa Reporting these metrics helps improve the efficiency of the consensus process and enables more holistic adoption of system driven forecasts by cross functional stakeholders.
Users can also view exceptions to cycle over cycle changes.
Sort on the gap to understand where the forecast has changed the most from previous cycles and drill down to skew level for more details.
Users have the ability to plan and review sell out and sell in forecasts and financial and volume forecasts in the same platform.
Product transition information can be managed and maintained in o9, and the history of a similar item can be added for forecasting per the system uses historical actuals and driver data as an input to generate driver based ML forecast.
The flow of information across participants of the process like sales and marketing schemes can be easily orchestrated here. After finalizing forecasts, the market manager can publish the forecast for further review to the next stakeholder. In this case, if the market owner submits the plan, The area manager receives the alert as seen in the R and O dashboard.
o9 demonstrated various forecasting approaches accommodated variations and business models, and executed the end to end demand forecasting process in the platform.
In addition to higher accuracy, automation of Excel dashboards, KPI is an exception based review, hugely improves the productivity and performance of demand planners and cross functional stakeholders involved in the process.
Can AI/ML forecasting help improve your forecast accuracy rapidly?
Learn how to efficiently manage internal and external data with an end-to-end digital platform.