Global Pharmaceutical Company
Forecasting & Backcasting
A global generic pharmaceutical manufacturer (own & 3rd party plants), was undergoing a major ERP system change, including demand planning. There was a need to establish trust in the statistical forecasting approaches that would be implemented, alongside an understanding of how to set up the algorithms to ensure the right balance between responsiveness and stability. The client was aware of the risks of "best fit" algorithms but wasn't sure of the best way to approach this vital step in ensuring baseline forecasts were fit for purpose.
Problem
- The client team were somewhat lost as to how to compare and improve forecast performance on a country-by-country basis and how to deploy statistical forecasting effectively.
- Sequoia was asked to join the existing project team and to focus on forecast analysis & improvement, alongside ensuring wide-spread understanding of and trust in the new ways of working.
- Pilot markets were tested, and processes defined that would drive a positive impact on results globally.
Solution
- Programme of comprehensive analysis and process review: to identify causes and correlations in performance.
- Forecast assessment tool: created to both test and quantify the opportunity for introducing a statistical forecasting approach for baseline forecasting on a SKU-by-SKU basis.
- Back-Casting process deployed: this technique quantified the performance of different algorithms on each SKU: Location to be forecast.
- Accessible results: dynamic charts to facilitate understanding and trust in the results of both the current and proposed statistical forecasting approaches on a SKU: Location basis.
- Winning Hearts & Minds: interactive workshop for global markets to explain and support the move to more statistical forecasting.
Impact
Clarity of forecast approach: each market team could use the Sequoia tools and approaches to identify the:
- SKUs suitable for statistical baseline forecasting
- Most appropriate algorithm to use for each SKU
- Best parameters to feed the algorithm
This created a virtuous demand planning circle:
- Appropriate, realistic forecasts, were implemented quickly, which further increased trust in the statistical approach
- Time freed up to then analyse & add value to those where the demand pattern was more challenging, resulting in better performance on the non-statistically forecast SKUs, particularly NPIs, market-specific and seasonal events.
- Forecast Accuracy improved by an average of 9% at both the mid-term and M-3 level
- 7.8% drop in Forecast Bias - achieved in less than 12 month which in turn had a direct impact on inventory levels
Embedding openness to an adoption of statistically driven baseline was achieved via:
- a collaborative pilot which enabled key market demand planning teams to understand the process of assessing and creating statistical forecasts
- a comprehensive, engaging workshop where the findings were explained and championed by the markets involved in the pilot
- Ongoing senior leadership review, measurement & feedback of the adoption and performance of the statistical demand planning processes
