How can you ensure your supply chain forecasting system scales with your business growth?

Let’s consider an ensemble learning approach to deliver the best forecast signal across all the available machine and human signals. We call this Touchless Forecasting at GitaCloud, where we generate optimal weights for all available signals to then produce a composite signal which is more accurate than any of the underlying signals, most of the time. We regenerate optimal weights every cycle and ensure stability and value add across cycles. We found this approach cuts forecast error by >20% in most cases. This is the pick-best equivalent functionality that goes beyond just stat forecast models, as demand planners / sales / marketing may win consistently for certain products in certain time horizons.

Well, for sure flexible foreacsting mehod is suitable as it keep evaluating the demand singnls and provides best fit forecast. Now a day ML models helps getting better results. As per my experience, forecasting should be done based on past history and upcomming signal. such as environment changes, competitors activities, events, possible entities which had impact on sales in past.. AI-ML has a lot of potential but its agains depends what you feed those model. Idally forecasting should be done on independent factors affecting dependent entities.. i.e. forecast events and based on those forecast what you expect to be forecasted. AI-ML is not magic it all statistic again which needs data points..

Automation of data definetely helps and saves time. In addition to that there should be user intervention based on exceptions on data patterns.. Automate data Generate exceptions on data patterns. Checks and balances base on exceptions Pass the input for further process.

Real time automatic data is still not wholistic as such. Sometimes you have to upgrade the data collection tools or system in use. New trend and new data collection tools are advancing regularly. Find a reliable tool and keep upgrading as it evolvers

The more data you collect the better. Well Said you don't have to do it alone. You will have to work with other departments and stakeholders.

One point to be noted in terms of foreasting is, " Forecast is always Wrong" All efforts are jut try to go close to foreast. if forecasting models and consensus demand plan is not giving correct results then it needs to be evaluated.. That is what foreast performance. Generate KPI find root cause where foreast accuracy is less than expected. Expectations should also be realistic.. measure current forecast accuracy and set expectations based on projected actions required. Generate action plan to improve foreacst accuracy. as per my experience maximum places its missing and user keep working on corrective action than preventive. All about change management again.

Update and refine forecasting system doesnt mean go for new systems unless there are real limitations to existing system. Many places its not about sytem but its all about the team working on that project. Maximum recent forecast solutions have capabilities but those are getting influenced by the leadership in the organization at many places. A bitter fact.