7 steps to better demand forecasting

The predictive nature of forecasting makes it a notoriously tricky process to manage. Stakeholders from multiple business areas with different perspectives, needs and often conflicting agendas, must come together and collaborate.

This post discusses seven key considerations that will help set you up for demand forecasting success.

1. Define how the forecast will be used

Start by determining how the business will use the forecast. This step is critical and will impact your process design, data requirements, and system configuration.

Documenting the business decisions and processes that the forecast needs to support will ensure that you meet all your requirements. Here are some examples where a business will benefit from forecast information:

• Short Term – order fulfilment, distribution between warehouses, production planning, purchase planning.
• Medium Term – balancing demand and supply constraints, sales and marketing tactics, promotions, supplier contracts, product lifecycle management.
• Long Term – identifying gaps to strategy, financial planning and budgeting, long-term supply contracts, segments to invest in or exit from.

2. Define relevant KPIs

Once you’ve determined how the forecast will be used, you’re ready to develop supporting KPIs. Take care to align all KPIs to your overall business goals. Poorly defined KPIs risk undermining what you’re trying to achieve as a business.

A good set of KPIs should be measurable, well understood by your organization and supportive of your overall business goals.

Forecast Accuracy KPIs such as Mean Absolute Deviation (MAD) and Mean Square Error (MSE) show the forecast’s overall accuracy and quality and inform planners what forecasting method performs best in a particular situation.

Understanding forecast bias (mean forecast error) is essential to improve demand planning over time. Including a forecast bias indicator will expose if there is a tendency to over or under forecast.

Knowing what success looks like will help guide your process design and system configuration.

3. Understand where detail and accuracy matters

It’s now time to determine the level of forecast detail and accuracy that’s required to meet your specified business goals. These levels will largely depend on your time horizons.

• Strategic Goals – Monthly or quarterly forecasts over a longer time horizon. High-level data is often sufficient, e.g. at the sales division and product category levels.
• Operational Goals – Daily or weekly forecasts over a shorter time horizon. More granular product, location and market information is usually required to fully support the planning of production, transportation, storage and sales.

A word of caution! There’s often a tendency to include as much detail as possible upfront. Although this can provide extra flexibility, it also adds risk in the form of:

• Information overload – Too much information can be intimidating to some users
• Focusing on the wrong thing – Consider a system that creates a weekly forecast by product and customer for a whole year—in most cases, forecasting at such a detailed level that far into the future will add little value. Also, the result is typically less accurate than a forecast generated at a more aggregated level.
• Unnecessary system load – Suppose a business has a large customer base managed at a customer segment level. The company can reduce both database size and processing time by excluding detailed customer data. Similarly, if a business effectively runs at a monthly level, adding daily or weekly information may not be needed.

4. Be mindful of too much complexity

A new demand planning system will have exciting features that seem appropriate for your business, but starting with the basics is both the safer and more practical approach. This advice is especially true for companies moving from Excel-based forecasting to a dedicated forecasting tool.

Running a statistical forecast improves the demand plan. However, expecting to rely solely on the statistical forecast without any overrides is unrealistic as it requires rigorous process monitoring and data management. Instead, use the statistical forecast as a benchmark for disaggregating higher level forecast changes.

Significant changes are possible, but they come with administrative overhead and data management challenges. Adding new functionality in stages as the forecasting process matures allows time to build trust in both the system and the process.

5. Get everyone on board

Demand planning is a core business process and vital to your company’s success. Support from senior management and key stakeholders is critical, as ambitious changes are often rolled back after meeting resistance.

Getting good forecasts out of people may require different tactics. “Reluctant forecasters” want a quick, simple process for making their projections. On the other hand, the “convince me it’s better” users want to understand more about the process and how statistical forecasting works.

The goal should be to get as good a forecast as is required, as quickly and efficiently as possible. Understanding the participants’ thought processes and pain points makes it easier to explain the benefits of the transition. Try to identify how the new solution can make things easier – e.g. through forecast automation and better demand collaboration tools.

6. Segmentation is king!

Product and customer segmentation is an essential part of forecasting. It lays the foundation for all analysis and reporting, forecast entry, statistical calculations, KPI definitions, and life cycle and promotion management.

The segmentation must provide the correct attributes and levels of detail. Unfortunately, existing segmentation is often based on what your finance, operations and/or sales departments needed at some point in the past – or what data was technically possible to extract. Therefore, you may need to update your master data to meet the new solution requirements. Consider the following when making decisions about segmentation:

• Data segments should support the level of detail required to meet the forecasting goals
• Areas of responsibility, e.g. product categories, geographies and customer segments
• Facilitate grouping of products and customers that behave similarly in the market to support more advanced analysis, statistical forecasting and life cycle management
• Aim for flexibility by allowing for future changes in focus and strategy

The ability to easily modify the segmentation is essential in delivering a usable tool. Without this functionality, master data management becomes more time consuming and error-prone.

7. It’s all about data quality

Last but not least, you need to consider data quality. If you want good forecasts, good master data is a must. Poor data quality will result in a variety of issues, including implementation delays, re-work and lack of user trust in the system.

Many data issues stem from inadequate setup and management of master data in your ERP. If the data is not solid, both forecast management and analysis will suffer. Address this issue by appointing owners for each of the master data fields used in the forecasting process.

Summary

An accurate forecast that supports business analysis and decision making is invaluable, and investing in your company’s demand planning process should be a top priority.

By carefully considering the seven areas above, you will set yourself up for a successful implementation. Remember! You’re aiming for a forecasting solution that will not only bring a few quick wins, but one that will grow with you and bring success to your business for years to come.

If you’d like to learn more about how Optimity can help your business, get in touch. We’d love to chat.

And if you’d like to read about the difference Optimity can make. Read the Bubbies story.