Supply Chain 101: Demand Planning

Special thanks to Ritika Singh for co-authoring this article and Felipe Matheson for comments on S&OP best practices.

If your company has a supply chain, it is likely that you will come across terms such as: demand planning, demand management, forecasting or planning organization. You will find yourself working within this team or somehow connected to the work that this team does.

Very simply, as the team that forecasts demand for your products, the Demand Planning team acts as a vital linkage between operations and the business.

But, before we talk about Demand Planning, let us understand the hierarchy of decision-making in a typical organization.

Topology of Decision-Making

Figure 1: Hierarchy of Decision-Making

Decision-making can be divided into 3 types:

1. Long-Range Planning: Typically, these decisions have a far-flung impact, looking 1–5+ years out. The intent of these decisions is to define important capital expenditures and relationships required to support business plans.
2. Medium-Range Planning: Very broadly, this is an exercise wherein teams try to forecast/shape customer demand, and find ways to intelligently match supply to demand. Other business functions such as Marketing, Finance, Sales and Merchandising will also spend time to understand how different scenarios of demand-and -supply affect metrics such as revenue, cost and margin. With these exercises, you are probably looking 1–6 months out (in some cases, up to 12 months) and viewing data in monthly or quarterly buckets. Companies may also choose to create separate planning processes by product portfolio.
3. Short-Term Planning & Execution: Short-term planning and execution strategies represent a more granular view of medium-range plans and detail specific actions that need to be taken to achieve stated goals. In this phase of decision-making, teams also take preventive or corrective actions if medium-term plans are based on incorrect assumptions or unforeseen events, e.g., raw material expedites, manufacturing overtime, order reprioritization, rationing of supply etc. Teams will probably focus on data 1 day — 1 month out, broken out in hourly, daily or weekly buckets.

If you’re interested in learning more, some great content is to be found in this webinar collection by o9 Solutions, and courses on inventory and capacity analytics from the edX Team.

Demand Planning — The Why and How?

In the previous section, we briefly talked about teams working to forecast and shape customer demand. But, what is the purpose of this?

To Intelligently Meet Customer Demand: Do any of your raw materials have long lead times? Do any of your products take a long time to manufacture? Can your customers wait for product to arrive if you need to start from scratch?

In many businesses, forecasting plays a great role in responding better to customer demand, especially when supply lead times can be quite large. Forecasting may be used to drive procurement of raw materials (procure-to-forecast), assembly of semi-finished goods (build-to-forecast) or even positioning of finished goods (make-to-stock) at different nodes in the supply chain.

However, the story is completely different if customers are willing to wait as long as it takes, e.g., luxury “collectors-edition” products. In such cases, planning organizations may do some basic forecasting to drive procurement or capacity planning decisions but may choose a make-to-order strategy to finally fulfil demand.

Figure 2: Lead Time and Demand Visibility; Source: DD-MRP V2 by Carol Ptak & Chad Smith

To Give and Support Business Direction: The business is interested in forecasting metrics such as revenue, cost and margin based on different demand scenarios to achieve business objectives, e.g., margin growth to improve shareholder profit-sharing incentives, margin growth for R&D etc. A key consideration of the demand planning process is to also see how well the demand plan supports business plans. It forms a strong basis for discussions on questioning the validity of certain business goals and also, demand shaping- a practice where planning organizations attempt to influence certain demand behaviors in customers through pricing, creative marketing strategies etc.

Demand Planning — Obtaining a Good Forecast

So, what does it take to develop a good forecast that minimizes lost sales, inventory build-up and promotes business metrics?

As mentioned earlier, forecasting is a collaborative process. A balanced view on forecasting with a mix of analytical methods and business intuition can be very helpful in predicting future outcomes.

However, at the end of the day, all teams in the demand forecasting process must reach an agreement or a consensus demand plan. This is locked and then sent to other teams to drive important decisions in operational and business planning.

A good demand planning forecast is built on the synergies of good people and their ability to effectively use data, business intelligence and system intelligence to create good forecasts.

Figure 3: Cross-Collaboration in Demand Forecasting Process

Best Practices

To see some other great articles on forecasting best practices, be sure to check out content by Nicolas Vandeput and Institute of Business Forecasting & Planning!

1. Keep Forecasts Unconstrained: Determine how much the end-customer will buy or can be influenced to buy. Assume there are no supply or manufacturing constraints. Some companies call this the unconstrained demand forecast.
2. Planning With Revenue Targets: A revenue target is an important guard-rail or a data point in the decision-making process. However, it is also important to actively identify ways to independently assess market appetite for different products wherever possible, e.g., independent assessments of market demand can help surpass targets.
3. Be Mindful of the Audience: Depending on the function and seniority, tools that are able to slice and dice data in multiple ways within time, region and product hierarchies are very helpful. Also, be mindful of units of measure, e.g., quantity units, quantity cases, weight, volume, currency etc.
4. The Perils of Demand Proxies: Remember that time when you delivered 2 months late on an important customer order? Using shipments data does not give you an accurate picture of historical demand, especially when the data has significant execution delays cooked into it. In retail environments, PoS (Point-of-Sale) data can get you pretty close but your visibility of actual demand could be severely limited by stock-outs.
5. Striving for the Complete Picture: Ensure standard system integrations and workflows exist to capture all historical customer orders for complete visibility of historical demand.
6. Normalizing Data for Product and Region Replacements: Normalize your data for important product replacements, e.g., phase-in/phase-out, and customer remapping. This can have a positive impact on your forecast accuracy as well.
7. Identifying the Right Granularity: Determine if it makes sense for you to forecast or predict forecasts at a given level of detail. For example, if you have centralized manufacturing and warehousing, does it make sense to forecast what specific customers want? Can the benefits of demand aggregation be used to create better operational forecasts? However, Area Sales leaders may be interested in seeing forecasts broken out by customer or Area. Create tools that are able to support the needs of different teams and smartly reconcile or compare forecasts.
8. Handling Outliers: This is a tough one. If you are using simple univariate time series forecasting methods, outliers can have devious effects on forecast output. In such cases, removing or correcting outliers may still not give accurate forecasts, especially if certain causal factors are at play and have to be modeled, e.g., markdowns, static/shifting holidays, temperature, quarter-end promotions etc. Explore using causal process models (e.g., dynamic regression time series models, random forests) to improve your accuracy. To learn a little more about outliers, check out this article by Nicolas Vandeput on outliers.
9. Looking Forward is Important: Being aware of the past is important when forecasting demand. But, only looking at the past is not a great idea. Leverage different leading/lagging indicators to add more intelligence to your demand planning process. Be mindful of the limitations of univariate time series models, e.g., they are not very good at dealing with inflection points, as explained in this great article. As we have seen with the recent COVID-19 pandemic as well, certain models may not perform well in black swan events.
10. Managing by Exception: Use alert rules to identify significant differences of opinion or departures in expected outcome. This would allow planning organizations to sanity check data for errors (incorrect assumptions, data entry mistakes) and also reduce operational planning complexity within lead time.
11. Evaluating Forecast Performance: Define simple metrics to monitor team forecasting performance with different methodologies. Explore metrics such as FVA (Forecast Value Add) to identify drivers of error in the forecasting process. Also, aligning forecast accuracy metrics to net lead time can provide insights on material shortages. Iterate on the process to reduce forecast error and bias. To learn more about using different accuracy metrics, check out this great article by Nicolas Vandeput.

Conclusion

It helps to understand how demand planning ties into the overall decision-making process in your planning organization. The demand plan is an important input in supporting operational and business plans. Strive to create an institutionalized demand planning process that is built on good data, strong process fundamentals and flexible tools.