Three major points to improve the forecast accuracy of demand planning.
When the month ends, spreadsheet reports start to pile up on your desk. Countless sheets hurt team’s productivity. Yet no sufficient time is given to clean up and upgrade. Product demand last month was low so you may assume next month sales will be similar. Only then, you realize the forecast accuracy is pretty off. Your quick reaction for intervention may totally be misguided also when no data evidence is supported to challenge your assumption. You ask, what factors drive our demand and what can we do?
Well, you might have done forecasting in statistically less reliable ways all along. The bad news is that forecasting accurately is hard, both offline and online. However, opportunities exist for your organization because accurate forecasts improve supply chain management, optimize inventory allocation, and save cost from inefficiencies otherwise go wasted. The good news is we share with you some of the forecasting lessons we learned from many projects worldwide.
1. Manual forecast helps, but data modeling is sweeter
Your business knowledge is precious. However, translating that into an exact sales number for hundreds of products (SKUs) is a headache, costly, and unreliable if not impossible at scale. We all now know that demand forecasting is done more accurate and reliable with data and algorithm than manual estimation. Whether you have a forecasting analyst or outside consulting, the forecasts may incorporate all sources of data to explain your sales next week, next month… If you are a beverage firm selling drinks, ask yourself if you need external data like weather beyond your sales history, inventory, and marketing investment. You know why? Because I’m unlikely to buy drinks outside when it’s raining. Nor do I’ll buy more in a raining season than a sunny one. In other words, the demand forecast needs to account for seasonality and other factors. And that can be handled well with data modeling, plus the benefits of insights discovery from all relevant sources of data.
2. Identify drivers and model shall shine
Well, you know that manual estimation is not working. You have tried to model sales using uni-variate time series or others but your forecast accuracy remains bad. Worse than your baseline. No wonder you produce and ship the number of items to various cities and provinces so off again.
Ask how much do your team know about your data of product life-cycles. Where do those data sit? The complexity of analytics tools and scarcity of great teams’ talent can be a huge barrier to uncover opportunities in your data. That’s specifically why we are here to help. You want to acquire all sales history and combine with other sources of data into accessible storage for manipulation and insights discovery. So you can visualize your sales trend and understand correlations. Depending on your business, sales next month correlates not only last month quantity sold but also the month last year. Of course, no one factor fits all; extracting variables, that are matter and true drivers of your sales, is key to inventory optimization. Also leveraging advanced machine learning models can capture complex relationships between these drivers and sales. Once you understand the exact relationship say marketing dollars on demand, you can take corrective and customized actions to cause the future: meeting the right level of demand while shaping more needs. At the end of the day, that matters a lot.
3. Embed forecasting into your business process
New months roll in. Messy spreadsheet reports come again. But it doesn’t have to continue this way. You can ask questions about your data and get answers quickly. You may have some access to a proper design of performance monitoring dashboard and analytics tools. To the extent this is true, a huge value from accurate demand planning can be realized more in the embedding stage that fits into your business process. So you can optimize your decisions at every stage on-demand, and value-added.
After all, combining domain expertise with data analytics to find drivers of your product demand will help improve your forecast accuracy. What it means improving by 1%, 5%, 10% in terms of revenues is to be found. So you can benefit fully in the operations of analytic results.
Reach out to me for comments or if you need to optimize your operations, inventory, and demand planning in consumer goods, energy, site tower traffic … etc or discuss for general analytic services.
The article originally appears at DataTicon.
