This website does not support Internet Explorer. Please, switch to Edge, Chrome, or Firefox browser to view this page.

Learn about IE support
Best Practices

Part I: How to improve your demand plan

Adoption and chage
Published: Reading time: 5 min
Simon Joiner Product Manager of Demand Planning
Simon JoinerProduct Manager of Demand Planning


Granularity of time

Demand plan cycle frequency

Move from univariate to multivariate



Demand plans are a core element in driving supply chain effectiveness, but there are many challenges that impact their efficiency and accuracy. A common question that planning managers ask is: How can I improve my demand plan? 

Whether you are considering an upgrade, or new implementation, or simply wondering about operational refinement, here are some situations that could leverage faster, neater, and more useful demand plans. 

The right granularity of time

Planning at a different granularity of time is an option to look at first. The granularity might be a finer or a more aggregated level of detail. Both choices can provide improvement opportunities through greater accuracy or process nimbleness. It all comes down to aligning time granularity with input and output data and the decisions to be made. It’s important to work at the best level for each activity.

Machine learning and the use of multiple drivers to generate refined forecasts are becoming the new gold standard in demand planning but driver effectiveness will be hampered if the levels of time available in your planning system are not appropriate for feature engineering. Weeks, partial weeks, or days may be required for optimal AI/ML forecast generation.

If your data arrives at a cadence not matched by the level at which you plan (say, you plan in months but load data every week), there will be opportunities for improvement to your demand plans. In the example given, the options would be to change to a monthly data load for simplification or to lower the time hierarchy level from months to weeks to obtain trend and seasonal insight.

If your planning system cannot hold data lower than the time level of months, then options for improvement will be limited to simplification. To improve trend analysis and begin the journey into demand sensing and shaping using AI/ML, you need to model a finer granularity of time.   Technology is the enabler here. Look to implement a planning solution that can utilize multiple dimensions and hierarchies because, with a greater set of levels at your disposal, the ability to do the right thing at the right level over the desired horizon is unleashed (see: What Level to Forecast at?)

Demand plan cycle frequency

Staying on the subject of time, another opportunity for improvement is changing the frequency of your planning. Typical demand plan cycles run monthly, but demand volatility doesn’t respect meeting schedules. Driver signals appear daily or weekly and can easily be missed in a monthly planning process. As a result, they may not get identified until the next planning cycle. This means you’re late to see events that impact demand and late to react to them.  

The solution is to set up a more agile demand planning process so that the supply chain can react to change faster. Of course, a faster cycle implies more throughput for the demand planners, so trust in the truth of the demand signal is the key here. The ability to focus only on where change has happened and not the row and column of every category can dramatically increase demand plan efficiency and accuracy.

Changing the frequency of the planning cycle doesn’t mean that all combinations need to be managed the same way. Try organizing your planning reviews so that stable, slow-moving, and trustworthy data is assessed less frequently or with a much lighter touch than erratic, promoted or critical combinations. Don’t slavishly follow a standard review formula for the sake of it.

Move from univariate to multivariate

Traditional demand planning solutions take historical data such as sales orders or shipments and use time-series methods to create statistical forecasts. This is called univariate or single variable planning. This traditional forecasting puts too much weight on last year. As a result, demand planners spend significant time adjusting upcoming events and recent trends.

Multivariate is the modern planning approach where machine learning can assess multiple variables such as prices, search trends and indices, ILI seasonal data, holidays, prime days, weather, and consumer sentiment to create more refined statistical forecasts. Univariate and multivariate approaches can be run for different segments of the business or the results can be merged to create hybrid forecasts.

If your current system is univariate and cannot use leading and lagging data from internal and external sources, then moving to multivariate will require resource and system change to be effective. Invest in data scientist resources, and use open source systems that can ingest, process, cleanse and present data driver information for meaningful decision-making. However, don’t think that this approach requires a digital transformation before it can begin.   Research for the right data sources and evaluate the granularity and suitability for your planning needs right now.


You don’t need to wait for next year’s IT budget to take some of these improvement steps. Undertake action now to evaluate whether granularity changes, cycle frequency or external driver data can provide the enhancements that your demand plan needs. 

Leaders gartner banner image

Gartner® Magic Quadrant™ 2023

o9 Solutions recognized as a Leader in the 2023 Gartner® Magic Quadrant™ for Supply Chain Planning Solutions. Download for free the full Magic Quadrant.

I agree to the privacy policy.

About the author

Simon Joiner Product Manager of Demand Planning

Simon Joiner

Product Manager of Demand Planning

Simon Joiner is a Product Manager of Demand Planning at o9 Solutions. He has over 20 years of experience in transforming Demand Planning Systems, Resources and Processes in such diverse sectors such as Pharmaceutical, Building Supplies, Agriculture, Chemical, Medical, Food & Drink, Electronics, Clothing and Telecoms. Simon lives in Hemel Hempstead in the UK with his wife and two (grown up) children and in his spare time likes to play guitar, research family history, walk the dogs and keep fit with running.