A critical element when implementing a next-generation demand planning platform is system configuration. However, like any aspect of implementation, from selecting potential vendors to allocating the proper resources, a successful system configuration process is also rife with challenges. Here are six aspects to keep in mind when entering this phase of implementing next-generation demand planning technology.
There is no “silver bullet”
A common expectation of a new planning solution—especially one that is using the latest artificial intelligence and machine learning technology—is that magical algorithms will save the day and resolve all your forecasting woes. In reality, statistical solutions rarely work straight out of the box. There is no “silver bullet” to remove error, increase agility and reduce the workforce.
Solutions rely on data to be evaluated, understood, organized and tested for forecasts to reach ultimate refinement. Do not underestimate the effort involved in this process. There will be lots of dead-ends, failures, and reforecasting but the learning process provides tremendous insight. You will become a master of your own data.
You will need data analysts and data scientists for this activity to produce results. These resources will be the key to aligning system structure and performing data mining, scrubbing, feature selection, segmentation, and partitioning. And let’s not forget the parameter settings, model, level, and horizon tuning and tournamenting.
The forecasting “silver bullet” can be obtained then but it will require the right resources, comprehensive training, and a forecast refinement ethic. Plan for continuous improvement by creating a long-term plan that will allow the planning solution to evolve and improve.
Aim for “just right”
A system-generating forecasting solution requires hierarchies that correspond to the “goldilocks” syndrome. They should not be too deep, nor too shallow, but “just right.” Too big and there will be dataset excess, complexity, performance, and allocation issues. Too shallow and there will be limited analysis and poor statistical generation.
Hierarchies that do not match existing (or expected) Budget or S&OP levels will result in forecast allocation and reporting issues. This potentially results in a chronic lack of engagement and inevitable spreadsheet usage.
Focus on approvals, not balancing the controls
Demand planning is about trust. Creating plans that the business can use to undertake critical buy, make, and distribute decisions. A stable and trustworthy demand forecast requires resources that can be allowed to perform adjustments that are fit for the business without censure.
Demand planning without evaluation is a risky proposition though. One misplaced decimal place could have catastrophic consequences. It is important then, to have appropriate checks and balances in the planning cycle. Common implementation errors are having too many approvals which cause cycle time delays and create frustration or having no approvals or validation points.
KPI analysis paralysis
Being able to measure the forecast at aggregate levels for guard rails and strategic sense checking is essential. The standard approach to manage this is through Key Performance Indicators. KPIs should be calculated with methods and intent that everyone in the process understands. Ensure that your configuration provides KPIs in easy-to-read and dynamically updated dashboards.
It’s easy to fall into the trap of implementing too many exception alerts. Too many checks in the process can create confusion and conflicting priorities. Use a limited number of KPIs. Consider aligning KPIs to achievement levels and change them once levels of trust or accuracy are reached. Developing a strategic set of KPIs to achieve provides a clear path of improvement for resources and the business alike.
Defining too many KPIs isn’t the only common mistake. It’s also easy to select the wrong KPIs (or no KPIs) and the wrong lags for the business. All of this results in poor decision-making.
Data, information, and insight
Not using the best data and not using the latest data are easy mistakes to make. Short Horizon Forecasting, where everyone is focused on firefighting, should be identified. Instead, teams should learn to focus on mid and long-term horizons. Getting a more accurate forecast at an earlier point in time results in less short-term firefighting and enables better long-term decisions.
Some businesses use the sales forecast or budgets as the starting point for their demand plan but since sales and budget are predominantly calculated by value (and often at aggregated levels) the conversions to volume can be skewed. This is a flawed approach and should not be considered without very good reason.
Striking the right cycle time balance
Forecast cycle times are the most difficult to course correct during an implementation. These processes are ingrained with meeting calendars and business culture, which makes changes to more appropriate cycles (whether shorter or longer) highly challenging. Customer service levels and product lead times should determine the forecast cycle, not historical systems or culture.
Cycle time constraints are not helped when new solution designs insist on planning levels that are far too granular. Often data is too chaotic and sporadic at low levels but executives can incorrectly assume that granular detail equates to insight. From a system generation point of view, aggregation can build a more seasonal and smoother number. Another point not to forget here is that too much data can slow down system screens and processes.
In summary, statistical and ML forecasting is not a fast track to forecasting nirvana. Reaching that state will require the right resources and a tuning philosophy. Solution design should prioritize number generation rather than conforming to legacy or reporting expectations. Create deep, long-term relationships with your alliance partners. Don’t focus on magic quadrants but instead look outside to competitors, peers, and Project Zebra. Creating an employee development plan for your supply chain team is a critical aspect of the success of a next-generation planning tool. The future is not like the past, so embrace the change.
