Caveats in a Forecasting System implementation project
Let me start with a clichéd statement - forecasting drives supply in your network. It sets the tone of supply chain planning. In many different contexts, we have heard about what different proponents of forecasting are saying- a one percent improvement in forecast accuracy saves sometimes millions of dollars in inventory and other short-term working capital requirements. Granted all this holds lot of water and consequently, we will probably have millions of dollars spent in forecasting and related consulting projects in coming years.
Well the intent of this blog is not to talk about the obvious, however talk about something that has been a subtle but very powerful learning over a course of two very diverse forecasting related implementation projects. These two projects spanned across different organizations, completely unrelated industry verticals and yes, different technologies altogether.
Two very important learning from these forecasting projects were:
1. Forecast is not about numbers, but about people
2. Focus on business process workflow is far more effective than the tool per se in a Forecasting project
Well let me talk a bit more about them. Quite often forecasting is made synonymous with statistics and consultants/organizations spend most of their time developing forecasting algorithms that can detect that slight trend or cyclicality or changing seasonality pattern. In most cases algorithms work on numbers that may not be even “forecastable”. Organizations sometimes not even draw a consensus on what numbers to forecast upon. There are unanswered questions like - when should an organization realize sales, how should returns be treated, should billing cycles be accounted for, should an organization forecast on shipment, invoice date/quantity or better still point-of-sales. Sometimes organizations are also constrained by what information they have and can get or have control over. Given all these variables, it is very difficult for a forecasting organization to process actionable projected sales numbers from even best of algorithms.
Imagine another situation where planners have to do forecasting distinctly on multiple geography-specific hierarchical structures or complex dynamic product hierarchical structures. One cardinal and painful truth that most organizations discover only after-the-fact is that a very complex forecasting business process breeds a very unmanageable demand planning solution with numerous customizations. The management consequently spends its attention and investment-worthy capital on maintaining this complex landscape. They spend time and energy in procuring and installing more powerful machines.
Forecasting solutions fail or do not deliver expected results if - either the solution is not embraced by the stakeholders or is way too complex to catch anyone's imagination.
Some recommended steps in a forecasting implementation would be:
1. Re-engineer the forecasting workflow to make it simple, repeatable, exception-managed and reportable.
2. Make sure you invest on people since they are bound to do a better job in conjunction with a forecasting system solution than the two working in isolation.
3. Clear integration of forecasting workflow with S&OP planning.
4. Tracking and enabling a feedback loop of ex-post released forecasts to reinforce unbiased forecasting management behavior.
5. Always driving a consensus with Marketing, Sales, Finance, external partners, Production and other important stakeholders.
6. Sharing and collaborating forecast numbers with downstream supply chain partners.
Whoever said Forecasting was number crunching! It is much more than that and may be none of that.
Two very important learning from these forecasting projects were:
1. Forecast is not about numbers, but about people
2. Focus on business process workflow is far more effective than the tool per se in a Forecasting project
Well let me talk a bit more about them. Quite often forecasting is made synonymous with statistics and consultants/organizations spend most of their time developing forecasting algorithms that can detect that slight trend or cyclicality or changing seasonality pattern. In most cases algorithms work on numbers that may not be even “forecastable”. Organizations sometimes not even draw a consensus on what numbers to forecast upon. There are unanswered questions like - when should an organization realize sales, how should returns be treated, should billing cycles be accounted for, should an organization forecast on shipment, invoice date/quantity or better still point-of-sales. Sometimes organizations are also constrained by what information they have and can get or have control over. Given all these variables, it is very difficult for a forecasting organization to process actionable projected sales numbers from even best of algorithms.
Imagine another situation where planners have to do forecasting distinctly on multiple geography-specific hierarchical structures or complex dynamic product hierarchical structures. One cardinal and painful truth that most organizations discover only after-the-fact is that a very complex forecasting business process breeds a very unmanageable demand planning solution with numerous customizations. The management consequently spends its attention and investment-worthy capital on maintaining this complex landscape. They spend time and energy in procuring and installing more powerful machines.
Forecasting solutions fail or do not deliver expected results if - either the solution is not embraced by the stakeholders or is way too complex to catch anyone's imagination.
Some recommended steps in a forecasting implementation would be:
1. Re-engineer the forecasting workflow to make it simple, repeatable, exception-managed and reportable.
2. Make sure you invest on people since they are bound to do a better job in conjunction with a forecasting system solution than the two working in isolation.
3. Clear integration of forecasting workflow with S&OP planning.
4. Tracking and enabling a feedback loop of ex-post released forecasts to reinforce unbiased forecasting management behavior.
5. Always driving a consensus with Marketing, Sales, Finance, external partners, Production and other important stakeholders.
6. Sharing and collaborating forecast numbers with downstream supply chain partners.
Whoever said Forecasting was number crunching! It is much more than that and may be none of that.
