|
One of the most important facts to remember about a forecast is that it will be wrong! The only question is how wrong can it be and still be valuable to your business. It is intuitive that a more accurate forecast is better than a less accurate one, and this is true. The really crucial questions are how much did it cost to improve the forecast, is it sustainable, and do the benefits outweigh the costs?
A very key question is what the forecast is going to be used for. There is a difference between a sales forecast used for financial planning which can be at a relatively high level and rolled up, and one used for sales and operations planning, where a much greater level of detail is required. Although there are circumstances where one forecast can be used for multiple purposes this is not advisable as statistical anomalies kick in when low level forecasts are aggregated, and errors become magnified.
Assess The Benefits
The first step in any forecasting project is to assess the value that will be gained by a forecast. How accurate does the forecast need to be to deliver the benefits and what is the incremental benefit change of an increase or decrease of the forecast accuracy obtained.
Altimus do this in two stages. Firstly past historical trends need to be analysed to see if the data is random, or if there is, for example, any seasonality in the data, which yields a chart like the one to the right, peaks and troughs show seasonality.
Secondly we build a simplified model of the fulfilment process and experiment with the model to see the effect of changing forecast accuracy on the key supply metrics (inventory, resource utilisation, availability and so on). This yields a target for forecast accuracy which will deliver value to the business. If the costs of achieving this are greater than the value which can be accrued then quite simply it is best to spend the money elsewhere.
The next step is to think about how the customers and products/services can be categorised. For example in a manufacturing environment a product which is bought in small quantities by thousands of customers will very likely be treated differently to one which is bought in large quantities by a single customer.
Other factors to consider at this stage would be expected lead time or acceptable queue times, customer expectations of delivery and so on. In some businesses it is a competitive advantage to reduce lead time; in others it is important to fix service time at a particular level.
Design The Process
Once we are confident that there is real benefit to be derived the next step is to design a process for delivering the forecast. All businesses are different, and will derive competitive advantage from their forecast in different ways. However most will use a variation of the flow on the right.
One of the key questions is how much should be done manually based on direct conversations between sales staff and the customers, how much should be algorithm based, and of the algorithm based ones how much should be manually revised. This differs between industries and companies. Our clients with a few hundred products and less than a hundred customers do things very differently to those with a million products and tens of thousands of customers.
Another important area to address is the level of the forecast. Forecasting can be done at a high level and blown down to product/customer level, or done at a low level and aggregated. It can also be done by value and converted to volume via an average selling price, or by volume and converted to value via known contract pricing. Again these are design questions which need to fit the nature of the business.
Sales data to power the forecast will almost inevitably come from a sales system. It is also important to decide if the forecast output will be automatically fed back into a system to allow ERP style forward planning, or if a manual process will suffice.
Choose The Tools The tools required for executing this process generally consist of the following elements:
- A database for storing all the data
- A set of appropriate algorithms for processing the data
- A spreadsheet or web based tool for manual input
- A spreadsheet or Business Intelligence tool for viewing the data
- An appropriate view to determine and evaluate forecast accuracy
Some specialist tools can be purchased to do all or part of this. However experience suggests that these are over-complex for most needs and are thus used poorly. Our approach is to use very simple tools, easy to use and understand – both in use and the output – and combine them in a powerful way.
Run The Process The process now needs to be run. It is important to consider this as an iterative process. In particular when there is manual intervention in the creation of the forecast there is the possibility of consistent over or under forecasting by particular people. This is known as forecast bias and needs to be addressed either algorithmically or through education. As forecast accuracy is monitored so the process can be continuously improved and fine tuned to deliver further benefits.
This review step is crucial to get benefit from the system. It needs to be rigorous and address the issues in a way which encourages participation and not finger pointing. The review must also at predetermined frequencies consider whether the expected benefits are accruing, and if not why not.
People Issues Finally a forecasting project involves people, and issues arise here due to management systems and incentives.
For example if a sales person is involved in the forecasting process are they held accountable for their forecast? They will often have sales targets on which compensation is based. How does this influence their forecasting behaviour? How does the company maintain forecast accuracy when a sales person wins a big unexpected order?
Many CRM systems assign a percentage win likelihood to an opportunity for business. From a sales revenue point of view two 50%’s make a whole. From an operational planning perspective both opportunities are either won or lost, each being for different customers and different services.
All of these factors, and more, means that education is important, in addition to training on the process. Staff from all areas of the Company should be taught together and explore the issues and how they are going to handle them. This ensures a consistency of purpose and understanding which is essential for delivering the benefits.
Conclusion Sales forecasting when designed and deployed correctly can deliver great benefits. There are however many pitfalls to be avoided in designing and implementing an effective forecasting process. Altimus has worked on many projects over the years and can assist in all phases of forecasting; from initial assessment through process design and implementation and help you maximise your return on investment from the available technologies.
To find out more about Forecasting and how it can benefit you, call Altimus on 0800 804 6442 and speak to one of our experts, or contact us through Information Request
|