Atlantec Software Limited - proFORE Strategic Overview proFORE is a unique tool for forecasting the daily sales of all types of product, particularly perishable and other limited-life goods. In addition to forecasting daily sales during non-holiday weeks, it finds particular and unique application when forecasting daily sales for the Christmas and Easter holidays when changes in people's daily shopping patterns skew the usual daily trading patterns. proFORE also has powerful segmentation tools to group products according to their daily or weekly sales profiles. The Problem But why is it particularly important to have accurate daily sales forecasts for Christmas? We all know that Christmas poses big challenges for retailers. If these challenges are overcome, the rewards in terms of financial performance and customer satisfaction can be very significant. Many retailers spend much of the year planning for Christmas, and indeed many retailers make most of their profits in the three months up to and including Christmas. If the sales forecasts for this period are correct, both retailers and customers will be satisfied — customers will be able to purchase everything they want to when they want to and retailers will not be left with unsold stock. The challenge is to tread the fine line between the extreme cases of overstocking (high availability just before the holiday but being left with a lot of unsold stock at the end of the holiday) and understocking (low availability just before the holiday but having no unsold stock at the end of holiday). The immediate financial costs of these extreme cases are obvious. Furthermore, understocking can lead to some dissatisfied customers not shopping again at retailers where they had bad experiences. Overstocking means that unsold stock must either be removed quickly or cleared by markdown, thereby reducing the margin and so possibly selling the stock at a loss. The time and effort required to process all this unsold stock cannot be ignored. These problems are serious during non-holiday periods, but even more so at Christmas for at least four reasons: the daily shopping profile during Christmas is different from that at non-holiday periods; the huge increase in sales at this time of year; shops are closed for at least 2 days; and when shops reopen people shop on a 'need to top-up' basis rather than on a planned basis. The Solution The solution to these problems lies in the fact that different daily sales forecasting models are required for different times of the year, particularly for periods of skewed daily sales profiles. The nature of the product also affects the type of forecasting model. Atlantec has isolated and modelled the various combinations of product type and time of year, and incorporated these models into proFORE. Thus, proFORE has a number of forecasting models, one for each combination of product type and time of year. These models form the heart of proFORE. The unique and proprietary algorithms in proFORE recognise that daily sales depend on the day of the week and the week of the year. As a result, proFORE forecasts daily sales directly based on the actual day in the year rather than forecasting the weekly sales and users then having to impose a daily sales distribution on the forecast weekly sale. This approach means that proFORE forecasts daily sales to far greater accuracy than do the forecasting models in other forecasting software. With respect to Christmas and Easter, the solution lies in understanding how the daily shopping profile during these holidays differs from the non-holiday daily shopping profile, and then adapting the non-holiday daily forecasting models to consider the additional factors present at Christmas and Easter. Indeed, we believe that no other forecasting product considers the special factors that cause people to change their daily shopping habits at Christmas and Easter — such holiday periods are usually treated in much the same way as non-holiday periods, and so the potential for significant under- or over-stocking during these periods is huge with all the consequent losses. Product Segmentation proFORE also has powerful cluster analysis tools to segment products into homogeneous groups based on their daily or weekly sales patterns. This allows the similarities and differences in the trading patterns of products to be analysed and compared. Traditionally, segmentation of products has been carried out by comparing the products' mean sales. However, this method fails to recognise that because the mean is, by definition, a summary statistic the daily and weekly variation of the raw data from which the means were calculated have been lost. In contrast to this summary approach, proFORE uses all the daily or weekly data for the chosen period to segment the products. An example of where proFORE's segmentation tool would reveal different groupings but analysis of the mean rates of sale would show the same groupings is for products which have the same mean daily rate of sale but different daily profiles – some products may be bought more at the end of the week than at the beginning, for example multi-pack yoghurts. These differences in daily demand profiles have implications for the supply chain right back to the suppliers. The segmentations can be carried out on daily or weekly sales, and on an absolute or percentage basis (the percentage basis removes the effect of the level of the sales for the same daily or weekly profile). The Philosophy of proFORE It is now clear that the philosophy of proFORE is very different to that of other forecasting software. Its core design is based on having one forecasting module for each combination of product type and forecast period, and each module is labelled accordingly in the interfaces. This design philosophy manifests itself to the users by listing these descriptors rather than having a long list of forecasting models in the interfaces. This approach is in marked contrast to other forecasting software which is based entirely around forecasting models with no explanation of the suitability of each model to different types of product and different forecast periods. Thus, the philosophy of proFORE leads to four very important advantages compared to its competitors. They are: The interfaces are much more business-focussed. It can be used much more quickly than other forecasting software. It requires minimum user-intervention. Its forecasting algorithms are very advanced and far superior to those in other software. Thus, users do not concern themselves with selecting the 'best' model or the values of its parameters — all the models are hidden from the users and all the parameter values are calculated in the software. After selecting the correct module by of product type and time of year, users only choose the store, product, forecast period and historic data. It is in these very significant ways that proFORE offers a radically new approach to demand forecasting. Sean Christie, finance director of Northern Foods, has been quoted as saying (Financial Times, 16 January 2002) that because Christmas Day fell on a Tuesday [in 2001], many shoppers had delayed buying food until the weekend or even Christmas Eve. This meant that Northern Foods had to offer premium payments to persuade extra staff to come in over the weekend to meet last minute demand. Now, assuming that Northern Foods used one of the standard forecasting models, they would have forecast that about 36% of the sales during the week 18 to 24 December would occur on Saturday 22, Sunday 23 and Monday 24. However, had Northern Foods used proFORE to forecast sales for this week, the forecast percentage sales for these three days would have been 50%. Gordon Mason, EMEA president of Retek, has written (Logistics Europe, February 2002) about the importance of having accurate [daily] forecasts for the Christmas period so that the extent of the markdowns is reduced to the absolute minimum. He stressed the importance of having accurate historical data and forecasts as everything is contingent on this. During the development of proFORE, one major food retailer admitted to us that their annual wastage of fresh produce is £30 million. These observations confirm that the design philosophy of proFORE, i.e. having one forecasting module for each product type and time of year combination, are correct. Thus, proFORE offers a radically new approach to demand forecasting in terms of its much more business-focussed design and its far superior forecasting algorithms. Features and Benefits proFORE has many features and benefits. They include: It only requires EPOS data (date, product, store, sales). Both sales values and sales quantities can be analysed. The model for forecasting daily sales just before Christmas considers the day of the week that Christmas Day falls on. The models for forecasting daily sales for the Good Friday and Easter Sunday weeks consider the four-day public holiday and the fact that shops are closed on Easter Sunday. The models for the Christmas and Easter holidays calculate both short and long lead time forecasts. The daily sales during non-holiday periods are forecast using a nested time series model which considers the week of the year and the day of the week. The models for forecasting daily sales can be enhanced to consider hourly forecasts. The products can be clustered according to their sales profiles using all the data based on their daily or weekly, and percentage or absolute sales. It has a powerful and flexible database that uses both product and store information. Technical Overview