Type of Data Needed

In order to take advantage of the ClarityRetail service, data must be provided by the retailer. At a minimum, the retailer must provide sales transaction data. However, the retailer can enhance the value obtained from the service by providing additional information. ClarityRetail uses the following type of data:

• Sales Transaction Data– This data at a minimum includes date and time of the sale, the quantity of the sale, sales transaction type and the UPC (Universal Product Code) the sale was identified with. This data is typically collected and generated by retail Point-of-Sale systems.
   
• Future Pricing and Promotion Data– This data will provide future item unit price and if available, associated promotional indicator determining what type of promotion it is being sold under. This data is typically managed in retail item management systems, price and promotion management systems or within the Point-of-Sale system.
   
• Special Events Calendar Data– This data will determine the special calendar events that occur from year to year where there is an expected influence on sales volumes. The data simply takes the form of an event name and the applicable from and to dates from each calendar year. This data does not typically reside within any retailer system and is manually created. Facility will be provided on the ClarityRetail web site for customers to input this data.
   
• Item Master Data– This data will detail various key attributes of the item such as UPC, description, unit of measure, item type, etc. This data is typically managed by retail item management systems or within the Point-of-Sale system.

Service Level Options

The following table describes the Service Level Options available and the type of data used by each service level. The Service Level Options are determined automatically based on the type of data provided.

 

Service Level	Description	Data Required
1 – Basic	This service level will provide sales demand forecasting on a UPC basis only with no item descriptions or hierarchy and will determine average unit pricing from each sales transaction. There will be limited ability to forecast several days in the future. Retailers with uniform unit pricing and a nominal number of price changes are ideal candidates for this option. Forecasting quality rating: GOOD.	- Sales Transaction Data
2 - Intermediate	This service level builds on top of Service Level 1 – Basic and adds to it advance pricing and promotional influence to the forecasts. This will greatly enhance the quality of the forecasts being generated. Forecasting quality rating: BETTER	- Sales Transaction Data - Future Pricing and Promotion Data
3 - Full	This service level builds on top of Service Level 2 – Intermediate and adds to it special events calendar data. The data will help improve forecasting quality by isolating special events such as Thanksgiving, Christmas, Senior Days, Homecoming, etc. It will determine sales volume impact leading up to and following the event. Forecasting quality rating: BEST	- Sales Transaction Data - Future Pricing and Promotion Data - Special Events Calendar Data
Enhanced Reporting Option	This is an add-on option that can be added on top of any of the Service Level Options described above. This option adds item master, department and category data that will greatly enhance the quality of reporting by providing full descriptions, department and category sorting. It will also allow the removal of erroneous undesired items from sales feeds, leaving only items found in the item master, thus controlling the scope of items on which the retailer is interested in reporting. Forecasting quality does not change with this service level.	- Item Master Data

Data Frequency and Timing

The frequency with which certain data is sent to ClarityRetail is almost as important as the data itself. The order or timing in which ClarityRetail receives data can be just as important. However, much of this really depends on the expectations of the retailer. For example, if daily reporting of forecasts is desired, then sales and price/promotion change data must be received daily and before the recalculation of any new forecasts. However, if weekly reporting of forecasts is all that is required, then there is more flexibility in timing of when this data can be received. This means all changes for the week are collected and only sent once per week before new weekly forecasts are recalculated. It is still important to note that all changes in data are required for the week. All that would change in this case is how often the retailer must deliver data to ClarityRetail. Depending on the size and capabilities of the retailer, this can help in reducing the management and burden of delivering data to the service.

Another concern in timing revolves around the minimum granularity of sales data to be received. The minimum is a single transaction per item per day representing an entire calendar day’s quantity and dollar amount of sales. This will support daily forecasts. As a future option provided by this service, it is possible to break the day into 2 hour segments and provide intra-day demand forecasting at that level. This might be useful for retailers that require insight into intra-day movement for highly perishable products. In order to support this capability, the retailer must be able to provide sales transaction data as close to real time as possible. We call this method of collecting data “real-time trickle feed”. This typically will require some form of software technology that resides on the POS controller and feeds it directly to the service as sales are completed. The providers of ClarityRetail can supply this technology as a value added option for common POS systems such as IBM’s 4690/Ace Systems or NCR’s ACS POS systems. 

The last remaining concern around timing of receiving data is with regards to how far in advance pricing and promotion data is received. For example, if this data is received only on the day of the change, then the forecasts provided can only reasonably be provided for the day of that change. Any calculation of forecasts beyond the receipt of accurate pricing and promotion data will introduce error based on incorrect assumptions related to that data. The ideal scenario would be to receive pricing and promotion data at least 7 to 10 days in advance of their effective date. This will provide users of the service with the ability to forecast at least a week at a time so that it might coincide with their ad or promotional cycles. It is also true that the further in advance pricing and promotional data is received, the further into the future the forecast can be reasonably calculated. If this is available from the retailer, there can be advantages to forecasting over larger spans of time into the future depending on the types of products the retailer sells.