Methods and apparatus to determine effects of promotional activity on sales
View Patent ↗Example systems, methods, processes, and apparatus for determining expected base sales for a product include obtaining sales data for a product sold at a point of sale location. The sales data can be organized in a time series according to a predetermined time period. The method further includes identifying a promotional event for at least one of the product and the point of sale location and excluding sales data corresponding to the promotional event. The remaining sales data is processed using a smoothed moving average model involving a plurality of passes through the remaining sales data. Expected base data for the product is generated based on the smoothed moving average model and output to a user.
1. An apparatus for product sales baseline determination, comprising:
a processor and a memory programmed to implement:
a data preparation and alignment engine to receive sales data for a product and causal data identifying a promotional event from a point of sale, the sales data organized in a time series for a time period, the data preparation and alignment engine to correlate the product sales data with the causal data to exclude sales data corresponding to the promotional event identified in the causal data to generate non-promoted sales data for the product; and
a modeling engine to process the non-promoted sales data using a smoothed moving average model including a smoothing constant, the smoothed moving average model comprising an exponentially smoothed moving average model and a smoothing constant to provide relative higher weight to newer sales data and relative lower weight to older sales data by assigning exponentially decreasing weights as the sales data becomes older in time,
the exponentially smoothed moving average model comprising a) a double exponentially smoothed moving average model or b) a single exponentially smoothed moving average model to be selected based on a determination of trend and seasonality in the time series data,
the exponentially smoothed moving average model involving a plurality of passes through the non-promoted sales data to generate expected base data for the product from the exponentially smoothed moving average model, the plurality of passes including a) executing a backward pass through the non-promoted sales data, b) executing a forward pass through the non-promoted sales data, and c) averaging the backward and forward passes,
the modeling engine to output the expected base data for the product to a user by at least one of generating a visual depiction of the expected base data for display to the user and generating a machine-readable representation of the expected base data for further processing.
2. An apparatus according to claim 1 , wherein the modeling engine is to compare expected base sales for the product to sales data corresponding to the promotional event to determine incremental sales for the product.
3. An apparatus according to claim 1 , wherein the modeling engine is to calculate short and long sales from the non-promoted sales data based on the averaged backward and forward passes and ignores outlying short and long sales data points in the time series in the plurality of passes through the non-promoted sales data.
4. An apparatus for product sales baseline determination, comprising:
a processor and a memory programmed to implement:
a data preparation and alignment engine to receive sales data for a product and causal data identifying a promotional event from a point of sale, the sales data organized in a time series for a time period, the data preparation and alignment engine to correlate the product sales data with the causal data to exclude sales data corresponding to the promotional event identified in the causal data to generate non-promoted sales data for the product; and
a modeling engine to process the non-promoted sales data using a smoothed moving average model including a smoothing constant, the smoothed moving average model comprising an exponentially smoothed moving average model and a smoothing constant to provide relative higher weight to newer sales data and relative lower weight to older sales data by assigning exponentially decreasing weights as the sales data becomes older in time,
the exponentially smoothed moving average model comprising a) a double exponentially smoothed moving average model or b) a single exponentially smoothed moving average model to be selected based on a determination of trend and seasonality in the time series data,
the exponentially smoothed moving average model involving a plurality of passes through the non-promoted sales data to generate expected base data for the product from the exponentially smoothed moving average model, the plurality of passes including a) executing a backward pass through the non-promoted sales data, b) executing a forward pass through the non-promoted sales data, and c) averaging the backward and forward passes,
the modeling engine to test the smoothed moving average model with the non-promoted product sales data to validate the model for use with the non-promoted product sales data, the modeling engine to output the expected base data for the product to a user.
5. An apparatus according to claim 4 , wherein the modeling engine is to compare expected base sales for the product to sales data corresponding to the promotional event to determine incremental sales for the product.
6. An apparatus according to claim 4 , wherein the modeling engine is to calculate short and long sales from the non-promoted sales data based on the averaged backward and forward passes and ignores outlying short and long sales data points in the time series in the plurality of passes through the non-promoted sales data.
7. A method of product sales baseline determination, comprising:
receiving, by a processor, sales data for a product and causal data identifying a promotional event from a point of sale, the sales data organized in a time series for a time period, the data preparation and alignment engine to correlate the product sales data with the causal data to exclude sales data corresponding to the promotional event identified in the causal data to generate non-promoted sales data for the product;
processing, by the processor, the non-promoted sales data using a smoothed moving average model including a smoothing constant, the smoothed moving average model comprising an exponentially smoothed moving average model and a smoothing constant to provide relative higher weight to newer sales data and relative lower weight to older sales data by assigning exponentially decreasing weights as the sales data becomes older in time, the exponentially smoothed moving average model comprising a) a double exponentially smoothed moving average model or b) a single exponentially smoothed moving average model to be selected based on a determination of trend and seasonality in the time series data, the smoothed moving average model involving a plurality of passes through the non-promoted sales data to generate expected base data for the product from the smoothed moving average model, the plurality of passes including a) executing a backward pass through the non-promoted sales data, b) executing a forward pass through the non-promoted sales data, and c) averaging the backward and forward passes;
outputting the expected base data for the product to a user by at least one of generating a visual depiction of the expected base data for display to the user and generating a machine-readable representation of the expected base data for further processing.
8. A method according to claim 7 , further comprising comparing expected base sales for the product to sales data corresponding to the promotional event to determine incremental sales for the product.
9. A method according to claim 7 , further comprising calculating short and long sales from the non-promoted sales data based on the averaged backward and forward passes and ignoring outlying short and long sales data points in the time series in the plurality of passes through the non-promoted sales data.
10. A method of product sales baseline determination, comprising:
receiving, by a processor, sales data for a product and causal data identifying a promotional event from a point of sale, the sales data organized in a time series for a time period, the data preparation and alignment engine to correlate the product sales data with the causal data to exclude sales data corresponding to the promotional event identified in the causal data to generate non-promoted sales data for the product;
processing, by the processor, the non-promoted sales data using a smoothed moving average model including a smoothing constant, the smoothed moving average model comprising an exponentially smoothed moving average model and a smoothing constant to provide relative higher weight to newer sales data and relative lower weight to older sales data by assigning exponentially decreasing weights as the sales data becomes older in time, the exponentially smoothed moving average model comprising a) a double exponentially smoothed moving average model or b) a single exponentially smoothed moving average model to be selected based on a determination of trend and seasonality in the time series data, the smoothed moving average model involving a plurality of passes through the non-promoted sales data to generate expected base data for the product from the smoothed moving average model, the plurality of passes including a) executing a backward pass through the non-promoted sales data, b) executing a forward pass through the non-promoted sales data, and c) averaging the backward and forward passes;
testing the smoothed moving average model with the non-promoted product sales data to validate the model for use with the non-promoted product sales data; and
outputting the expected base data for the product to a user.
11. A method according to claim 10 , further comprising comparing expected base sales for the product to sales data corresponding to the promotional event to determine incremental sales for the product.
12. A method according to claim 10 , further comprising calculating short and long sales from the non-promoted sales data based on the averaged backward and forward passes and ignoring outlying short and long sales data points in the time series in the plurality of passes through the non-promoted sales data.