Description:SYSTEM AND METHOD TO DYNAMICALLY ADJUST BIAS IN STOCK KEEPING UNIT FORECASTING
TECHNICAL FIELD
[0001] The present disclosure generally relates to inventory policy planning system. Further, the present disclosure particularly relates to a system to dynamically adjust bias in stock keeping unit (SKU) forecasting.
BACKGROUND
[0002] Accurate bias correction in forecasting is essential for calculating safety stock norms. Traditionally, fixed historical periods are used to compare historical sales data with forecasts. Said comparison helps identify over-forecasting and under-forecasting trends, along with their respective magnitudes. Such identified trends are subsequently applied to correct forecasts for future periods.
[0003] However, said fixed-period approach demonstrates limitations in scenarios where stock-keeping units (SKUs) exhibit fluctuating behaviors. For example, certain SKUs may display over-forecasting for specific months followed by under-forecasting in subsequent months. In such scenarios, averaging the forecasting bias across said fixed period may lead to a near-zero value, thereby misrepresenting the actual bias. For instance, an SKU displaying over-forecasting during the first two months and under-forecasting in the last two months may result in the calculated bias value appearing insignificant despite significant deviations.
[0004] Various existing techniques attempt to address the limitations associated with fixed-period bias correction. For instance, heuristic-based methods incorporate manual adjustments to historical data to correct bias values. Said methods rely on subjective analysis and may fail to capture the dynamic behaviors of SKUs. The dependency on human judgment limits the accuracy of bias correction, particularly for SKUs exhibiting irregular demand patterns.
[0005] In another example, statistical techniques apply weighted averaging techniques to emphasize recent historical periods for bias correction. Said techniques attempt to account for variations in SKU behavior by assigning higher weights to recent historical data. However, such techniques operate under pre-defined assumptions, making them unsuitable for SKUs with non-linear demand fluctuations. Additionally, the reliance on statistical techniques without considering SKU-specific trends and historical data characteristics may result in suboptimal bias adjustments.
[0006] Other known techniques employ rules-based approaches to identify periods for bias correction. Said approaches define fixed thresholds for over-forecasting and under-forecasting patterns to determine forecasting errors. While rules-based approaches reduce subjectivity, said approaches lack adaptability to varying SKU demand patterns. For instance, SKU experiencing short-term seasonal demand may fail to meet predefined thresholds, leading to inaccurate bias detection and correction.
[0007] Machine learning models have been introduced to enhance bias correction accuracy by analyzing historical data. Said models utilize historical SKU data to identify trends and patterns for forecasting bias correction. However, conventional machine learning models often consider fixed historical periods without dynamically determining the optimal period for bias analysis. The reliance on pre-set periods limits the adaptability of machine learning models to SKU-specific demand variations.
[0008] In light of the above discussion, there exists an urgent need for solutions that overcome the problems associated with conventional systems and/or techniques for dynamically adjusting bias in SKU forecasting.
SUMMARY
[0009] The present disclosure provides a system to dynamically adjust a bias in stock keeping unit (SKU) forecasting. The system comprises a server having one or more processors that execute executable routines stored in a non-transitory storage device. The server imports historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU. The server analyses the imported historical sales data and historical forecast data to compute a bias value for each SKU. The server determines an optimal time period for bias adjustment based on the computed bias value by applying a machine learning technique. The server adjusts the imported future forecast data for each SKU based on the bias value of the determined optimal time period. A computing device operatively connected to the server receives the adjusted future forecast data through a transceiver and displays the received adjusted future forecast data on a display screen for user interaction.
[0010] In another aspect, the server analyses the imported historical sales data to classify the SKUs. The classification is based on factors such as sales trends, demand patterns, or sales volume. The classification enables customized forecasting strategies for distinct SKU categories, such as fast-moving, slow-moving, or seasonal SKUs.
[0011] In a further aspect, the server compares the imported historical sales data with the imported historical forecast data to compute the bias value. The server identifies over-forecasting patterns, where forecasted quantities exceed actual sales, and under-forecasting patterns, where forecasted quantities are lower than actual sales, for each SKU.
[0012] In another aspect, the server applies corrective actions based on the computed bias values. The corrective actions dynamically adjust forecast quantities to reduce discrepancies in future forecasting. Adjustments may increase or decrease forecasted quantities for upcoming time periods based on historical trends.
[0013] In another aspect, the server incorporates additional SKU attributes into the analysis of the bias values to refine the bias-adjusted forecasts. The additional attributes comprise on-hand inventory levels, stockouts, and replenishment frequency, which are integrated into the analysis to improve forecasting accuracy.
[0014] In another aspect, the additional SKU attributes are prioritised using a weighted analysis. The server assigns weights to each attribute based on their historical impact on demand patterns, enabling refinement of bias-adjusted forecasts for each SKU.
[0015] In another aspect, the server aligns the forecasted sales data for the pre-set time period with the imported historical sales data. The alignment process detects misalignments between forecasted and actual sales data, and corrections are applied to rectify the identified discrepancies.
[0016] In another aspect, the computing device receives a customization input to adjust the pre-set time period for forecasting. The customization input modifies the pre-set time period, allowing the server to recompute bias values and generate bias-adjusted forecasts based on the updated parameters. The adjusted forecasts are displayed on the computing device for user interaction.
[0017] In another aspect, the present disclosure provides a method for dynamically adjusting a bias in SKU forecasting. The method comprises importing historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU. The method further comprises analysing the imported data to compute a bias value, generating forecasted sales data, determining an optimal time period for bias adjustment, dynamically adjusting bias values, and adjusting future forecast data. Bias-adjusted forecast data is generated and transmitted to a computing device for user interaction.
BRIEF DESCRIPTION OF THE DRAWINGS
[0018] The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein.
[0019] Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams.
[0020] FIG. 1 illustrates a system to dynamically adjust bias in stock keeping unit (SKU) forecasting, in accordance with various implementations of the present disclosure;
[0021] FIG. 2 illustrates exemplary steps of a method for dynamically adjusting bias in SKU forecasting, in accordance with embodiments of the present disclosure;
[0022] FIG. 3 illustrates an interaction diagram for a system to dynamically adjust bias in SKU forecasting, in accordance with embodiments of the present disclosure; and
[0023] FIG. 4 illustrates an exemplary operational flow diagram for dynamically adjusting bias in SKU forecasting, in accordance with embodiments of the present disclosure.
DETAILED DESCRIPTION OF EMBODIMENTS
[0024] The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in art would recognize that other embodiments for carrying out or practicing the present disclosure are also possible.
[0025] As used herein, the term "import" refers to the process of retrieving or transferring data from an external source into a system or database. Such import involves accessing historical sales data, metadata, or related information stored in external systems, such as enterprise resource planning software, customer relationship management databases, or file storage systems. The process may utilize data exchange mechanisms such as application programming interfaces (APIs), data pipelines, or direct file uploads.
[0026] As used herein, the term "generate" refers to the act of creating or producing data, models, or outputs through computational processes. Such generation involves processing input data to produce a forecast, bias value, or other actionable information.
[0027] As used herein, the term "compute" refers to the process of performing calculations or operations to derive quantitative results. Such computation may involve applying mathematical formulas, statistical methods, or machine learning techniques to input data. For example, computing a bias value involves comparing forecasted sales data with actual sales data to quantify deviations such as over-forecasting or under-forecasting patterns for each SKU.
[0028] As used herein, the term "determine" refers to the act of identifying, establishing, or deciding a specific value, parameter, or condition based on data analysis or pre-defined criteria. Such determination may involve the application of machine learning models, statistical analysis, or decision-making algorithms. For instance, determining the optimal pre-set time period for bias adjustment involves analyzing historical sales patterns and applying predictive models to identify a time interval that minimizes forecasting errors.
[0029] As used herein, the term "adjust" refers to the act of modifying, altering, or refining data, parameters, or outputs to align with specific requirements or conditions. Such adjustment involves updating bias values, forecasts, or system settings dynamically based on real-time or pre-analyzed data. For example, adjusting bias values involves recalculating forecast data to correct systematic errors identified through historical data analysis. Similarly, adjusting operating parameters for an SKU makes sure the bias-adjusted forecasts are relevant for future sales scenarios.
[0030] As used herein, the term "classify" refers to the process of grouping or categorizing data, objects, or entities based on shared characteristics or attributes. Such classification is performed using predefined criteria or machine learning models. For example, classifying SKUs may involve grouping them into categories such as high-demand, seasonal, or low-demand based on sales trends and inventory turnover rates.
[0031] As used herein, the term "prioritize" refers to the process of ranking or assigning importance to data, attributes, or tasks based on predefined criteria or calculated impact. Such prioritization may involve applying weighting factors or decision-making frameworks
[0032] FIG. 1 illustrates a system 100 to dynamically adjust bias in stock keeping unit (SKU) forecasting, in accordance with various implementations of the present disclosure. The system 100 comprises a server 102 that comprises one or more processors, with said processors operatively associated with a non-transitory storage device. Said non-transitory storage device stores one or more executable routines that allow server 102 to perform operations essential for dynamically adjusting bias in stock-keeping unit (SKU) forecasting. Said routines allow importing historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU. Said historical sales data comprises past transaction records representing actual sales. Said historical forecast data consists of previously predicted values corresponding to said historical sales data, while said future forecast data represents anticipated sales figures for an upcoming period. Said routines process said data to assure compatibility, normalization, and completeness. Said imported data undergoes systematic validation to address missing or erroneous entries and assure readiness for subsequent operations.
[0033] In an embodiment, said system 100 processes said historical sales data and historical forecast data using server 102. Said processing comprises analyzing said historical sales data against said historical forecast data to identify discrepancies. Said discrepancies are quantified as bias values for each SKU. Said bias value represents the extent of deviation between actual sales and forecasted values. Said routines employ statistical measures, such as mean absolute error or standard deviation, to calculate said bias value. Said routines classify SKUs based on forecast accuracy, grouping them for further adjustments. Said analysis phase identifies patterns of over-forecasting or under-forecasting, enabling targeted recalibration of said future forecast data.
[0034] In an embodiment, said system 100 determines an optimal time period for bias adjustment for each SKU using server 102. Said determination is performed by processing said bias values through a machine learning technique stored in said server 102. Said machine learning technique analyzes said historical sales data, said historical forecast data, and said computed bias values to identify temporal patterns indicative of demand fluctuations or seasonality. Said determination involves evaluating data across multiple time intervals to select a time period that aligns with observed demand trends for each SKU. Said routines incorporate SKU-specific attributes, such as replenishment frequency, sales volume, and demand variability, into said analysis. Said determination accounts for both internal and external factors influencing forecast reliability, assuring that said adjustments are performed at an optimal time period for each SKU.
[0035] In an embodiment, said system 100 adjusts said imported future forecast data for each SKU using said computed bias values and said determined optimal time period. Said adjustment is performed by recalibrating forecasted values to account for historical discrepancies identified in prior analysis. Said adjustment process integrates said computed bias value as a corrective factor and modifies said future forecast data to align with real-world demand patterns. Said routines utilize statistical methods to apply corrections while preserving SKU-specific characteristics. Said adjustments compensate for historical over-forecasting or under-forecasting patterns, providing recalibrated forecast data for each SKU. Said optimal time period serves as a reference point for performing said adjustments, assuring that said recalibration aligns with both temporal and SKU-specific attributes.
[0036] In an embodiment, said system 100 comprises a computing device 104 operatively associated with server 102. Said computing device 104 receives said adjusted forecast data from server 102 through a transceiver. Said computing device 104 processes said adjusted forecast data and presents said data on a display screen. Said display screen allows users to review, analyze, and interact with said recalibrated forecast data. Said computing device 104 supports additional functionalities, such as filtering and sorting options, for detailed examination of said data. Said transceiver facilitates secure transmission of said adjusted forecast data, maintaining integrity and reliability during communication between said server 102 and said computing device 104. Said computing device 104 also supports user inputs, including modifications to said pre-set time period, enabling customization of bias adjustments for specific SKUs. Said customization inputs are processed by server 102, allowing dynamic recalibration based on operational requirements.
[0037] In an embodiment, said routines executed by server 102 incorporate additional SKU attributes into said bias adjustment process. Said attributes comprise on-hand inventory levels, stockout occurrences, replenishment frequency, and other SKU-specific characteristics. Said routines analyze said attributes alongside said historical sales data, said historical forecast data, and said future forecast data to enable recalibration of said forecast data. Said analysis evaluates said attributes using weighted parameters, prioritizing their impact on forecast accuracy. Said additional attributes enhance the granularity of said bias adjustment process, enabling corrections customizations to each SKU.
[0038] In an embodiment, said system 100 detects misalignments between said historical sales data and said historical forecast data during said bias adjustment process. Said routines analyze alignment between said datasets to identify significant deviations where said historical forecast data diverges from said historical sales data. Said misalignments are corrected by adjusting said historical forecast data to more accurately represent said historical sales data. Said correction process is important for establishing reliable input data for subsequent bias computations. Said routines address said misalignments systematically to enable consistency and accuracy throughout said bias adjustment process, maintaining the integrity of said recalibrated forecast data.
[0039] In an embodiment, said system 100 enables a computing device 104 to receive customization inputs for modifying said pre-set time period. Said customization inputs are processed by server 102, which adjusts said bias computation and said recalibrated forecast data accordingly. Said customization allows users to refine said bias adjustments based on specific operational needs or changing market conditions. Said customization process integrates with said routines, maintaining consistency across all stages of said data processing. Said adjustments are dynamically applied to make sure that said recalibrated forecast data aligns with user-defined parameters while adhering to historical and SKU-specific characteristics.
[0040] In an embodiment, the computing device 104 receives adjusted future forecast data from the server 102 through a transceiver. The adjusted future forecast data is generated after the server 102 processes historical sales data, historical forecast data, and future forecast data to calculate bias values for each stock keeping unit (SKU). The computing device 104 facilitates the communication link between the server 102 and the user by transmitting and displaying the adjusted future forecast data. The transceiver within the computing device 104 establishes a communication pathway with the server 102 to affirm reliable data transfer without introducing delays. The computing device 104 translates the data into a format suitable for display on a display screen, allowing users to interact with the adjusted data efficiently.
[0041] In an embodiment, the computing device 104 incorporates components to handle the reception and display of adjusted future forecast data. The transceiver embedded within the computing device 104 processes incoming signals from the server 102, converting the transmitted data packets into usable information. The transceiver operates within defined parameters to maintain compatibility with the server 102 while minimizing disruptions during data exchange. The processing circuitry within the computing device 104 synchronizes with the transceiver to interpret the adjusted future forecast data. Such processing circuitry operates in conjunction with a memory unit to store temporary data related to the adjusted forecasts, enabling accurate translation of the received information.
[0042] In an embodiment, the display screen of the computing device 104 is configured to present the adjusted future forecast data visually, providing a clear interface for user interaction. The display screen integrates advanced graphical rendering capabilities to accurately depict forecast adjustments made by the server 102. The screen supports various formats of visual representation, including charts, graphs, or tables, as per processed data. The computing device 104 aligns the displayed data with user preferences, permitting interaction with the displayed forecasts to modify time periods or explore additional SKU attributes. Through such configurations, the computing device 104 facilitates an efficient interaction interface for understanding forecast adjustments transmitted by the server 102.
[0043] In an aspect, a retail company managing multiple stock keeping units (SKUs) utilizes the system 100 to dynamically adjust bias in SKU forecasting. The server 102 imports historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU. The historical sales data comprises recorded sales quantities, while the historical forecast data comprises prior predictions for the same pre-set time period. For example, the historical sales data and the corresponding historical forecast data for SKU-001 for six months are shown in the tables (table. 1 and table. 2) below.

Month Historical Sales
SKU-001
January 100

February 110
March 120
April 90
May 95
June 115
Table. 1
Month Historical Forecast
SKU-001
January 120
February 130
March 140
April 110
May 100
June 130
Table. 2

Month Future Forecast
SKU-001
July 125
August 130
September 135
Table. 3
[0044] The server 102 analyses the imported historical sales data and historical forecast data for each SKU to compute bias values. The bias values represent discrepancies between the forecasted and actual sales quantities over the pre-set time period. Bias values are calculated as the difference between historical forecast data and historical sales data. For instance, the bias values for each SKU over six months are provided in the table. 4 below.

Month Bias Values
SKU-001
January 20
February 20
March 20
April 20
May 5
June 15
Table. 4
[0045] The server 102 processes the computed bias values to determine an optimal pre-set time period for bias adjustment for SKU-001. The server 102 applies a machine learning technique to analyse trends and patterns in the bias values across different pre-set time periods. The machine learning technique evaluates factors such as demand trends and the magnitude of forecast deviations to identify a time period that minimizes forecasting errors. For SKU-001, the server 102 determines that the optimal pre-set time period is four months, as shown in the table. 5 below.
SKU Initial Time Period Optimal Time Period
SKU-001 6 Months 4 Months
Table. 5
[0046] Using the determined optimal pre-set time period, the server 102 dynamically adjusts the bias values for SKU-001. Such adjustment recalibrates discrepancies in the forecast data to reduce errors. The adjusted bias values for SKU-001 for selected months are shown in the table. 6 below. Negative values indicate underestimation in forecasted sales for the respective months.

Month Adjusted Bias
January 10
February 15
March 10
April 15
Table. 6
[0047] The server 102 uses the adjusted bias values to generate bias-adjusted forecasts for future periods for SKU-001. Such bias-adjusted forecasts reflect recalibrations made to improve prediction accuracy. The bias-adjusted forecasts for SKU-001 for the months of July, August, and September are displayed below.
Month Bias-Adjusted Forecast
July 110
August 115
September 120
Table. 7
[0048] The computing device 104 receives the bias-adjusted forecast data for SKU-001 from the server 102 through a transceiver and displays such data on a display screen. The display screen provides a visual representation of the adjusted forecasts for user interaction. Users can analyse the forecast data for planning purposes and refine operational strategies based on SKU-001-related attributes. The server 102 and the computing device 104 collectively enable improved forecasting for inventory and sales management specific to SKU-001.
[0049] In an embodiment, the server 102 may analyse the imported historical sales data to classify the stock keeping units (SKUs). The server 102 applies classification techniques to segment the SKUs based on specific criteria, such as sales volume, demand frequency, and seasonal trends. The historical sales data, which comprises recorded sales quantities over a defined period, is initially pre-processed to filter out incomplete or anomalous data points. After pre-processing, the server 102 evaluates the data using clustering or classification techniques to categorise the SKUs into distinct groups. The categories may comprise fast-moving SKUs, slow-moving SKUs, and seasonal SKUs, depending on their observed sales patterns. For fast-moving SKUs, the server 102 identifies consistent, high-volume sales trends, while slow-moving SKUs are characterised by low or infrequent sales activity. Seasonal SKUs are identified by evaluating periodic spikes in demand. The classification enables the system 100 to adopt forecasting approaches for each SKU category. For example, fast-moving SKUs may require shorter forecasting intervals to meet rapid demand changes, while slow-moving SKUs may benefit from broader intervals to optimise inventory management. Seasonal SKUs are subjected to additional analysis to identify peak and off-peak demand periods for more accurate planning. The classification results are stored in a database and used to refine subsequent processes, such as bias computation, bias adjustment, and the generation of bias-adjusted forecasts.
[0050] In an embodiment, the server 102 compares the imported historical sales data with the imported historical forecast data to compute the bias value for each SKU. The comparison identifies discrepancies between forecasted and actual sales quantities over a pre-set time period. The server 102 calculates the bias value as the difference between the historical forecast data and the corresponding historical sales data. Positive bias values indicate over-forecasting, where the forecasted sales quantities exceed the actual sales, while negative bias values denote under-forecasting, where the forecasted quantities are lower than actual sales. The server 102 evaluates the bias values across the pre-set time period to detect trends, such as consistent over-forecasting or under-forecasting patterns for specific SKUs. The server 102 also analyses the magnitude of the bias values to identify SKUs with significant forecast inaccuracies. For instance, SKUs with recurring over-forecasting patterns may be re-evaluated to identify causes, such as changes in demand dynamics or external factors influencing sales. Similarly, SKUs with under-forecasting patterns may indicate underestimated demand, requiring upward adjustments in future forecasts. The server 102 further identifies specific time periods within the pre-set range where deviations are most pronounced, enabling targeted corrections. The bias values and associated patterns are stored in a database for use in subsequent processes, such as determining optimal pre-set time periods and generating bias-adjusted forecasts.
[0051] In an embodiment, the server 102 applies corrective actions based on the computed bias values to address forecasting discrepancies for each SKU. The server 102 determines the appropriate corrective measures by analysing the magnitude and trends of the bias values across the pre-set time period. For SKUs exhibiting over-forecasting, the server 102 reduces future forecast quantities proportionally to the computed bias values, assuring that the adjusted forecasts align more closely with observed demand patterns. Conversely, for SKUs displaying under-forecasting trends, the server 102 increases future forecast quantities to account for underestimated demand. The server 102 dynamically adjusts the corrective measures based on ongoing data analysis, incorporating factors such as recent sales fluctuations and changes in market conditions. The recalibrated forecasts are generated as bias-adjusted forecasts, which reflect the impact of the corrective actions on future periods. The server 102 stores the adjusted forecast data in a database and transmits said adjusted forecast data to the computing device 104 for user interaction. The corrective actions also comprise provisions for handling SKUs with irregular demand patterns, where historical bias values alone may not fully capture variability. In such cases, the server 102 employs additional metrics, such as demand volatility, to refine the adjustments. By systematically applying corrective actions, the server 102 makes sure that the bias-adjusted forecasts improve planning accuracy for upcoming time periods. The recalibrated forecasts provide a reliable basis for inventory management, procurement planning, and demand fulfilment, effectively addressing the discrepancies identified in the bias values.
[0052] In an embodiment, the server 102 incorporates additional SKU attributes into the analysis of the bias values to refine the bias-adjusted forecasts. The additional attributes comprise on-hand inventory levels, stockouts, and replenishment frequency. The server 102 imports data corresponding to these attributes from connected databases or inventory management systems. Once imported, the server 102 integrates the additional attributes with the historical sales data and computed bias values to create a dataset. The analysis considers the relationships between the additional attributes and historical sales trends, quantifying the impact of each attribute on demand patterns. For example, frequent stockouts in historical data may suggest unmet demand, prompting the server 102 to adjust future forecast quantities upward to accommodate such demand. Similarly, high on-hand inventory levels may indicate surplus stock, leading to downward adjustments in forecasts to prevent overstocking. The replenishment frequency data is used to align forecasts with procurement cycles. The server 102 employs statistical techniques to evaluate the interactions between the additional attributes and sales data, affirming that the bias-adjusted forecasts reflect the combined influence of multiple factors. The server 102 stores the refined forecasts in a database and transmits them to the computing device 104 for display.
[0053] In an embodiment, the server 102 prioritises additional SKU attributes using a weighted analysis to refine bias-adjusted forecasts. The server 102 assigns weights to each attribute based on predefined criteria, such as historical correlations with sales trends and their significance in driving forecast accuracy. For example, on-hand inventory levels may receive higher weights for SKUs with irregular demand patterns, as such levels provide insights into inventory sufficiency relative to observed sales trends. Stockouts, on the other hand, may be prioritised for SKUs with consistent high demand, where the absence of inventory directly impacts sales. The server 102 dynamically updates the assigned weights based on incoming data to reflect changes in SKU behaviour or market conditions. The weighted analysis quantifies the relative importance of each attribute and incorporates said evaluation into the refinement of bias-adjusted forecasts. For instance, a SKU with high replenishment frequency and moderate on-hand inventory levels may require adjustments to affirm forecast alignment with procurement cycles. The server 102 uses the prioritised attributes to fine-tune forecast adjustments, affirming that the recalibrations address the most influential factors for each SKU. The refined forecasts are stored in a database and made available for user interaction via the computing device 104.
[0054] In an embodiment, the server 102 aligns the forecasted sales data for the pre-set time period with the imported historical sales data to detect a misalignment and correct the detected misalignment. The server 102 first imports the historical sales data and forecasted sales data into a processing unit. The alignment process involves mapping the data points from the historical sales data to their corresponding entries in the forecasted sales data for the same time period. The server 102 identifies discrepancies between the two datasets by calculating differences at each data point. For instance, a forecasted sales data point that deviates significantly from the corresponding historical sales data point indicates a misalignment. The server 102 analyses the identified misalignments to determine causes, such as shifts in demand trends, inaccurate forecasting assumptions, or external factors influencing sales. Once the misalignments are detected, the server 102 corrects the forecasted sales data by applying adjustments based on historical trends or recalculating projections for the affected time period. The correction process involves recalibrating forecast values to reflect the patterns observed in the historical sales data. The server 102 stores the corrected forecast data in a database, where said corrected forecast data is integrated into subsequent forecasting processes. By aligning and correcting the forecasted sales data, the server 102 improves the accuracy of bias computations and bias-adjusted forecasts for each SKU. The corrected data is subsequently transmitted to the computing device 104 for user interaction and analysis.
[0055] In an embodiment, the computing device 104 receives a customization input to adjust the pre-set time period to generate the bias-adjusted forecasts for each SKU. The computing device 104 comprises an interface that displays forecast data received from the server 102. The interface presents options for users to modify the pre-set time period, such as by extending or reducing the duration of the forecast. The computing device 104 processes the customization input to reconfigure the pre-set time period and communicates the adjusted parameters to the server 102. Upon receiving the updated pre-set time period, the server 102 reanalyses the historical sales data, historical forecast data, and any other relevant inputs to recompute bias values. The server 102 generates updated bias-adjusted forecasts that reflect the customization input provided by the computing device 104. The computing device 104 retrieves the updated forecasts from the server 102 and visually represents the data on a display screen. The displayed data incorporates the recalibrated forecast values, enabling users to evaluate the effects of the customized time period on the bias-adjusted forecasts. The computing device 104 also allows users to make iterative adjustments to the pre-set time period, assuring that the forecasting process remains dynamic and responsive to user preferences. The updated data is stored in a database and remains accessible for subsequent analysis or operational planning.
[0056] The present disclosure discloses a system and method for communication between a server and a computing device, wherein the server comprises one or more processors configured to execute one or more executable routines stored in a non-transitory storage device. The executable routines are operable to generate bias-adjusted forecasts for transmission to a computing device.
[0057] The computing device, which may be a smartphone, tablet, or similar portable device, comprises a transceiver configured to receive bias-adjusted forecasts from the server. The server transmits the bias-adjusted forecasts to the computing device over a communication network. Such communication may be implemented using wireless protocols, including but not limited to Wi-Fi, LTE, 5G, or Bluetooth, or wired protocols such as Ethernet or USB. The server formats the data in a manner compatible with the transceiver of the computing device, enabling reliable and efficient data transfer.
[0058] Upon receipt of the transmitted data, the transceiver of the computing device processes the received signals to extract the bias-adjusted forecast data. The computing device comprises a display screen operatively connected to the internal processing components, enabling the rendering of the bias-adjusted forecasts for viewing by a user. The forecasts may be displayed in graphical, textual, or other user-intuitive formats, providing actionable insights in real time or near-real time.
[0059] The system facilitates secure and seamless communication between the server and computing device by employing encryption protocols or authentication mechanisms to prevent unauthorized access to the transmitted data.
[0060] FIG. 2 illustrates exemplary steps of a method 200 for dynamically adjusting bias in SKU forecasting, in accordance with embodiments of the present disclosure. At step 202, the one or more processors of the server 102 import historical sales data, historical forecast data, and future forecast data for each stock keeping unit (SKU) from a non-transitory storage device. The historical sales data comprises recorded sales quantities for a pre-set time period, while the historical forecast data contains previously predicted sales quantities for the same period. The future forecast data represents sales projections for upcoming periods. The imported data is temporarily stored in server memory for subsequent processing.
[0061] At step 204, the one or more processors analyze the imported historical sales data and historical forecast data for each SKU to compute a bias value. The analysis involves comparing forecasted sales quantities with corresponding actual sales quantities for each data point in the pre-set time period. The bias value is calculated as the difference between the forecasted and actual quantities, where positive values indicate over-forecasting and negative values indicate under-forecasting. The computed bias values are stored in a database for further analysis.
[0062] At step 206, the one or more processors determine an optimal time period for bias adjustment for each SKU based on the computed bias values by applying a machine learning technique. The determination process evaluates trends in the bias values over varying time periods to identify a duration that minimizes forecast inaccuracies. The optimal time period for each SKU is stored in the server for use in subsequent processing steps.
[0063] At step 208, the one or more processors adjust the imported future forecast data for each SKU based on the bias values of the determined optimal time period. The adjustment recalibrates the future forecast quantities to align with the observed trends in the bias values. Over-forecasted quantities are reduced proportionally, while under-forecasted quantities are increased to reflect the adjusted bias values.
[0064] At step 210, the one or more processors generate adjusted future forecast data for each SKU. The adjusted future forecast data incorporates the recalibrations made in the previous step to address inaccuracies observed in the bias values. The generated adjusted forecasts are prepared for transmission to the computing device 104 for user review and interaction.
[0065] At step 212, the one or more processors transmit the adjusted future forecast data to the computing device 104 through a transceiver. The transmission occurs over a secure communication link, making sure that the adjusted data is reliably delivered to the computing device 104 for display and interaction. The server 102 maintains a copy of the transmitted data for record-keeping and audit purposes.
[0066] At step 214, the computing device 104 receives the adjusted future forecast data transmitted by the server. The computing device 102 processes the received data and prepares said data for visualization on a display screen. The received data is stored temporarily in the memory of computing device 104 to facilitate interactive operations initiated by the user.
[0067] At step 216, the computing device 104 displays the received adjusted future forecast data on a display screen for user interaction. The display interface presents the adjusted forecasts in a visually interpretable format, such as tables or graphs, allowing users to analyze the recalibrated forecasts and make informed decisions about inventory and sales planning for each SKU.
[0068] FIG. 3 illustrates an interaction diagram for a system 100 to dynamically adjust bias in SKU forecasting, in accordance with embodiments of the present disclosure. The system 100 comprises a user, a computing device 104, and a server 102. The process begins with the user initiating a forecast adjustment request through the computing device 104. The computing device 104 transmits the request to the server 102, which processes the request to dynamically adjust the forecast. Upon receiving the request, the server 102 imports data from external sources, including historical sales data, historical forecast data, and future forecast data for the pre-set time period. The server 102 analyses the imported data to compute bias values for each SKU by identifying discrepancies between historical sales data and historical forecast data. Using the computed bias values, the server 102 applies a machine learning technique to determine an optimal time period for bias adjustment for each SKU. The server 102 then adjusts the future forecast data for each SKU based on the computed bias values and the determined optimal time period. The adjusted future forecast data is returned to the computing device 104, which displays the received adjusted forecast data on a display screen for user interaction and analysis.
[0069] FIG. 4 illustrates an exemplary operational flow diagram for dynamically adjusting bias in SKU forecasting, in accordance with embodiments of the present disclosure. The process begins with analyzing historical sales data and forecasted sales data for multiple SKUs across different categories over a specific time period. For instance, in the M-1 dataset, SKU1 under Category A has a historical sales value of 100 and a forecasted sales value of 140, resulting in a bias of 1.4. SKU2 under Category B, with a historical sales value of 100 and a forecasted sales value of 70, has a bias of 0.7. Similarly, the M-2 dataset comprises historical and forecasted sales data for Category A and Category B, where SKU1 under Category A exhibits a bias of 1.25, and SKU2 under Category B has a bias of 0.68. Analysis of additional datasets, such as M-n, continues to capture SKU-specific biases across multiple periods. The system processes said biases to determine the optimal number of months required for bias correction for each category. For example, Category A requires three months, while Category B requires four months for bias adjustment. The system dynamically applies bias correction to adjust forecasted sales values, generating bias-adjusted forecasts. For instance, the forecasted value for SKU1 under Category A is adjusted from 100 to 120, while SKU2 under Category B is adjusted from 100 to 95. Said bias corrections account for SKU-specific deviations and sales behaviors, producing more reliable bias-adjusted forecasts.
 

CLAIMS
What is claimed is:
1. A system to dynamically adjust a bias in a stock keeping unit (SKU) forecasting, comprising:
a server, wherein the server comprises one or more processors which execute one or more executable routines, which are stored in a non-transitory storage device;
the one or more routines allow to:
import, the historical sales data, historical forecast data, future forecast data for a pre-set time period, for each SKU;
analyze, for each SKU, the imported historical sales data and the historical forecast data to compute a bias value;
determine, an optimal time period based on the computed bias value, for each SKU, by applying a machine learning technique;
adjust, the imported future forecast data, for each SKU, based on the computed bias-value of the determined optimal time period;
a computing device operatively connected to the server, wherein the computing device is configured to:
receive, through a transceiver, the adjusted future forecast data from the server; and
display, on a display screen, the received adjusted future forecast data for user interaction.
2. The system of claim 1, wherein the server is configured to analyze the imported historical sales data to classify the SKUs.
3. The system of claim 1, wherein the server is configured to compare the imported historical sales data with the imported historical forecast data to compute the bias value by determining the over-forecasting patterns and the under-forecasting patterns for each SKU.
4. The system of claim 1, wherein the server is configured to apply corrective actions based on the computed bias values.
5. The system of claim 1, wherein the server is configured to incorporate additional SKU attributes, including the on-hand inventory levels, stockouts, and replenishment frequency, into the analysis of the bias values to refine the bias-adjusted forecasts.
6. The system of claim 5, wherein the additional SKU attributes are prioritized using a weighted analysis based on their impact on the bias-adjusted forecasts.
7. The system of claim 1, wherein the server aligns the forecasted sales data for the pre-set time period with the imported historical sales data to detect a misalignment and correct the detected misalignment.
8. The system of claim 1, wherein the computing device receives a customization input to adjust the pre-set time period to generate the bias-adjusted forecasts for each SKU.
9. A method for dynamically adjusting a bias in stock keeping unit (SKU) forecasting, the method comprising:
importing, by one or more processors of a server, historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU from a non-transitory storage device;
analyzing, by the one or more processors, the imported historical sales data and historical forecast data for each SKU to compute a bias value;
determining, by the one or more processors, an optimal time period for bias adjustment for each SKU based on the computed bias value by applying a machine learning technique;
adjusting, by the one or more processors, the imported future forecast data for each SKU based on the bias value of the determined optimal time period;
generating, by the one or more processors, adjusted future forecast data for each SKU based on the computed and adjusted bias values;
transmitting, by the one or more processors, the adjusted future forecast data to a computing device through a transceiver;
receiving, by the computing device, the adjusted future forecast data transmitted by the server; and
displaying, by the computing device, the received adjusted future forecast data on a display screen for user interaction.

SYSTEM AND METHOD TO DYNAMICALLY ADJUST BIAS IN STOCK KEEPING UNIT FORECASTING
ABSTRACT
The present disclosure provides a system to dynamically adjust bias in stock keeping unit (SKU) forecasting. The system comprises a server and a computing device. The server imports historical sales data for a pre-set time period for each SKU, generates forecasted sales data for each SKU to compute a bias value, determines an optimal pre-set time period for bias adjustment using a machine learning technique, and dynamically adjusts the bias values based on the determined optimal pre-set time period. The server further generates bias-adjusted forecasts for future periods. The computing device receives the bias-adjusted forecasts from the server and displays said forecasts for user interaction.
FIG. 1

, C , Claims:CLAIMS
What is claimed is:
1. A system to dynamically adjust a bias in a stock keeping unit (SKU) forecasting, comprising:
a server, wherein the server comprises one or more processors which execute one or more executable routines, which are stored in a non-transitory storage device;
the one or more routines allow to:
import, the historical sales data, historical forecast data, future forecast data for a pre-set time period, for each SKU;
analyze, for each SKU, the imported historical sales data and the historical forecast data to compute a bias value;
determine, an optimal time period based on the computed bias value, for each SKU, by applying a machine learning technique;
adjust, the imported future forecast data, for each SKU, based on the computed bias-value of the determined optimal time period;
a computing device operatively connected to the server, wherein the computing device is configured to:
receive, through a transceiver, the adjusted future forecast data from the server; and
display, on a display screen, the received adjusted future forecast data for user interaction.
2. The system of claim 1, wherein the server is configured to analyze the imported historical sales data to classify the SKUs.
3. The system of claim 1, wherein the server is configured to compare the imported historical sales data with the imported historical forecast data to compute the bias value by determining the over-forecasting patterns and the under-forecasting patterns for each SKU.
4. The system of claim 1, wherein the server is configured to apply corrective actions based on the computed bias values.
5. The system of claim 1, wherein the server is configured to incorporate additional SKU attributes, including the on-hand inventory levels, stockouts, and replenishment frequency, into the analysis of the bias values to refine the bias-adjusted forecasts.
6. The system of claim 5, wherein the additional SKU attributes are prioritized using a weighted analysis based on their impact on the bias-adjusted forecasts.
7. The system of claim 1, wherein the server aligns the forecasted sales data for the pre-set time period with the imported historical sales data to detect a misalignment and correct the detected misalignment.
8. The system of claim 1, wherein the computing device receives a customization input to adjust the pre-set time period to generate the bias-adjusted forecasts for each SKU.
9. A method for dynamically adjusting a bias in stock keeping unit (SKU) forecasting, the method comprising:
importing, by one or more processors of a server, historical sales data, historical forecast data, and future forecast data for a pre-set time period for each SKU from a non-transitory storage device;
analyzing, by the one or more processors, the imported historical sales data and historical forecast data for each SKU to compute a bias value;
determining, by the one or more processors, an optimal time period for bias adjustment for each SKU based on the computed bias value by applying a machine learning technique;
adjusting, by the one or more processors, the imported future forecast data for each SKU based on the bias value of the determined optimal time period;
generating, by the one or more processors, adjusted future forecast data for each SKU based on the computed and adjusted bias values;
transmitting, by the one or more processors, the adjusted future forecast data to a computing device through a transceiver;
receiving, by the computing device, the adjusted future forecast data transmitted by the server; and
displaying, by the computing device, the received adjusted future forecast data on a display screen for user interaction.