{"id":2310,"date":"2023-05-30T11:23:28","date_gmt":"2023-05-30T05:53:28","guid":{"rendered":"https:\/\/blogs.infosys.com\/infosys-cobalt\/?p=2310"},"modified":"2023-06-01T15:05:30","modified_gmt":"2023-06-01T09:35:30","slug":"forecasting-with-accuracy-in-todays-world-clone","status":"publish","type":"post","link":"https:\/\/blogs.infosys.com\/infosys-cobalt\/cloud-applications\/oracle\/forecasting-with-accuracy-in-todays-world-clone.html","title":{"rendered":"Forecasting with Accuracy in Today’s World"},"content":{"rendered":"

Forecasting in Today\u2019s World<\/strong><\/p>\n

Frequent volatility in marketplace in the post pandemic era caused by multiple events leading from war in Ukraine to inflation and high interest rates globally have had resultant impacts on businesses around the world. The short lead times these fluctuations have offered to businesses to adjust their business plans have made it even more challenging for them to calibrate changes in planning, forecasting and resultant business actions.<\/p>\n

Enter Anaplan: A platform that allows flexible and nimble forecasting, moving away from rigid structures enforced by transaction processing systems<\/p>\n

\u2022 Enables truly connected planning, with the ability to start from one business area: say Sales Planning adding others: Demand Planning, Capacity Planning: FP&A Revenue &; OpEx Planning, etc.
\n\u2022 Allows customizable multi-dimensional hierarchies.
\n\u2022 Low\/Simple Coding
\n\u2022 Intuitive interactive Dashboards \u2013 forecasts on the fly
\n\u2022 Stakeholder collaboration
\n\u2022 Real-time integration with ERPs<\/p>\n

Defined Models in Anaplan maintain data in prescribed modules enabling generation of forecasts in Time Series or PlanIQ, a Machine Learning (ML) based platform to generate forecasts. These can be generated across Supply Chain, Human Resources & Finance at the lowest level of measure required.<\/p>\n

PlanIQ offers multiple forecast outputs using algorithms leveraging ML. It is a native platform within Anaplan offering that combines the unique flexibility and in-memory capabilities of Anaplan with the advanced forecasting analytics to deliver high forecasts with minimal setup and process time.<\/p>\n

Known PlanIQ solutions where it has works well are:
\n1. Supply Chain: Demand Sensing, Demand planning, Demand Management
\n2. Finance: CapEx planning, OpEx planning
\n3. Workforce: Workforce planning, Attrition planning<\/p>\n

To take the case of Demand Planning, following would be the derivation of Demand Plan, for a Manufacturing Company<\/p>\n

Data Organization<\/strong><\/p>\n\n\n\n\n\n\n\n\n\n\n
Sales History for the past x years<\/td>\n<\/tr>\n
Data Normalized \u2013 outliers removed, trade promotions, extra-ordinary events, and occurrences<\/td>\n<\/tr>\n
Statistical normalization: UCL, LCL, USL, LSL, Standard Deviation, etc. applied<\/td>\n<\/tr>\n
Run Time Series algorithms\u2026Anaplan provides downloadable models in AppHub and 17 results derived<\/td>\n<\/tr>\n
Run PlanIQ ML forecasting (explained below)<\/td>\n<\/tr>\n
System suggested Best Fit Forecast based on MAPE% closest to 0<\/td>\n<\/tr>\n
Over-ride of best-fit by Planners, if needed<\/td>\n<\/tr>\n
Unconstrained Demand Plan derived<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
Anaplan processes:<\/strong><\/p>\n
– Sales History and periodic updates are interfaced from ERP\/Transaction Processing systems into Anaplan at defined frequency into the Data Hub Model
\n– Downloadable Models are available in AppHub \u2013 a repository of pre-built customized apps that have Time Series Statistical Forecasting Apps
\n– AppHub Models downloaded need to be modified and within those data is appropriately transformed and loaded from the Data Hub. This data is normalized to remove outliers, statistical normalization applied and trends, seasonality applied.
\n– To apply Statistical Normalization and Trends & Seasonality statistical formulae can be built into Anaplan with excel-like functions available.
\n– Run Time-Series measures to derive results from algorithms reflected below (in blue)
\n– Run PlanIQ ML to derive results as shown below (in green)<\/p>\n
The Statistical measures and their application are explained below. Data below is illustrative:<\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Algorithm<\/strong><\/td>\n MAPE%<\/strong><\/p>\n
(Mean Absolute Percentage Error)<\/td>\n
RMSE<\/strong><\/p>\n
(Root Mean Square Error)<\/td>\n
MAD<\/strong><\/td>\n BIAS%<\/strong><\/p>\n
\u00a0<\/strong><\/td>\n<\/tr>\n
Amazon Auto ML Forecast<\/td>\n 2%<\/strong><\/td>\n 482<\/td>\n 415<\/td>\n 30%<\/td>\n<\/tr>\n
Anaplan Auto ML Forecast<\/td>\n 12%<\/td>\n 468<\/td>\n 390<\/td>\n 50%<\/td>\n<\/tr>\n
Prophet Forecast<\/td>\n 14.3%<\/td>\n 640<\/td>\n 350<\/td>\n 40%<\/td>\n<\/tr>\n
ARIMA Forecast<\/td>\n 8.3%<\/td>\n 680<\/td>\n 450<\/td>\n 40%<\/td>\n<\/tr>\n
DeepAR+ Forecast<\/td>\n 14.5%<\/td>\n 900<\/td>\n 750<\/td>\n 20%<\/td>\n<\/tr>\n
CNN-QR Forecast<\/td>\n 3.8%<\/td>\n 750<\/td>\n 650<\/td>\n 25%<\/td>\n<\/tr>\n
Linear Regression<\/td>\n 22.6%<\/td>\n 340<\/td>\n 300<\/td>\n 7%<\/td>\n<\/tr>\n
Logarithmic Regression<\/td>\n 24%<\/td>\n 330<\/td>\n 250<\/td>\n 6.5%<\/td>\n<\/tr>\n
Exponential Regression<\/td>\n 24.8%<\/td>\n 340<\/td>\n 300<\/td>\n 6%<\/td>\n<\/tr>\n
Power Regression<\/td>\n 8%<\/td>\n 200<\/td>\n 280<\/td>\n 4%<\/td>\n<\/tr>\n
Moving Average<\/td>\n 5%<\/td>\n 220<\/td>\n 120<\/td>\n 4.2%<\/td>\n<\/tr>\n
Double Moving Average<\/td>\n 8%<\/td>\n 300<\/td>\n 150<\/td>\n 5%<\/td>\n<\/tr>\n
Simple Exponential Smoothing<\/td>\n 5.5%<\/td>\n 150<\/td>\n 130<\/td>\n 5%<\/td>\n<\/tr>\n
Double Exponential Smoothing<\/td>\n 7%<\/td>\n 160<\/td>\n 120<\/td>\n 5.5%<\/td>\n<\/tr>\n
Triple Exponential Smoothing<\/td>\n 8%<\/td>\n 200<\/td>\n 150<\/td>\n 6%<\/td>\n<\/tr>\n
Additive Decomposition<\/td>\n 40%<\/td>\n 1,000<\/td>\n 1100<\/td>\n 30%<\/td>\n<\/tr>\n
Multiplicative Decomposition<\/td>\n 5%<\/td>\n 130<\/td>\n 100<\/td>\n 4%<\/td>\n<\/tr>\n
Multi Decomp Logarithmic<\/td>\n 6%<\/td>\n 142<\/td>\n 120<\/td>\n 4.5%<\/td>\n<\/tr>\n
Multi Decomp Exponential<\/td>\n 5%<\/td>\n 130<\/td>\n 130<\/td>\n 5%<\/td>\n<\/tr>\n
Multi Decomp Power<\/td>\n 6%<\/td>\n 140<\/td>\n 140<\/td>\n 5%<\/td>\n<\/tr>\n
Winter\u2019s Additive<\/td>\n 6.5%<\/td>\n 170<\/td>\n 160<\/td>\n 5.2%<\/td>\n<\/tr>\n
Winter\u2019s Multiplicative<\/td>\n 7%<\/td>\n 200<\/td>\n 140<\/td>\n 6%<\/td>\n<\/tr>\n
Prior Year%<\/td>\n 14%<\/td>\n 300<\/td>\n 280<\/td>\n 8%<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
<\/p>\n
\u2022 The Algorithm with MAPE (Mean Absolute Percentage Error%) closest to Zero is considered as the industry measure for the Best Fit Forecast. MAPE is a statistical metric where the where the value corresponds to the average amount of errors that predictions have. Therefore, a lower MAPE is better, where the lower the value the more accurate the prediction is.
\n\u2022 RMSE or Root Mean Squared Error Root error helps understand how accurate predictions are and, what is the amount of deviation from the actual values.
\n\u2022 MAD or Mean Absolute Deviation represents the deviation of actual independent values from their aggregate average value.
\n\u2022 BIAS% is the variance between actual outcomes and previous forecasts.<\/p>\n
Best Fit forecast measure and forecast values available on a Dashboard enable Planners and Business Leaders to determine the Demand Planning Forecast to be carried forward. They can also over-ride the System suggested Best Fit considering:<\/p>\n
\u2022 Other measures such as BIAS% which becomes crucial given the deviation of past outcomes from forecasts for a certain algorithm.
\n\u2022 RMSE for deviation of predictions from actual values
\n\u2022 Actual pulse on the business and markets<\/p>\n
Forecasting accuracy being an evolving process, data organization, the underlying parameters to drive outputs would play a key role. This would get refined as the business iteratively executes this as there is no getting it first time right theory.
\nML Based forecasting using PlanIQ provides opportunities to improve the forecasting accuracy.
\nBenefits:
\nIn this instance for Demand Planning
\n\u2022 Reduce cycle time for scientific forecasting integrated with transaction systems.
\n\u2022 Improve forecasting accuracies leading to informed downstream decisions i.e.,
\n\u2022 Streamline Production Planning, Inventory Optimization \u2013 avoid overstocking and unavailability of raw materials and inputs.
\n\u2022 Plan for temp workers
\n\u2022 Arrange for Working Capital as per seasonality.
\n\u2022 Fulfilling Market Demands while reducing risk of overstocking of finished goods, stock obsolescence or buy finished goods to sell at lower margins.
\nAll these lead to adding improved Profitability and better Cash Flow Management.
\nConclusion:<\/p>\n
Anaplan with PlanIQ offers a very flexible forecasting options across Business Areas for improved business performance.<\/p>\n","protected":false},"excerpt":{"rendered":"
Forecasting in Today\u2019s World Frequent volatility in marketplace in the post pandemic era caused […]<\/p>\n","protected":false},"author":180,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"inline_featured_image":false,"footnotes":""},"categories":[85,69],"tags":[],"coauthors":[98],"class_list":["post-2310","post","type-post","status-publish","format-standard","hentry","category-anaplan","category-oracle"],"acf":[],"_links":{"self":[{"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/posts\/2310","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/users\/180"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/comments?post=2310"}],"version-history":[{"count":8,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/posts\/2310\/revisions"}],"predecessor-version":[{"id":2562,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/posts\/2310\/revisions\/2562"}],"wp:attachment":[{"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/media?parent=2310"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/categories?post=2310"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/tags?post=2310"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/blogs.infosys.com\/infosys-cobalt\/wp-json\/wp\/v2\/coauthors?post=2310"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}