Getting Familiar with the Business Technology Platform and its Intelligent Technologies

SAP Business Technology Platform (SAP BTP) brings the intelligence to the Intelligent Enterprise. It is an integrated offering comprised of four technology portfolios, giving users flexibility to choose SAP technologies that provide an intrinsic understanding of data and processes in SAP and 3rd-party applications. Business-centric services in the four technology portfolios offer agility for companies to quickly turn their data into business value.

SAP Business Technology Platform (SAP BTP) brings the intelligence to the Intelligent Enterprise. It is an integrated offering comprised of four technology portfolios, giving users flexibility to choose SAP technologies that provide an intrinsic understanding of data and processes in SAP and 3rd-party applications. Business-centric services in the four technology portfolios offer agility for companies to quickly turn their data into business value.

SAP Business Technology Platform offers fast in-memory processing, sustainable agile services to integrate data and extend applications, and fully embedded analytics and intelligent technologies.

Let's now clarify some of these intelligent technologies:

• Artificial Intelligence (AI) and Machine Learning (ML):

  The focus of these intelligent technologies is building bots that understand humans, leveraging robotic process automation (RPA) tools and software, harnessing natural language processing (NLP) technologies and capturing user interactions.

• Internet of Things (IoT):

  IoT tries to innovate and enhance processes with data-driven intelligence from machines, products, supply chains, customers, and partners. The goal is to convert sensor data into intelligent, real-time insights and to recognize patterns and automate events. All this while managing devices (using integration services) and while securing data consumption and executing complex processes.

• Blockchain:

  Blockchain tries to simplify complex multiparty processes and create trust among participants. Distributed ledger solutions from SAP are also used in this context. SAP aims to ensure a seamless cloud deployment, leverage an open blockchain platform where pilot and productive blockchain applications can be built and monitored in real time. The solution integrates enterprise and blockchain data.

An example of using Natural Language Processing (NLP) features in the area of SAP S/4HANA Sales is the SAP Fiori app used to create a subsequent sales order:

This app has the following characteristics:

• NLP enables internal sales representatives to create sales orders from sales quotations using Natural Language Processing (NLP).
• NLP allows a user to gain insights and suggestions from Natural Language Processing (NLP) to decide on the proper follow-up step for a certain process and to quickly create a business object with minimal input.
• NLP allows a user to use natural language to show one or more open sales quotations, select an open quotation from the list, and create a subsequent sales order.
• NLP provides navigation to related applications to display for example sales order details.
• NLP allows for communication using natural language through both voice and text.

The app Track Sales Orders allows an internal sales representative to use natural language processing to get instant insights and track sales orders. It has the following characteristics:

• Natural Language Processing (NLP) is available as a side panel.
• It has is fully integrated with all related applications.
• It provides insights which can be turned into actions: resolve issues instantly by drilling down.
• It pushes real-time figures to internal sales representatives.

The goal of a machine learning algorithm is to determine a mathematical model, that is used to improve the algorithm's outcome automatically through experience. Machine learning algorithms build a mathematical model based on sample data, known as training data. This is done in order to make predictions or decisions without being explicitly programmed to do so.

SAP provides a vast library of machine learning algorithms developed by SAP and by its partners, such as TensorFlow by Google. These algorithms are part of SAP HANA. SAP HANA provides data governance engines necessary to prepare, govern, ensure quality, and treat in real time, the data necessary to train its own predictive analysis engines, and other machine learning algorithms available on the SAP Cloud Platform. The automated predictive library (APL) in the SAP HANA database optimizes machine learning models automatically, without the need for a data scientist.

Machine learning models are so powerful because they mimic the behavior of historical data. An actual predictive model is trained on live customer data that never leaves the SAP S/4HANA system, not even for training. Model creation and activation is done by an analytics specialist. Model training is a one-click action and the assessment of a model is also highly simplified. The analytics specialist decides if a predictive model should be activated. When the usage of the model in an app yields new actual data, the model can be easily retrained if needed, so that the updated model is used for new predictions.

A few examples will now be discussed where machine learning can be used in a certain SAP Fiori app for SAP S/4HANA Sales (including its billing functionality).

This SAP Fiori app allows a company to identify the risk of potential delays for its open sales orders. It provides important insights into the current sales order fulfillment situation:

The goal is to be able to take action early on to avoid the predicted delays. The app does this by focusing on the predicted delay of delivery creation and the predicted delivery processing delay. The system can predict the delivery delay based on what it has learned from its training of the predictive model.

For the training of the predicted delay of delivery creation, the system uses the planned delivery creation date of deliveries as follow-up documents to open sales orders. To do this, the system compares the planned delivery creation date from a confirmed schedule line of a sales order item (that has already been delivered), with the actual delivery creation date of the corresponding delivery.

For the training of the predicted delivery processing delay, the system uses past data to compare the planned goods issue date of all deliveries for which the goods issue is completed, with the actual goods movement date of the follow-up deliveries. This calculation is carried out for every delivery for a corresponding sales order item. The system then takes the maximum delivery processing delay of the analyzed deliveries. This means the delivery with the highest delay as the predicted delivery processing delay.

Quotation conversion rates can be used to track to what extent quotations are being converted to sales orders before expiring. By leveraging machine learning capabilities, predictive insights into quotation conversions can be gained by comparing actual and predicted results.

The following demonstration will show you some features of the SAP Fiori app Quotation Conversion Rates - Valid / Not Completed:

The analytical SAP Fiori app called Quotation Conversion Rates - Valid / Not Completed has the following characteristics:

• The app can be used to predict quotation conversion rates.
• The app gives insights and provides reliable predictions for a sales manager for achievable Incoming Sales Orders.
• The app provides an increase accuracy with respect to sales planning.

The quotation conversion rate measures the percentage of the net value of order items that have been converted from a quotation item, based on the total net value of quotation items.

Only quotations that meet the following criteria are included:

• Net value > 0

• Valid on the current date

• Not fully referenced (header of the quotation)

• Overall status is not completed (header of the quotation)

The app allows to View Quotation Conversion Rates:

• The tile displays the overall conversion rate of quotations. In the app, it is possible to drill down into quotation conversion rates by different dimensions, such as sales organization, customer, product, and employee responsible.

• View Top 10 Quotations by Net Value:

  It is possible to analyze to what extent the 10 quotations with the highest net values are being converted into sales orders within the validity period.

• View Bottom 10 Quotations by Conversion Rate:

  It is possible to quickly identify the 10 least converted quotations, especially those that are close to expiring.

The app also allows to Predict Quotation Conversion Rates:

• It is possible to compare the actual and predicted conversion rates dynamically while drilling down by different dimensions.

  To see the predicted conversion rate, the mini chart (hidden by default) needs to be shown that displays the predicted rate.

• The net value of the converted quotations can be compared with the prediction by different dimensions.

When training the predictive model, consider the following points:

• Use data that is still accurate and relevant.

  Training a predictive model with data that is no longer relevant for the current business situation, could impact the quality of the model. An example is data that is too old or data from exceptional cases. To improve the results, filters can be set to exclude irrelevant data from the training data sets. If the conversion rate of sales quotations is normally never more than 100%, the data of quotations whose conversion rate is higher than 100% can be excluded from the training data sets since these were exceptional cases. This is done by setting the high value of the training filter for quotation conversion rate (SLSQTANCONVERSIONRATE) to 1.

• Train the model with a sufficient amount of data.

  A successful training of a model requires a sufficient amount of data which is reasonably distributed among the key business dimensions, such as customer and product. This helps the model learn the key aspects of a company's business and thus positively affects the model’s predictive capabilities. Insufficient data may result in a failed training. Therefore, it is recommended to train models only after the systems of the company, especially newly installed systems, contain a sufficient volume of business data.

• Train and activate the predictive model at least once a month.

  The amount of time required for each training depends on the volume of training data. It is recommend to train the model during non-working hours, for example, at night.

With the app called Sales Performance - Predictions, it is possible to compare currently achieved sales (for example, sales volume) with modeling-based predictions:

By selecting an existing sales plan, it is possible to analyze to what extent certain sales targets are being achieved. Using this app, predictive insights into current sales performance can be gained.

Currently, this app is only available for SAP S/4HANA Cloud (for example the 2108 and 2111 releases and onwards).

The intelligent product proposal functionality is an enhancement of the dynamic product proposal functionality, which has been available for several releases already in SAP S/4HANA. The intelligent product proposal functionality itself was introduced with SAP S/4HANA release 2020.

The dynamic product proposal combines data from several data sources to create a product proposal: order history, listed materials (that is, products), excluded materials, item proposal, customer material info record, customer-specific data sources. The system then accesses the data sources either online or in the background.

The intelligent product proposal functionality allows users to receive product and quantity proposals as an input help to enable faster sales order entry in the SAP S/4HANA system. The quantity of products ordered can be predicted with the help of an algorithm based on second order exponential smoothing. The algorithm is able to predict trends, but cannot take seasonality effects into account. This would require a third order exponential smoothing algorithm which is planned for a future release of SAP S/4HANA.

Since the intelligent product proposal is based on the dynamic product proposal, it is essential to carry out all the configuration activities for the dynamic product proposal.

In addition to the dynamic product proposal, the intelligent product proposal allows you to predict whether the quantity of products ordered will increase or decrease. The necessary settings can be found in Customizing for Sales and Distribution, under Basic Functions → Dynamic Product Proposal.

The intelligent product proposal functionality for sales order creation has the following features:

• Receive product and quantity recommendations based on historical data, while considering product listings and exclusions for the proposed products
• Fine-tune the provided proposals based on additional criteria such as customer or sales document type (for example, standard order)
• Enable external consumers to use the provided API to call the intelligent product proposal function

Using the intelligent product proposal functionality has several benefits. And it provides business value for various areas of a company:

• Improvement of sales force efficiency by using product and quantity proposals for faster sales order creation
• Increased profitability and avoidance of errors through an improved sales order creation process
• Support of customer oriented activities by learning form historical customer data