Regardless of the type of end product, product planning is deployed in all branches of business, such as health care, pharmaceuticals, the stock exchange, automotive, insurance, wholesale, etc.
Everyone knows the phases of product planning, which are commonly divided into
Essentially, product planning consists of recognizing a demand market (or developing one yourself) and providing the right product at the right time to those customers who develop a need for this product out of wants.
If the product is on the market, it must be adapted to the changing needs of the customers and improved, and new needs must be anticipated. Ultimately, the goal is to have one’s products be placed as far down as possible on Maslow’s hierarchy of needs. Apple has demonstrated this in an awe-inspiring way with the iPhone.
What really matters in product planning today is the speed of the global markets. Today, wants become needs more quickly, and these are awakened more quickly and also mitigated more quickly. Until the middle of the 1970s, almost every household wanted to have a telephone connection, which up until then was the exception rather than the rule.
In 1983, Motorola came along and invented the mobile phone. It took a few years before Nokia was able to fulfill the desire for a personal mobile phone for almost everyone at the end of the 1990s.
A mobile phone was a status symbol and was used for many years. With the introduction of the iPhone by Apple in 2007, something happened on the market. The mobile phone became the smart phone, and it was brought further and further down on Maslow’s hierarchy of needs.
Almost everyone wanted to have a smart phone, and if possible always the newest version. But not only that. The market changed so much that, in many parts of the world, almost everyone had to have a smart phone, as the M-Pesa system for the transfer of money tellingly demonstrates.
Rather by chance, a product was generated from an idea and awakened demand. Years later, the demand became a need. And some would maintain today that their mobile phone has become a psychological need for them.
The whole thing happened within a time frame of about 30 years. Product planning was a tailored and well-established process. By the time Nokia recorded a loss of over €1 billion in 2011, it was clear that conventional product planning can no longer meet today’s requirements. Modern forecasting must analyze as much information as possible from all areas (text, podcasts, videos) and make assertions.
Too much information from the entire world must be processed at all times in order to make out what direction customer demand is headed, what the needs consist of, and how the competition is positioning itself as a result. What is the market composed of, which events influence the market, and how do customers and rivals react to these?
Collecting and processing this information was and is the field of activity of business intelligence and big data, and this can be implemented well using appropriate architecture, such as the BOSON architecture.
Understanding (perception) and drawing conclusions from and making forecasts (prediction) from the ever growing amount of information that is being exponentially increased through, for example, IoT /IoE or the Industrial Internet Consortium or Industry 4.0 can only accomplished rudimentarily, or practically not at all, with big data alone. The results of efforts to solve this with big data are more time-consuming and costly than meaningful.
Attempting to obtain competition-critical predictions solely through BI (business intelligence) systems, big data processes and deep learning alone will not lead to optimal results. Whenever customer behavior is discerned from collected data, then it involves knowledge from the past. Many customers must have already demonstrated a specific pattern of behavior in order for this behavior to be recognized. However, if that is already the case, then a large portion of the need in this market segment has already been satisfied. Wouldn’t it be good to know which completely new demands and needs are being formed right now?
Business intelligence and big data are indispensable but not sufficient for perception and prediction.
In all phases of the product life cycle, current information must be correlated with company goals in order to determine the right point in time for market introduction, product adaptation or the elimination of the product. In the process, the most important thing is always to take the right steps at the right time.
The demise of Nokia in the year 2011 and the insolvency of SolarWorld [LINK https://en.wikipedia.org/wiki/SolarWorld] in the year 2017 show that traditional product planning is nowhere near sufficient any more.
Many companies are aware of this, and count on artificial intelligence for product planning. And this is just where something is coming into existence that will give the market fresh momentum.
In the area of artificial intelligence, most companies think of machine learning or deep learning.
Machine learning defines the opportunity to parse data and learn from it, supported by software. Deep learning mostly involves the use of neural networks that are trained to recognize and classify certain patterns. This means that, broadly speaking, machine learning is a fairly general concept of artificial intelligence.
With deep learning, systems are trained so that they can recognize handwriting, voices and images, or make predictions from data streams and draw conclusions from these. The big disadvantage of these applications lies the fact that most of these systems must be trained by experts and the whole training must start from scratch with each modification, such as for example new types of images to recognize.
With these systems, input data can be screened for previously trained characteristics and categorized in order to make statements, for example during predictive maintenance (prediction for service intervals).
Deep learning delivers impressive results. However, one should not lose sight of the fact that deep learning systems are static. They can only do what they have learned to do.
But in order to be able to ingest the enormous data flows in business planning and specifically in product planning today, and to “understand” these automatically and make predictions and trend-type landing forecasts from such data, systems must be able to independently process these data and learn while doing so. They cannot continuously require new training, for this expenditure of time and money would not be profitable for any company.
Systems that learn independently are referred to as artificial general intelligence (AGI).
Artificial general intelligence is also described as “strong AI” or “full AI.” With AGI, systems have the potential to execute “general intelligent actions.” They can learn and understand independently, at least in the context for which they have been developed and in which they are supposed to operate. The major objective is the so-called singularity, though we are still somewhat removed from this.
AGI is the next generation in the area of artificial intelligence. An AGI system collects all incoming data and independently associates the information contained therein within the context of the system.
Depending on the context, systems for games, for research (e.g. in medicine) or even for robotics can arise in this manner. The context plays an important role in this, because systems that can process all data on all topics are still unthinkable today.
What is becoming truly interesting are AGI systems for companies in support of product planning.
If AGI systems are supplied with all available information from patent data banks, the social web, community sites and so on, then correlations can be detected that make whole new insights into a company and its market possible.
At the same time, the goal is not just to predict events, but rather to draw inferences regarding how one’s own company can profit from them or protect itself from them. This is referred to as prescriptive analytics.
With neural product planning, we are working in an area that takes up precisely these opportunities.
Every company is a part of at least one industry, in most cases more. These industries are standardized and classified in Germany under WZ 2008 and internationally under ISIC.
By matching business sectors to information, new findings can be gained for the supply chain of a company, because impacts on an industry in the company’s supply chain have general impacts on the company and its product planning.
Findings from these impacts about the company’s supply chain can now lead to conclusions that are decisive for competition.
Similarly to the human brain, an AGI system constantly asks itself why, where, and how, and what for. In order to answer these questions, ever more information sources are autonomously added, and little by little, a broad knowledge is built up. This course of action is executed according to a Rasmussen ladder model and takes place using algorithms that search for patterns.
These patterns will in turn be used to create genetic algorithms that search for optimal solutions. The genetic algorithms transmit their results to neural networks that, in turn, activate further neural networks, until a result can be delivered.
The search for patterns is performed in the system at all times, and leads to ever more knowledge that is manifested in ever newer and better genetic algorithms.
A self-learning system that assembles more and more knowledge and identifies relationships is generated. Completely autonomously and independently.
The knowledge discovered with these methods becomes ever more concentrated and interlaced in various layers with itself. Millions of small neural networks are combined with one another and procedures and action plans are formed using sparse coding algorithms.
Action plans are best practice behavior patterns, similar to the process in the cerebellum, a region in the brain that is responsible for movements. For instance, one can catch a ball in two different ways. One can calculate the flight curve using complicated differential equations, and incorporate the angle of the eyes and the arm movement to be made, so that the hand is in the correct position at the correct time to catch the ball.
Because this would take too much time for the brain, these equations are simplified and transformed into a simple trend model. Using this trend model, the anticipated target location of the ball is calculated. Taking into account the field of vision, the same is done for the movement of the hand. Should the something about the flight curve change, new calculations are corrections are made ad hoc. Sometimes, nevertheless, the ball will not be caught. The brain is very exact indeed, but not 100% correct.
In an AGI system, this method of proceeding saves memory capacity and simultaneously increases the processing speed of the system. The outputs have a high level of plausibility, but cannot be unequivocally correct.
The cumulative knowledge in an AGI system is mapped in lists upon lists of genetic algorithms, neural networks and procedures. These, in turn, refer to knowledge that is archived in graphs and hypergraphs, and a recursive interplay develops. In order to detect the importance of discrete pieces of information, each piece of information is continuously matched using parameters for long-term and short-term knowledge, along the lines of Hebbian theory.
Visual and auditory information is pre-processed using Deep Spatio Temporal Inference Networks (DeSTIN) and then is transformed into new information again using genetic algorithms.
Similar to the human brain as we understand it today.
In contrast to a human brain, however, an AGI system can become so parameterized that information is never forgotten. For this purpose, a delicate trade-off between costs and usefulness must take place so that the storage capacity and processing speed (and therefore the costs) of the system remain positive with the expected added value of less important information. With the NFI (never forget information) feature, unique and exceptional events can still be earmarked for perpetual memory. The speedy and reliable recognition of NFI, and the determination of the importance of these pieces of information, for example during the pattern recognition and generation of new genetic algorithms and neural networks within the AGI system, is one of the interesting challenges.
In order to obtain a reference to the supply chain and thus on procurement, neural product planning automatically correlates all information to individual lines of business through various levels of recognition algorithms. New findings arise using time series analysis, cluster analysis and other complex methods. This is analogous to the “thinking” that occurs in the human brain or what is described as a “flash of insight” that occurs even during dreaming.
After this, the information from the individual sectors is analyzed for what impact it has on the supply chain of a company. Analogies to previous similar impacts are taken into consideration in order to obtain a better confidence factor.
For this purpose, it obviously must be known which business sectors are in the supply chain of a company and how to proceed with the impacts on them. What influence do the individual links have on the whole chain and how should these be substituted in case that is needed? Which action is required when a link of the supply chain suddenly represents a risk for one’s own business goals due to exogenous or endogenous risks? What does this mean for procurement?
For example, these risks can be unrest, epidemics or other catastrophes that hamper the ability to deliver raw materials. A fusion similar to that between Bayer and Monsanto can impact one’s own product planning just as much as a newly emerging trend. For instance, the strong effect of the #SugarFree trend did not just surprise Coca Cola.
A modern application of SCRM, similar to a modern data provenance application, must be in a position to quantify these effects on the company. It must identify the level of risk of a link on the entire supply chain and the countermeasures provided for this. Because the risks change continually, the indices in such systems are, as a rule, adapted with automated processes.
In a neural product planning system, impact clusters among other things are formed in order to recognize these impacts on the supply chain. Through these impact clusters, it is determined why there are impacts and whether these directly or indirectly concern product planning.
Using cascades of forward and backward reasoning processes, these impacts are put into context, and new findings are obtained with the help of abductive inference.
Among other things, using outputs from Dietrich Dörner’s PSI theory and inspired by Joscha Bach's MicroPsi AI system, neural product planning works on capturing this knowledge and gaining new knowledge and possible actions from this in order to reach critical conclusions.
In doing so, neural product planning interacts with a variety of other levels in the system, and there ensues a synergy of the most modern methods and algorithms that reflects the current view of human brain function.
A process that is performed around the clock. Every day, at all times, and completely autonomously.
By integrating patent databanks, predictions about the actions of competitors can be made. Statements in blog entries and social media channels reveal current opinions and trends. All manner of websites report on events and new topics. All of this information is put into context.
Like a human, a system can only analyze what it knows. The more knowledge flowing into neural product planning, the better the conclusions that can be reached and the support that is given to product planning.
The integration of all company information and open information is therefore a logical consequence. Through cognitive BI and linked open data, additional important information sources are brought together and analyzed. The video “Introduction to Linked Open Data” demonstrates this interaction.
As with human learning, it is crucial not to overwhelm the “brain” with knowledge, but rather to achieve a cognitive development in the desired context. For this reason, an AGI system must, analogous to Jean Piaget’s stages of cognitive development, pass through certain phases of learning. This means that the knowledge in the context of the AGI system must be constructed according to certain rules and become more and more sophisticated. In the future, this job will belong to a neural cognitive engineer.
Neural Product Planning is
autonomous assessments of all available information with the goal of delivering conclusions and predictions that are key for business success.
a system that works similarly to the human brain, that learns independently using artificial general intelligence and puts events and information into a meaningful association with the company’s supply chain.
a system that recognizes the bearing of positive or negative impacts on product planning by identifying and processing all possible information, then evaluates it, and reaches appropriate conclusions.
the best opportunity to apply successes from AGI research and development in the area of robotics, intelligent games and healthcare to product planning.
the support of all colleagues who are too valuable to be kept busy for protracted periods of time with analyzing an immense amount of information, training and programming of software, and trial and error. Neural product planning operates like a human colleague who handles these tasks 24 hours a day, every day of the year, gladly and conscientiously.
the solution for anticipating, independently and without prior training, exogenous and endogenous events that are critical to the company. Today, event prediction is laboriously trained and programmed using methods such as time series analysis in connection with reasoning processes and neural networks. There are successful examples such as the conclusions of the CIA, which can detect social unrest up to five days in advance, or the identification of epidemics months before they arise. However, these examples are all based on methods and systems that were laboriously prepared in advance for their exact task. With neural product planning, such impacts, just like new trends, new demands and needs, and new market opportunities and market risks, are recognized without having to be trained or programmed in advance for a specific issue.
What’s more, neural product planning analyzes reasons for customer behavior in that it searches automatically for cluster conclusions for a laddering process, for example, or for answers for Herzberg’s theory.
But these are only simple examples of the possibilities of neural product planning.
We work on neural product planning and, in the future, will support businesses in reaching better predictions and optimizing their product planning, their procurement, and their forecasting.
Neural product planning, together with cognitive BI, delivers important information, insights and knowledge and will help the decision-makers, personnel in marketing and product planning, and data scientists be able to deliver the right products at the right time, in the right model, at the right place to the right customer. Furthermore, the system will point to risks and trends in the supply chain and to possible effects on procurement and thereby make it possible to act in a timely manner in the overall product life cycle.
Neural product planning always interacts with the colleagues of the company and relieves them of tedious and overly repetitive work steps and analyses, so that they can concentrate on what they do best – being ahead of the competition.
Get in touch or send us an email at contact@MySecondWay.com