Activate the existing collective intelligence with machine learning

‍

Year after year, vast amounts of valuable strategy information from many stakeholders pours into our organization. Given the big volumes of data and the complexity of our businesses, we hope to gain insights across markets and divisions, but find that the flood of information pushes us to human limits and the effort required to evaluate it often outweighs the insight gained in the end.

As part of the strategy and planning process, not only quantitative data is collected, but also a lot of qualitative information. Standardized templates contain the accumulated knowledge of brilliant minds about underlying assumptions and their interpretations, descriptions of trends in the environment, opportunities and risks, strategic and operational challenges, and so on. All this is stored in a multitude of folders and files and often in different languages. In addition, there are different formats, different levels of detail and qualities that hinders its valuable "activation". Instead of exploiting this treasure trove of data to the maximum and generating knowledge to do the right things, it is usually only well archived.

For strategy work, it is crucial to intelligently connect, cluster and correctly evaluate the knowledge already available of many people from the markets in order to be able to actively work with the findings at key points in the strategy process. In practice, gathering information of markets, competitors, customers, opportunities or even ideas for innovation from the submarkets is the smaller problem. It is more a matter of being able to handle the abundance of qualitative and quantitative information correctly and to draw the right conclusions from it. This is undoubtedly a mammoth task, but the use of machine learning makes it less daunting.

Many companies are not aware that they already fulfill several prerequisites for the use of such innovative approaches. With the appropriate amount of data available and the right choice of methods, they can discover and lift the data gold and at the same time open new gates of "strategic insight". Especially in strategy work in more complex organizations, there is no shortage of use cases.

‍

AI – decision support for strategy development

Machine learning and artificial intelligence address one of the core questions of strategic management: How do we get from data to information and, via this, to actionable knowledge? Intelligent algorithms provide us with a change of perspective and a new approach to answering relevant questions of strategy development and strategic planning – especially in more complex businesses.

Here are a few examples of practical, versatile and technically scalable use cases in the context of strategy work:

• Automated clustering and gain deep understanding of large amounts of text by Text Clustering: Text clustering techniques use existing language models to identify semantic similarities from a large number of populated templates. Intelligent, self-learning algorithms extract relevant information from large amounts of text and recognize strategically usable patterns in all available qualitative data and information. This means that all texts - even the smallest fragments - from the strategy and planning process are considered, processed, and synthesized about their content-related meaning and "proximity" to other content. This can be a large number of text elements on, for example, strengths, trends, opportunities, challenges or even ideas, which can be grouped at any point using defined filter criteria. The algorithm uses an unimaginably large number of existing texts, compares them with our own written text in a multi-dimensional vector space, and provides us with a suggestion as to how, just as one example, 500 trends from different market segments can be meaningfully clustered.

• Automated classification of strategy content: Almost every company has already introduced some base categorization dimensions in one form or another across the group. In one case, BSC categories or variations thereof are used. In another case, there is talk of Building Blocks, Key Performance Indicators, or other categories. These categories help to group strategy content, goals, initiatives, measures or, for example, ideas into meaningful groups. What is resource-intensive today, done manually for each individual piece of content, can be automated. The machine can make a suggestion for the categorization that is in no way inferior to categorization by humans. If I change suggested categorizations manually, I thereby train the language model and sharpen my own "corporate language." The overarching benefit is obvious. We can sort strategically relevant information across units at the push of a button according to different, predefined categories and draw conclusions from them. In addition to saving time by eliminating manual tagging, we have the advantage of identifying overarching patterns faster and in better quality.

• ‍Better forecasting and plausibility of planning through predictive modeling: For many years, we have been estimating market volume developments in finely segmented submarkets and come up with sales forecasts for these markets. We should not underestimate the importance of the existing history of data and use it more actively. The longer and more high-frequency the available historical data series, the more encouraging this is for predictive modeling algorithms. These methods provide predictions based on patterns from the past. We all know that in times of high dynamics and complexity, the past is not suitable for plausibly producing predictions for the future. Nevertheless, in some practical cases we have found that the machine is very often capable of making conclusive and also confirmed forecasts for, for example, market volume developments based on patterns from the past. We can use the technology as a plausibility aid for our own forecasts. If individual data series are additionally correlated with data series from other markets, we get an even better feel for which of our markets might follow similar rules of the game. The principle of "human in the loop" naturally has high priority here as well.

A completely new quality of strategic interpretation becomes reality

With technologies such as text mining, natural language processing (NLP) and predictive modeling, it is possible for the first time to use the complete historical and constantly growing strategy data treasure automatically. And this at any time and always up to date.

For strategy development, this means an extraordinary gain in quality, because strategy content is discussed in a cross-functional and over time in a fact-based and insight-oriented manner. Thanks to the latest technologies, justifiable and comprehensible answers can be given to questions that were practically impossible to answer using old methods.

In doing so, we are not necessarily dependent on data sources external to the company. The use of existing knowledge from the market and product-related areas is the most effective and efficient way to gain new insights and bring the strategic dialog to a new level of quality. The prerequisite is the active involvement of the many people who know the submarkets and their rules of the game in detail.

It is right and good to permanently preserve the quantities of quantitative and qualitative high-quality strategic information for the organization. To actively use this hidden treasure of information is even smarter and possible. We should start now to actively leverage the knowledge in the organization. It is the only way to increase necessary agility and to mobilize those self-organizational forces that are crucial for success in dynamic times. The company will be surprised what the company already knows!

‍

In this articles series have been published so far:

Article 1: Revolutionize your Strategy Development

Article 2: Agility and speed – key success factors in Strategy Development

Article 3: Top-down or Bottom-up in Strategy Development

Artikel 4: The hidden strategy treasure – If the company knew what the company all knows

‍