From the Stockpile of Documents to the Map of Knowledge

Wednesday, 19. June 2019

Preserving Qualitative Strategy Data Is Good, but It Is Better to Let Action-Guiding New Knowledge Emerge from It
Strategic Insights: Part III

A guest article by Ronald Herse & Florian Dreher


In part II of the publication series Strategic Insights, we demonstrated practical ways of improving the quality and reliability of forecasting in planning.

This article is dedicated to a much greater potential for strategic insights – which can finally be realized thanks to modern methods such as Machine Learning (ML) and Artificial Intelligence (AI): the full use of the available qualitative strategy data set, which slumbers in countless folders, PowerPoint and Excel files of an organization.

The already existing and constantly emerging qualitative information is − in terms of quantities and qualities − even more interesting than the quantitative data.

The reasons are plausible: The planned turnover and yield values are the final result of an intensive dialogue and personal experience. The underlying assumptions and the interpretation services that result from them are decisive and in the best case lead to justifiable actions. This is the actual core of strategy work and is already reality in most companies in the form of qualitative descriptions. The "weapons of choice" of such considerations are standardized templates that are processed and updated annually as part of the strategy and planning process.
 
In most cases, descriptions of drivers and trends from the environment are also available; the resulting opportunities and risks for the company itself and its competitors. Strengths and weaknesses, also in relation to relevant players, are identified, strategic and operational challenges are formulated and strategic options and activities to be promoted are recorded. This is done on a regular basis for a large number of planning units, taking into account minimum methodological standards. The knowledge of these brilliant minds is thus selectively documented and preserved as information. This does not lead to a gain in knowledge. But it is possible!

Simple but not easy ...

Stockpile of documentsAs catchy and great as the potential benefit for strategic interpretations from the use of all these text modules may be, the devil is often once again in the detail.

This information is distributed in hundreds of orders and files and is often available in different languages. Different formats, degrees of detail and qualities do the rest and successfully prevent you from doing more with this data treasure than "just" preserving it and searching very selectively for combinations and possible interpretations.

Text Mining and Natural Language Processing (NLP) Are the Key

The challenges that the much-quoted VUCA world holds for us can no longer be mastered with the conventional methods of pure text recognition. New challenges require new solutions and the existing technologies open the doors to a completely new world of high-quality and action-guiding strategic interpretation.

Text Mining is essentially about extracting relevant information from (very) large amounts of text using intelligent, self-learning algorithms from the field of NLP. In the best case, Text Mining provides information that users previously did not even know if and that was contained in the processed text.

Typical application areas of Text Mining/NLP are document classification and automatic indexing of documents, translations, pattern recognition in texts up to automated text summaries and hypotheses. The rapid progress in NLP in recent years is based on the high computing power available, the amount of data available on the Internet and deep learning technology.

Transfer Learning Creates a Uniform and Comparable Strategy Language

In Transfer Learning, a universal language model that has been trained on a massive amount of data is adapted to a specific problem by fine-tuning.


How Transfer Learning works

Illustration 1: Schematic illustration of how Transfer Learning works


The amount of data in ML and AI is often the critical factor that determines success or failure. Transfer Learning offers a decisive advantage: Success is not prevented by too little data. We can build on a state-of-the-art language model and use its existing capabilities for our own strategic planning work. The language model thus "translates" the many text modules of our planning units from the different segments and regions into a uniform strategy language for our own organization.

Using a practical example: 30 planning units individually describe developments in the environment, opportunities and risks along individual regions as part of the planning process. In addition, trends are recorded for product groups, some of which are relevant across regions. This happens on a rolling basis with a history of ten years, with varying degrees of detail in the description, in four languages and with different grammar and choice of words. At the same time, a large number of factually not systematically evaluable and partly contradictory studies are available.

We can no longer see the forest for the trees. At the latest when we want to look beyond planning units to our business, we know where our human interpretation boundaries lie.

Text Mining and NLP solve some of the known knots and pain points of strategy work in a new way. All existing texts and even the smallest fragments from the strategy and planning process are considered, processed and synthesized over time with a view to content meaning and "proximity" to others.

Textual clustering of regional trends
Illustration 2: Symbolic illustration for textual clustering of regional trends of a year


Tydiying up the Babylonian Strategy-Language Confusion

Text Mining and NLP offer a new way to identify significant, strategically exploitable patterns based on all available qualitative data and information! In addition, a common understanding is also generated linguistically.

With KI/ML and NLP, we are using the historically completely available strategy data set for the first time – and completely automatically, too. This makes the development of strategy content cross-sectional and over time fact-based and knowledge-based. Now, justifiable and comprehensible answers can be given to questions that could practically not be answered using old methods.

For example, we can also see much better how individual trends in the regions develop over time and with what time offset, and which opportunity patterns emerge for our individual businesses. The clever clustering and alignment of our own strengths, weaknesses and activities over time raises the gain in knowledge and strategic dialogue to a completely new level of quality and enables superior strategic planning with maximum efficiency.

In the fourth part of the publication series Strategic Insights, you will gain important insights into environmental monitoring through correlation and time series analyses.


Credits (from top):
© istock.com I magnifying glass independently added
© istock.com
© Ronald Herse/Florian Dreher (own visualization)
© Ronald Herse/Florian Dreher (own visualization)
© Ronald Herse
© Florian Dreher



About the Authors

Mag. (FH) Ronald HerseAs a partner at EVOLUTIONIZER, Ronald Herse is responsible for the development and integration of business content into software for strategy development and implementation.

He was managing director and partner of a leading software company for nine years. As a pioneer in the combination of integrated management concepts and modern software for anchoring excellent management in complex organizations, he implemented concrete and sustainable solutions in the entrepreneurial practice from very early on. 

Previously, he worked in more than 80 international projects for six years as a consultant and spiritus rector of the Competence Center for Strategy at the Malik Management Center St. Gallen, Switzerland.

Already at the beginning of the digital era, he founded a digital marketing agency and still teaches at the University of Applied Science in Vorarlberg, Austria, on strategic management, transformation and change processes. You can reach him via ronald.herse@evolutionizer.com.

Florian DreherAs Senior Data Scientist at EVOLUTIONIZER, Florian Dreher is responsible for everything from simple statistics to artificial intelligence.

He uses machine learning methods for topics such as predictive analytics, data mining and text mining to make EVOLUTIONIZER products even more intelligent.

Florian Dreher worked for an automotive supplier for four years. There he used machine learning to automatically forecast complex product characteristics in order to optimize product development.

As a trained engineer in mechanical engineering, he built up his data scientist skills in a joint project with the Fraunhofer Institute and from then on continuously expanded them on the job. You can reach him via florian.dreher@evolutionizer.com.