Part I: New insights through use of existing strategy data

Developing a robust strategy has never been easy. High business complexity, market dynamics and shorter planning horizons create uncertainty and make it harder to make the right decisions at the right time. The half-life of a strategy becomes shorter, the importance of a strategy increases. The call for more agility in strategic planning is understandable. A lever lies in a new handling of the already existing strategy data estimates. Machine learning methods are developing into useful service providers of the "strategic command center".

Many companies are already successfully using intelligent data analysis and evaluation for decision support in purchasing, production or warehousing. Patterns are discovered, outliers identified and possible developments predicted as a basis for future scenarios via the frequent and structured availability of large amounts of data from the past. Extended statistics and machine learning make this possible and the practical successes speak for this "digital consultant".

The strategy process and strategic planning are among the essential drivers of a business. Despite their central importance, the use of these "opaque" methods to support decision-making is still in its infancy.

The basic prerequisites for this are good:

• Most medium-sized companies have systematized their strategy process and have already anchored it organizationally at the various planning levels
• This process has been going on for several years at regular intervals
• Methodologically, minimum standards exist and comparable data structures for strategy development with recurring input variables can be found
• Both quantitative and qualitative data are collected and plan and actual data are available on a rolling basis

Strategy data potential for intelligent evaluation is greater than one might think

In practice, it is not uncommon for the strategy process to pass through a more complex business for a total of 30 to 50 planning units annually in varying granularity. Let us assume that a business unit consists of several submarkets (e. g. market segments, application fields). Within each planning unit, data is often collected for further segmentation dimensions (e. g. region, products, customers). This collection has a comparable data structure and common denominators. This produces a wealth of new quantitative and qualitative information every year, which can be actively used to gain knowledge.

The hypothesis is that there are common qualitative and quantitative denominators across planning units with rolling, recurring updates:

• Market volume developments for the respective planning unit and detailing by regions, product groups or similar. (plan/forecast and actual)
• Sales and profit developments of the own planning unit and the direct competition (both plan/forecast and actual)
• Sales and profit developments by regions, countries, product groups (plan/forecast and actual)
• Drivers, trends and hypotheses on the relevant environment
• Strengths, weaknesses, opportunities and threats (SWOT) for our own company and for our competition
• Key strategic issues or challenges to be resolved
• Strategic options or alternatives, initiatives and projects to be taken forward

Compared to other machine-learning applications, our strategy work involves relatively small amounts of data and a manageable number of time series points. It is manageable for the machine, but already difficult for the human being to grasp its entirety and to process it further for knowledge gains. The right methods applied in the right form open new doors of "strategic knowledge" for us.

Core question of corporate management

Machine Learning and Artificial Intelligence address one of the core questions of corporate management: How do we obtain information from data and action-guiding knowledge from it? Today we do not suffer from a lack of data or strategy information. On the contrary. Due to the wealth of data available and the complexity of our business, we reach practical and mental limits in order to generate relevant and action-guiding knowledge.

Machine-learning approaches will help us to overcome some of the limitations and to use existing empirical data even smarter for better decisions:

• The entire wealth of available information across all planning units and targets will be usable in real time, taking into account all possible combinations and data cuts (e. g. correlating patterns of market segments crossed with product groups and regions in terms of sales and profitability).
• For the first time, quantitative and qualitative strategy data can not only be analyzed over time using intelligent methods, but can also be correlated in a meaningful and comprehensible way.
• The machine identifies relevant patterns (breaks) and conspicuities from the available data and their series themselves. The annoying search for "the needle in the haystack" by manual comparison and filtering is reduced to a minimum.
• We train the models by human corrections and updating the data over time. The machine learns and gets better.

Change perspective: known questions rethought …

Intelligent algorithms provide us with a change of perspective and a new approach to answering relevant questions of strategic control and planning.

A few examples:

• ‍Improvement of your own planning quality through relatively "simple" statistics
  How reliable is our planning work really? Where and why do we regularly over- or underestimate each other and why? What is the "right" planning horizon for us along our submarkets?
• ‍Understanding common patterns in our businesses through time series and correlation analyses/clustering methods
  Which of our (sub)markets have common patterns in their development? Which submarkets are similar and with which pattern or time lag? Where are there "clusters" with striking similarities in business?
• ‍Supported forecasting and plausibility checking of planning through predictive modeling
  How could our business develop in the next two years on the basis of all the patterns of the past?
• ‍Deeper understanding of the content of the business through text mining and semantics
  ‍What are the similarities in content, relevant differences and overarching patterns in market, competitive and customer trends? Are our assumptions regarding the development of the market, sales and profitability plausible, taking into account the environmental conditions and defined strengths, weaknesses, opportunities and risks?

Good answers to these questions consider the individual planning unit as well as the cross-sections across several units through different "cuts" in the planning dimensions (e. g. through regions, product groups, application fields) and planning variables (e. g. turnover, profit, volume). The dimension time and the rolling data update are taken into account in the form of plan values and the actual data. Qualitative and quantitative findings are "partners at eye level" and confirm each other or question each other for reasons.

Discovering and actively using existing strategic knowledge potential

Every year, many of the actors involved provide our organisations with quantitative and high-quality strategy information in considerable quantities and in various formats.

Conserving this for the organisation is right and good. To actively use this hidden treasure of information seems even smarter and is practically possible. The company will wonder what it knows!

‍This article is the first part of the publication series Strategic Insights. Ronald Herse will give you more exciting insights into the world of strategy and data in the future.‍