Geo-deep-learning for the best possible decision – Meet Geospin


Global digitization encompasses trends such as the Internet of Things, Industry 4.0 and Smart Cities. Mobility and logistics service providers, for example, have to deal with ever changing customer needs, new technologies and legal requirements. To cope with these challenges the service providers can use their proprietary data. However, a firm understanding and prediction of customer needs often requires the combination of internal and external data sets.

Geospatial intelligence approach; picture Geospin

Imagine you are planning a new inland freight terminal and would like to examine the full market potential of your investment thoroughly. Imagine further that you possess comprehensive internal data sets for example on demand forecast, infrastructure connectivity, and capacity planning. In a second step, you enrich these data sets with external data sources for instance on congestion of road and rail network; freight flows in the target location with a 150-kilometre radius; and density of nearby logistic service providers. Lastly, a software connects the relevant parameters in an intelligent way, machine learning algorithms compute the vast amount of data sets and visualizes the results in a user-centric view. Now you cannot only analyze where the market for your freight terminal is likely to exist but also when and why the demand incurs.

In exploiting the full potential of geospatial data, companies can precisely forecast demand and take better-informed investment decisions. This is precisely the value proposition of Geospin, a spin-off of the information system research department of the University of Freiburg. The approach integrates innovations from the fields of machine learning, deep learning, neural networks and predictive analytics to a unique service. Geospin’s technology is deeply rooted in state-of-the-art academic research (e.g. Wagner and Brandt’s publication on analytics for free-floating carsharing) which has been published in re-known scientific journals. Furthermore, the start-up has been awarded at competitions like code_n,  startinsland, and the Erasmus Energy Forum.

Co-Founders from left to right:
Dr. Bendler (CTO), Niklas Goby (COO), Dr. Wagner (CEO), Dr. Brandt (CFO), Dr. Gebele (CMO)

Best possible decision? Intuition meets Geo-deep-learning

Whether in the mobility sector, logistics, construction, the automotive industry, mobile communications or branch network planning for banks — in the future, company decisions will no longer be possible without the analysis of dynamic and fine-grained geographic data. For this, the handling of large data sets, as well as experience with current procedures in the fields of machine learning and deep learning are necessary. By smartly merging companies’ data with external geospatial data Geospin methods help companies to evaluate, analyze and predict behavior in cities.

To see how Geospin’s technology is leveraged for a particular use case, let’s have a look at the challenge of branch network planning for a bank. Mobile and online banks are increasing the market pressure on traditional incumbents, who carry a large overhead cost for their brick and mortar branch network. Thus traditional banks are facing a number of decisions to adjust their strategy: Fuse branches? More self-service counters and ATMs? Close branches? What is the right mix for which city? Et cetera.

To answer the above questions, traditional banks have to crunch massive data sets and rely on their judgment of soft factors. What is the right balance between profitability and being close to the customer? Do I become more competitive with cutting costs or driving expansion? In which location do customers demand a branch and which ones can be shut down? Add uncertainty and volatility of economic factors and decisions making for the entire branch network becomes a true Sisyphus challenge.

Machine Learning empowers branch network planning

To explain consumer behavior in a region or a quarter of the city requires an in-depth understanding of the underlying causes why certain services and products are demanded under specific conditions seen in context with the direct surrounding of each branch.

At this point, Geospin’s data insight capability comes into play. Depending on the bank specific portfolio, customer segmentation, strategy, and business model, some possible solutions – each with a unique set of advantages and disadvantages – can be articulated. For instance, the closure of an individual branch leads to substantial cost savings, but can also result in losing valuable customers.

Before data sets can feed the algorithms of Geospin’s software, they are collected, for example, local maps, strategic points of interest, and flow of people.  Additional data sets, for instance, on purchasing power and buying behavior of customers, opening hours of surrounding businesses like supermarkets, malls, and restaurants, or open data of mobility service providers enrich the existing data sets with valuable information and disclose why some branches are performing more profitably than others. In combination with customer movement profiles and frequency analysis, the data finally displays the tendency of when customers spend money where and this can be crucial to understanding a branch’s performance.

This slideshow requires JavaScript.

Thus the impact of strategic decisions on the branch network can be better predicted: By opening a new branch, what is the number of clients that are reallocated? How does this affect revenue streams per branch? In which area are additional ATMs profitable? Where do customers prefer a counter service? As a result, soft factors, which have always played a significant role in decision-making can now be quantified and weighted. This highly-granular geospatial analytics empower banks, to generate novel insights into a single branch and the entire branch network.

To compute data, Geospin deploys methods like machine learning, artificial intelligence, neuronal networks and predictive analytics. However human judgment and experience are of utmost importance to avoid, that the algorithms identify correlations, where no causally determined connections exist in reality. Thus the selection of data sets and interpretation of results by statistics experts are indispensable. In the end, algorithms require training and learning of key performance indicators and correlations within the particular context before they deliver robust answers to defacto questions.

Today Geospin offers its customers two different services: data insights and analytics as a service (AaaS). The AaaS product is charged based on a software license and customers can access the analytics tools by using an API for their accurate analysis. The Data Insights are individual projects to answer and train customized models which can be accessed by an API afterward. Contact Geospin today and enable your company to make better-informed decisions empowered by geospatial intelligence.

Leave A Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.