At Modelyzr GmbH, we always keep an eye on the latest future technologies and are inspired by the new possibilities they offer. This naturally includes AI applications and machine learning processes. The digital experience agency kuehlhaus was also interested in this, so they invited our founder and CEO Nils Niehörster for a podcast interview.
Our data analytics engine MODELYZR is based on the SAP HANA high-speed database and uses powerful algorithms to refine company data into meaningful results. Here we present excerpts from “brennstoff”, kuehlhaus AG’s podcast format, in which kuehlhaus host Clemens Weins talks to Nils about a clean database, among other things.
Clemens Weins: Nils, how do you ensure that the MODELYZR works with clean data?
Nils Niehörster: When we go into companies, we often hear that the data quality in the connected system, for example in CRM or marketing automation, is too poor. However, one of the strengths of pattern recognition systems is that the data doesn’t have to be perfect. Today, algorithms are able to eliminate the noise – i.e. faulty data. The strength of MODELYZR is therefore that it can work with data that is not perfect and extract the most probable truth from it.
This creates a picture that is as clear as possible: for example, the result of a query could be that you are dealing with 142,000 companies that represent a certain market. It is possible that there are actually only 141,000 companies, leaving a certain amount of blurring. If you then take a closer look at these companies, it becomes clear that they belong to 18,000 groups, 15,000 of which offer selling opportunities.
It doesn’t matter to marketing whether there are actually only 14,800 as long as the error rate remains below ten percent. Marketing only needs to know in which target market the best potential is hidden and can then work well with it.
However, this is not enough for sales. This is about a targeted approach. At this individual company level, the MODELYZR does not play out certain information if it could be misleading.
The MODELYZR therefore removes the noise and thus sharpens the result. Data staging, i.e. the merging of data sources and their cleansing, is state of the art today. Other applications are also capable of this. Collecting data, i.e. connecting the various data silos in companies, is only the preliminary work.
Clemens Weins: Data staging is the preliminary work, but what is the secret of MODELYZR?
Nils Niehörster: The “magic” is that our system can recognize corporate structures. Knowing that Allianz, Daimler or UBS are not just one, but hundreds of companies that belong together. And that there are branches of companies that are independent consumers. Which of these are allowed to make independent decisions and which are not. So the trick is to aggregate the raw data into a market model.
Clemens Weins: What happens structurally in a company when the introduction of MODELYZR means that the decisive factor is no longer many years of experience, but data that is available to practically everyone?
The structures have to be adapted accordingly. If I introduce a new tool but don’t change anything else, I can’t expect anything to improve. It is crucial to adapt your own processes to new possibilities.
For example, if a company uses a machine learning system that predicts the probability of turbine failure, it is no use simply extending the service intervals. Instead, the service and support organization, the associated processes and ticketing must be adapted in order to make optimal use of the new possibilities.
If a tool such as MODELYZR is used, there is no point in instructing marketing to sort the data according to “most likelyhood to buy” and then initiate the mailing as usual. Instead, the processes need to be adapted right from the planning stage.
Planning is usually based on the past few years. So you look in the rear-view mirror and want to deduce from history what the future might look like. If a system is now available that can show what potential can still be tapped, planning must also be adapted. Planning then no longer means estimating, but segmenting, uncovering whitespaces and knowing where a product is still undersold and where the market is already saturated.
And so the data-driven approach runs through the entire process chain, from planning to execution and after sales.
Listen to the whole episode We would like to take this opportunity to thank kuehlhaus AG for the invitation and the exciting conversation with Clemens. If you would like to listen to the entire episode “Man vs. machine” of the kuehlhaus podcast “brennstoff”, simply follow the link.