POC In Three Days: How We Did It

By Mateusz Ciesielski
Business Development Director

This post discusses a recent POC done with DataWalk in Europe.

For many vendors, Proof of Concept (POC) projects for advanced Enterprise software often take months. However, with DataWalk software, such projects can be completed in a couple weeks, or even just a few days. Here we discuss a recent example with a fraud detection use case.



Day 1

Kickoff for this POC was pretty much standard. Our goal was to do a POC in order to prepare the business case for the possible implementation of the DataWalk platform.

Our team consisted of our Account Executive, one Analyst, one System Engineer, and two Business Consultants. Until this time only the Account Executive had talked with the customer, and the rest of the team members were at the customer facility for the first time.

Our customer had two people participating at this point: the director of the fraud department, and an Analyst….and neither had much time.

From the beginning the work was intense. We first reviewed the customer’s current internal processes, as well as available data sets. We then diagnosed the customer’s current situation:

  • The number of detected fraud incidents varied significantly by line of business, and there was a need for improvement.
  • The customer had the expertise to increase fraud detection, but existing tools and processes did not allow for this to be put into practice.
  • There was a critical need for data integration in order to enable more effective work on fraud detection.

We then began work on the data model design. In the DataWalk system, modeling is done at the business level using business terms, and is intended to reflect how the customer perceives his business and wishes to address the most important and urgent needs. A key capability of DataWalk is a flexible logical data model that can easily be modified, such that there is no need to agonize about the structure of the physical model. With DataWalk, a first-pass model can easily be generated, and this model can then be easily, quickly changed as needs evolve and/or become better understood.

Day 2

We reviewed the data model design with the customer to ensure the design was on target. From our work the previous day, we already knew the types of data sets the customer had, and what the relationships were between data in different databases. With this we were able to correctly create the first-pass data model, which could now be directly modified in our tool.

At this point the customer was quite surprised; they had expected to only see the equivalent of a powerpoint slide, not a working data model in the application. This led them to quickly mobilize and obtain some of their real data (from a previous quarter), anonymize it, and provide it to us to load into the application.

This began a long and inspired discussion on the details of the model, covering both technical and logical business relationships between data sets, as well as the expectations of cross-database analysis of the data.

That afternoon the customer made their anonymized data available, and we put this data into our system, under the first-pass data model. By the end of the day, we had a unified view of all the data which had previously been in various data silos.

Day 3

On Day 3, the customer was again surprised to be able to see all their data, and the relationships between the data, on one screen! Business users immediately started to explore their data via the visually-oriented DataWalk system. After watching our team use the tool a bit, these business users were able to effectively use the system without training, and without any limitations.

To confirm our analysis, we asked for a larger sample of data. We also began detailed work on the implementation of specific rules for testing various hypotheses for detecting frauds.

By the end of the day, we had provided business users with the capabilities and the freedom to work independently, with data from multiple sources integrated under one visually-oriented interface. By the end of the week, the customer had already identified significant frauds that had previously gone un-detected.

How It Was Possible

  • The DataWalk platform has a flexible, logical data model that can be easily modified. We could constantly modify the model and adapt it to the customer needs, even when the model already was loaded with data.
  • DataWalk includes connectors that enable data to be easily extracted from various SQL databases, Hadoop HDFS, Microsoft Excel and various other sources, and be integrated into a single database for analysis. We could easily provide a unified view of all data and the links between those data elements.
  • The enthusiastic engagement of our customer was critical. They helped us understand their key needs in detail, and once they fully appreciated the possibilities of DataWalk, they aggressively engaged to obtain additional data and accelerate the process.
Umów prezentację!
Przesyłając swoje dane osobowe, zgadzasz się na otrzymywanie wiadomości od DataWalk SA. Możesz zrezygnować z otrzymywania dalszych informacji w dowolnym momencie. Aby uzyskać więcej informacji, zapoznaj się z naszą Polityką prywatności

Get A Live Demo!
Get A Live Demo!
Please note that by submitting you agree to receive messages from DataWalk. You may opt out of receiving further communications at any time. For further information, please see our Privacy Policy