Lack of knowledge main reason for Knowledge Graph POC failures

A lot of problem today can  easily be resolved with a Knowledge Graph, and out of box  technologies do exists today. Our favorite technology for such problems is the Resource Description Framework (RDF), often referred to has Web Semantic. It relatively simple yet extremely powerful and best of all it's an industry standard.  

Unfortunately this or other graph data store for that matter are often set aside in-lieu of building an in-house custom graph engine using their favorite programming language  over a relational database and in memory cache.  Why? Ironically the reason is because of lack of knowledge.

For example, I was invited to an architecture review for an application to resolve a problem that was a great use case for using RDF. As you can guess, their solution included building an in-house custom graph engine over a relational database. The first question I ask was how much effort to building this custom engine logic, to which they answered 4 weeks. My second question was, why not use RDF or even another graph technology to resolve this.  This is when it was revealed that they had indeed performed a proof of concept (POC) with less than acceptable result.

Rather than dismiss their finding, I requested whether I could look at their POC. The first thing that was apparent was that the person that build the POC had just been introduce to the technology used.  I then proceeded to understand their data around the problem they wanted to resolved. It took around 3 days. In the following 2 days, I proceeded to build the logic that would return the result on par with what would be expected from the custom software. 

At this point I had showed that the logic could be built 20 times faster without any custom code and the best of all the query was easy to read and very intuitive even thought I had elected to use a federated approach to minimize the need to ETL the data. In other word the cost of operation/maintenance will be much less.  I had also indirectly showed the main reason why the POC had failed. It was not because of technology but rather because of a human, They simply did not  have the proper knowledge to use it properly. 

It extremely important to have the proper knowledge when implementing a POC to evaluate a new technology . For this reason, we at MindStep offer corporate training on Web Semantic (SPARQL,OWL) as well as a  service where our architect assist your own architect get started on converting your data stores into knowledge. 

We invite you to visit our site and our blog on a regular basis. We will publish articles periodically detailing how you can transition from disparate data sources to a single federated information store and/or Knowledge Graph.

If interested in our training or if you have any questions or feedback you would like to share with our team, simply contact us at info@mindstep.ca. 

We look forward to staying connected.