Data Science Methodology
a Data Science Methodology structures your project
Getting insights out of the data, that’s what it’s all about in data science. After we have defined the business goal you try to solve, our data scientists jump in, try to get the data and start their process.
We adhere to a standard Data Science Methodology including the following steps:
We translate the business problem at hand into data and identify the right data sources and fields. We sit with the BI and data teams and jointly define the extraction and load methods we can leverage to get access to the data.
Before building the model, we spend sufficient time on Data Discovery or running various discovery analyses to challenge or confirm initial hypothesis. We get to know the data, iteratively provide feedback and thus crystalize the case at hand of what we want to predict and how we can best do it.
In this phase the actual modeling takes place. Our Data Scientists do feature definitions based on our disovery analyses and run and compare different predictive models, evaluate their performance, finetune and rerun. This is an iterative process where we iteratively communicate feedback and performance back to the client.
Fine Tune and testing
The model needs to be relevant and actionable. Therefore, we set-up a fine-tune phase where we go, with the business, through the results and where possible even set-up a real-live test. Furthermore, we help you to embed the resulting model within your business processes so you can fully leverage the newly created model. Every Proof of Concept delivered by our team results in an actionable working process which can be leveraged for a set of weeks of months providing real added value.
How we use it
All our proposal and projects are build around this methodology.
Based on our experience, this methodology is giving us a structure to tackle a problem step-by-step and to deliver, at every stage of the project, clarification and alignment to the project team.