Data Science Consulting Can Be Frustrating!
3 Things to Do to Fight the Frustration & Wow Your Clients!
When I first started consulting in data science, it was easy to get frustrated.
My first few gigs really had nothing to do with data science. They had a lot more to do with data analysis and research methods. A typical scenario was:
“Hey, I need someone to use data science to help me understand what is driving our cashflow.”
The result was a lot of PowerPoint presentations with spiced up visualizations of the data, showing how different variables in the data related to their key performance indicator (e.g. cashflow, bottom line, or however else they wanted to describe revenue).
The analytics were correlations, averages by groups, frequencies by groups, and various ways of visually presenting those rather basic statistics.
But then I realized something one day. I realized that despite the fact that their simple requests could often be answered with much simpler analytics better suited for a data analyst, I realized that slowly layering in my data science knowledge could be a real differentiator.
Combining skills I had acquired before I was a data scientist, as a researcher, helped me to set up the kinds of comparisons that would answer their questions. Layering in some more advanced analytics but explaining them with simple visuals, would then really wow them and keep them on as my clients.
So if you find yourself frustrated by the time you spend performing rather basic data analysis, consider spicing it up a bit with these 3 data science strategies:
Blend with open-ended data
Natural language processing is a powerful data science tool that can unlock value from unstructured data. Showing clients how you can blend with data sources that are unstructured to help add more insight gained from customers can be a game changer.
Imagine being able to inform your client that not only did product A, B, and C drive the most revenue but that customers are also telling other customers that they love how product A, B, and C can do X, Y, and Z for them.
Unpack a predictive model
Although a predictive model may not have been requested by the client, building one can help to uncover key features that drive particular outcomes or KPIs. This technique may be particularly powerful when combined with NLP derived features from unstructured data sources.
In other words, it’s one thing to say that product A, B, or C drove revenue and another thing altogether to say that across all products, features X, Y, and Z were the most important predictors of revenue.
Provide a different way of looking at the data
Finally, data science and data visualization go hand-in-hand. The reason for this relationship is that data science can be difficult to comprehend and so visualizations can help to capture some of the complexity in ways that words simply cannot.
It’s always a good idea to leverage different visualization techniques to help “see” data in new and possibly valuable ways. For example, heat maps set up correctly help to focus the eye on the most important relationships. Or treating your data as a network graph may help to visualize useful communities or groups of relationships among variables.
Like engaging to learn about data science, career growth, life, or poor business decisions? Sign up for my newsletter here and get a link to my free ebook.
