A training programme on data organisation and visualisation within the pharmaceutical industry, that I shall be hosting with a colleague, has forced me to reflect on the current state of play within the commercial departments of pharma. Dashboards that display KPIs have become commonplace but I wonder whether the analytical engine that drives those KPIs is actually fit for purpose. The development of visualisation software has been positive but have the pretty dashboards given a spurious legitimacy to the analytics that underpin them. In other words have we become so obsessed with the visual output that we fail to question the robustness of the analytical input?
Observations from various commercial and data science conferences suggest that there remains a strong reliance on Excel as the analytical engine for many dashboards. Excel, while being an excellent spreadsheet package is not suitable for statistical analysis of most pharmaceutical data. The chart at the top of this article shows quite graphically, albeit stylistically, how one could generate strong correlation coefficients from a dataset and still get the wrong model. Exploration and visualisation of your dataset is necessary before your choice of model to apply to the data. Is this done routinely or have we become reliant on Excel and visualisation software to drive some of our most important and expensive commercial decisions?
If you are involved with this area can you answer a few simple questions to help shed more light data exploration? Your thoughts on even a few of these would be helpful.
1. What software do you routinely use in your office to present visualisations of sales and marketing data AFTER analysis?
b Power BI
d. some other visualisation package – please name
2. What software/techniques do you routinely use, if any, to visualise and understand your data before BEGINNING your analysis of sales and marketing data?
3. What are the five most common problems you have to deal with when starting your analysis of sales and marketing data downloaded from your CRM system?
4. How do you deal with data cleaning eg duplications of customer names.
5. What relationships between input variables (eg face to face calls, group lunches, clinical conferences etc) and sales per postal brick do you normally expect to see? Eg straight line, Vmax, exponential, s-curve, cubic, other polynomials?
6. What are your main frustrations with respect to providing/presenting data to senior sales and marketing management?
Stewart Adkins was a Pharmaceutical Analyst at Lehman Brothers for 23 years and was involved with the Pharmafutures projects.
Pharma Forensic Limited. Registered in England with Registered number 08647917.
Registered office The Courtyard, Pershore, Worcs, WR10 1AA.