Title: Uncertain Data Science Subtitle: Does your business need Data Science? Date: 2018-09-03 10:20 Category: advice Tags: uncertain, ds, advice Authors: Varun Nayyar Status: draft

When I tell people what I do, a lot of people are interested. Data Science, Analytics, AI and Machine Learning have been buzzwords for the past few years and many people not in this space are wondering at it’s possibilities in their workspace. I’ve seen many applications of DS outside tech, a few examples being

  • Medical: Automatically interpreting X-Rays to diagnose TB

  • Banking: Estimate the interest rate on a load for a house and buyer

  • Charities: Estimate which areas of a city will result in best return from phone calls

  • Building Management: Estimating energy usage of building and building autonomous control systems

  • Retail: Recommending products to shoppers using loyalty card data. Or putting diaper’s next to beer

Tech companies like amazon, google and facebook have obviously led the way in Data Science and most new tech focussed companies are built with data collection in mind from the ground up, but there are many non-tech companies deciding whether or not they need to get serious about data. Generally speaking, most companies are small outfits that have a very direct contact with their customers and not many employees, and data science isn’t very helpful without scale. In many ways, data science can be seen as a way of extracting greater efficiency from existing processes. Prior to DS, let’s compare how the above situations would have been approached

  • Medical: TB xrays would have been sent to an expert to interpret. The number of xrays that could be handled would be limited by the number of xray experts on hand. For charity organisations, training people of skill is difficult and there is always the danger of talent being poached/churn and rendering the time investment null.

  • Banking: A manager who knew the area and the people would use compiled tables to estimate the risk and thus interest rate applicable in this situation. The tables would be compiled by head office collecting data, but there was significant leeway allowed. My parents spoke to bank managers who gave lower rates, but these managers no longer exist.

  • Charities: People would simply target areas with high foot traffic when it came to collection buckets and focus on calling existing donors. Volunteers would suggest areas to target with phone banking and cafes/businesses to send material.

  • Retail: Again managers would be keeping an eye on foot traffic and seeing what products sold together. Also seeing which areas resulted in congestion and moving aisles around to ensure more efficient movement.

What Data Science has done in the above examples has really removed the need for medium skilled employees/middle management. There is an argument that the inference is greater than what could be done by humans (charity being biased to locations they know of is an example) and that they can handle far greater amounts of data than otherwise possible by human (fraud detection in credit cards), both of which are fair arguments, but in general these benefits are only obtained later in the DS lifecycle, the initial benefit is the reduction of labour required. This can be felt as an indirect benefit too, in that new staff is not required to scale up business, or many tasks that are time-sinks are now much faster which allow more of an employees time to work on more challenging problems which add value to the company.

Do I need Data Science Checklist#

If you’re asking this question, the answer is probably. But an incomplete checklist asks if you can benefit from Data Science:

  • Do you feel like you make decisions based on your experience and instincts?

  • Are customer faced decisions being made by people other than yourself? And are these decisions less than ideal?

  • Do you questions you need investigated? Perhaps systemic problems or estimating how price changes affect customers?

  • Do you have a need to apply domain knowledge at a large scale.

  • Do you have data but no idea how to interpret it or use it?

  • Is a significant amount of time spent on data collection/cleaning?

  • Is losing an untalented but experienced worker worse than an inexperienced but talented employee?

These are usually signs that your organisation/company is beginning to need a Data Scientist. Combined with an issue of scale

And another checklist to see if Data Science will be effective from the get go

  • Do you have a strong IT/tech system in place already?

  • Can you affored the prevailing salary in your area?