Putting the data into data science

The opportunities for data science, internet of things, machine learning and artificial intelligence are well documented. Process automation, predictive forecasting of events, driverless cars, website personalisation, stock control, health management, transport efficiencies and so. So how do you choose? Which is priority? Who decides?

The challenges with finding the right people with the right mix of technical, data, commercial and mathematical skills is also well documented. It’s a team sport not an individual pursuit. High price to capture the best. Expensive training. Location. All make it challenging to build the type of team required to exploit the opportunities. So what’s the right team to build and how? Do you need a Chief Data Officer to sort this?

There is also truck loads of technology out there to choose from in order to access, gather, store, sort, analyse, visualise and publish value from data. The IBMs, SAPs, Oracles of this world have pivoted their business models. Newer but now already established capabilities like Cloudera, Mapr, Hortonworks, R are re-writing the big data and analytics space. The tech startup community is now full of data science, machine learning and AI generalist and niche technologies to go after some of the world’s biggest challenges. So which do you choose and when?

Data. We aren’t short of that. Traditional internal business data sets like sales, stock, logistics can be augmented with tricky data sets like web clickstreams, social media posts, verbatim, RFID movements and so on. External data sources like weather, census and flight paths. Newer usable data like machine logs, wearable movements, geolocation.  Loads to use to throw together, find patterns, understand consumers and identify new business opportunities.  So which sources are right to use and what order should you get them?

As a Data Scientist, Quant, Operations Research Analyst it should be easy to identify the right opportunities, get the data and solve the problems. However, I hear time and time again the difficulties with rapidly getting value from data and delivering the analytical products needed. Why? Mostly I hear it’s because there is no coherent strategy across the business helping to answer the questions from above. 

The key is defining a framework that works for your organisation that means you can continue to deliver value in the short term whilst building towards a set of capabilities that allow you to quickly react and respond to business changes and priorities. 

This framework should allow you to:

1. Identify the right business challenges and opportunities where data and data science can help and there is likely return. Obvious to say but should align to business objectives (not just the most interesting thing to go after!!)

2. Construct a well-rounded team (physical or virtual) that can build and support a data platform, ingest data in a consistent and repeatable way, build data models and run analytics. 

3. Articulate the value of data and communicate the upside being created from the activities you work on. This creates buy-in, wins hearts and minds and turns complex problems into consumable internal messages of success. 

4. Choose the right technology for data management and storage, general data science and specialist solutions for certain problems. There is no 1 size fits all so align to what objectives you are trying to achieve.

5. Deliver in an iterative and agile way to release value as often as possible. Increments of improvement are fine.

This framework then becomes your strategy for success and a way to demonstrate value, continually improve and have enough flexibility to adjust and react to the performance of your business and unexpected events.

Cynozure helps organisations get value from data through business and technology strategic advice and solution delivery.

Published in

Leave a reply

Log in with your credentials


Forgot your details?

Create Account