Five tips on how not to build a data science team

The business value of having a strong analytics capability can be transformational, especially when insight is used in a timely manner to support effective decision-making across an organisation.

While a small majority of organisations seem to have cracked the code and have capitalised on the promise of data and analytics, most are still finding their way, and many remain unsure even where to start. Though data maturity has almost doubled in the last four years, it remains low in general terms. A 2021 study by NewVantage found that only 39% of executives believed their organisations managed data as an asset, and even fewer (24%) viewed their companies as being data-driven. There isn’t a strong reason to think that these numbers have significantly changed over the last year. 

Not having strong data insights and analytics capabilities is a lost revenue opportunity – and, in today’s economic environment, this is not sustainable.  So, what should business and technology leaders do to rectify the situation? For most, establishing a data analytics or a data science practice is the first step. 

The stark reality is building an analytics capability is not quick and easy. In fact, it is hard work that requires investment. Perhaps not surprisingly, many organisations rely on a partner to fast-track time to value, which is an effective way to shorten time to insight and value creation.

For many, the most difficult part of this equation is not buying, integrating or using technology such as Tableau, BigQuery, Apache Spark or PowerBI.  In reality, it might be where problems start, according to studies – between 60% and 87% of data analytics and data science projects fail to make it to production. It’s the people.  Finding the right talent can be expensive as organisations compete to build their data insights and analytics capability.

As the Head of Data Science and Machine Learning at Xiatech – and as a long-time lecturer teaching the next generation of data scientists – I have been fortunate to help set-up data teams around the world. In my role as an advisor and practitioner in organisations big and small (private and public), I have witnessed many attempts to build data insights and analytics.

While there isn’t a magic recipe for success, I have seen what definitely doesn’t help. I have decided to compile the most damaging pain-points and causes of conflict, so you don’t waste precious time and money, and that you aren’t included in the Gartner’s static that only 20% of analytic insights deliver business outcomes.

1. Not having clear business objectives

Talking technology first and business outcomes second is a sure sign of potential failure in the future. The purpose of a data insights and analytics team is to serve the business in making better and faster decisions, using the data available to them. Many executives make the mistake of shaping data analytics projects to confirm hypotheses they already assumed as valid – introducing a bias on the expected result from the beginning. Unfortunately, this approach is incorrect. It increases the likelihood that your analysts only use data you already had, and your project will “reveal” insights into issues your company has already identified or place focus on areas of improvement that should not be prioritised.

Before investing in technology, organisations should step back and identify the business problems or opportunities they seek to address through data analytics. This can be some vague long-term roadmap, or a clearer bigger picture of the horizon where the organisation wants to walk towards. Only then should a plan be developed to build the analytics capability, looking at what skills and tools are required for success.

2. Not spending time on data quality 

67% of business decision-makers aren’t comfortable basing decisions on data pulled from their current technology.  Yes, that is a very high number which is concerning for anyone who is responsible for building a data analytics capability. 

Having your organisation’s data in one place does not mean that it’s ready to be used. Actually, this data might be pretty useless in its raw form. For example, some text-based inputs or addresses might need to be cleaned up (to fix typos, grammar mistakes or postcode formats), or additional information to be extracted from them (such as keywords from text; or the area, region or state of a postcode for example). Also, there might be corrupted records or entire fields that are causing analyses to be incomplete or false. 

I have seen organisations spend large sums of money on building some of the most beautiful dashboards but failed because of poor data quality. This leads to not only poor decisions but damages your reputation as data leader.

3. Not having the right data culture

Sometimes, even with all of the right ingredients in place, you might find yourself in a position where it is very difficult to fully unleash the power of your data, just because your organisation is not mature enough. 

Maybe some employees need upskilling, or some unhealthy habits might need to be removed or polished before committing to insights, analytics or machine learning. Manual reporting generation, data compiling, and other easily automated tasks, usually performed over spreadsheets, should be replaced and automated by other modern technologies. Otherwise, there is a big risk that old practices are carried over into the new world, leading you to needing more resources to just produce more insights or work with more data. 

Not focusing on creating the right data culture in your team, and across your organisation, will result in not generating the value you and your colleagues expect in the long-term.

4. Not integrating data practices into the rest of the business 

While only three of the 10 most valuable businesses were actively taking a data-driven approach in 2008, that number has risen to 7 out of 10 today. From Apple to Microsoft, from Facebook to Amazon – they all rely on data to drive their key decision-making processes. They have also created a data-led culture. 

Failing to integrate data and data best practices across your organisation will result in not creating a data-led culture so important today. In practical terms, this means:

• Integrating your analytics team into the wider business so they can help colleagues improve data literacy – the ability to read, write and communicate data in context.  
• Integrating your team’s analytics processes into the running of your organisation so data and insights flows to systems and people who make the decisions.
• And in the same way, having ideas almost constantly flowing from the rest of the business into the data team so it can shape and refine projects, making sure that the outcomes coming out from the team are useful and relevant.

Those three points talk about the connectivity between the data analytics team and the rest of the business – this is important not just to ensure short-term success with the first data projects delivered, but also to implant a data-led culture across the organisation that can redefine how the entire business works.

5. Not relying on a community of experts

We rely on the expertise of others every day in our lives: doctors, plumbers, electricians, engineers. You should be turning to the flourishing worldwide data community for support – and they want to help you. 

From freelance data architects, analysts, scientists and ML engineers to organisations that provide consulting services and technology companies that offer off-the-shelf data platforms that include holistic solutions combining data integration, insights and advanced analytics, there is no shortage of expertise. These individuals and companies can help you to accelerate your data roadmap, so feel free to tap into the community and let them inspire your imagination to discover the infinite possibilities of using data to create value in and for your organisation.