Francine is a data scientist and CEO and co-founder of Mastodon C. She's a trustee at DataKind UK, a board member at Bethnal Green Ventures, and advises the Cabinet Office on data science.


Seven steps to starting data projects on the right track


© Luckey_sun 2012 big-data_conew1

Only 13% of data projects reach completion, and of those that do, only 8% of leaders report being satisfied with the outcome.

It’s not news that executives in commercial organisations are taking advantage of data to be more competitive. City and local government leaders are starting to do the same, to achieve their own complex, and arguably more important goals. From using data models that predict the impact of different policies, to deploying data platforms that improve collaboration, the tools have never been more accessible, or more needed.

But starting data projects isn’t easy. Getting results even harder. Only 13% of data projects reach completion, and of those that do, only 8% of leaders report being satisfied with the outcome. To stand a chance of making an impact, it’s therefore essential that data projects are kicked off in the right way.

Work by Mastodon C, a data science firm that helps local authorities benefit from data-driven approaches, suggests that leaders should consider these key areas before getting data projects started:

  1. Scope for impact. Before doing detailed planning leaders should first check that they will be able to take meaningful action based on the results. Ethics should also be considered at the outset - can the project be delivered without adversely impacting a particular group or infringing anonymity?

  2. Refining the question. Many data projects start with what’s perceived as “interesting data” and a misguided belief that insights will be gained by doing something to that data with machines. The most effective projects start with a clear question, how the question might be answered, and how the answer might inform useful action or decisions.

  3. The business case. Data projects should focus on delivering results that align to strategic goals. Which means being clear how action will create a useful impact. Good business cases are also developed collaboratively - with leadership peers and relevant departments. This builds support, reveals potential obstacles and focuses the project on things people want and need.

  4. Agile working. Data projects have a better chance of success if they start small and deliver an initial impact in weeks and months, rather than years. That might mean reducing the scope, but scope reduction in itself can help achieve better results by making things simpler. It’s better to learn fast from early insights than attempt to crack everything in one go.

  5. Analysis. A key step in the process is to determine that the intended analysis method is fit for purpose. Project teams should search for similar projects and write a clear description of why the approach is appropriate. Conducting an initial exploration of the data will help test and refine the analysis approach.

  6. Data requirements. As the core project resource, the data itself needs careful consideration. Is it easily accessible or will it be sourced externally? If it’s sourced externally is it open data that anyone can use or does it come with restrictions?

  7. Skills and infrastructure. It’s important to check internal capacity and capability to support the project. To get meaningful results the project will need an appropriate infrastructure to support it - from access to tools, to having the right people with the rights skills available.  

For a more detailed free guide to starting a data project, download “Get Started With Data”, Mastodon C’s practical checklist for planning data science projects.


Back to homepage