Hunting unicorns

Exploiting data as a resource for insight not only needs high levels of skills and subject matter expertise, it requires the bringing together of a number of distinct roles – not all of which are tied up with statistics or computer science abilities.

In fact, the key problem is how to link these core skills with the problem we are trying to solve, and more specifically with the people doing the solving. To do this requires two roles: translator and storyteller.

The role of the translator is to conceptualise policy problems in the language of data and analysis, working out how available data could be used to understand the problem, identify potential causes, describe relationships and suggest solutions. The role of the storyteller is to tell compelling stories using the data, explaining why the results matter and how they can be used to improve policy and implementation, debunking myths along the way.

The ability to play these roles well is not necessarily widespread among people with very strong programming and statistical skills. Indeed, the stereotype of people with very strong programming and statistical skills is of a relatively stunted ability to tell stories. While the rise of data journalism (of which New Zealand has some superb examples) to some extent challenges this stereotype, the occurrence of single people who can play all the roles effectively is so rare that the joke about them occurring as commonly as unicorns applies.

Given the rarity of naturally occurring “unicorns”, how best can we make sure that all the required roles are being filled? Beyond bromides about “good communication” (which as advice goes manages to be entirely correct and entirely useless simultaneously) some approaches that organisations have adopted include the following:

1. Recognising the different strengths already in data teams and bring bringing together project teams based upon complementary strengths. 
2. Establish teams with different skill sets and roles but require them to work together on projects. Thus, the teams have a natural similarity and have the opportunity for specific and targeted sharing of knowledge, while the demand that they work with other teams of experts on a problem requires an ability to work across teams and development of respect and appreciation of other people’s skills, and the ability to bring them together.
3. Building the capacity to undertake especially the translator roles into the data scientist’s skill set – this is undoubtedly essential, but leaves the story teller role out so some variant of 1 or 2 would need to be followed. 

In fact, the diversity of successful approaches points to an important fact that there is no one right answer – the roles must be done by someone, and the roles must be understood, respected and well-coordinated by the manager – but precisely how they are best integrated depends upon local cultures and ways of working. This undoubtedly a good thing – it’s much easier to innovate what we do when we are not having to bring in entirely new processes and structures at the same time.