Data science is a multidisciplinary field. Its strength comes from the diverse backgrounds of both technical and non-technical teams, working together to solve challenging business problems. To achieve great outcomes, extensive collaboration with both technical and non-technical colleagues is essential. Over a series of three blogs, we’ll present some of the key motivations for ensuring your data science teams work collaboratively.
In this first post in the series, we’ll examine the interactions between data science teams and other internal stakeholders, such as those in Customer Success, Marketing or Finance functions. In blogs two and three, we will then look at collaboration within a data science team and with external data scientists, respectively.
Great data science work only creates meaningful business value when it leaves the lab and is completely integrated into day-to-day working life for end users. Data science teams can’t hope to design and implement these solutions in isolation; they must draw on the expertise and experience of colleagues from around the business.
There are three driving forces behind the need to work closely with non-technical stakeholders during a data science project:
- Without the right buy-in, your solution may never be commissioned
- Effective AI solutions need user and organisational context
- AI must deliver tangible results
Without the right buy-in, your AI solution may never be commissioned
Solving difficult problems using AI is technically challenging. But it can be exhilarating when it all starts to come together. However, data science teams exist in the real world; budgetary and time constraints demand that new projects present a compelling business case for their sponsorship. Even in the most innovative of firms, an exciting AI solution won’t receive sign-off unless non-technical stakeholders in the business can also see its value and are willing to provide their support.
When evaluating potential projects, data scientists should consider more than just the technical feasibility of the task. They should also ask: ‘What creates the most impact for the business?’. My colleague, Jonny Howell, has written previously about how we assess business needs at Faculty and you can read a more detailed overview here.
In the context of collaboration, however, data scientists cannot precisely define these business opportunities alone. Prospective end users and department leaders will have a much more refined understanding of their day-to-day pain points. Working with them to tweak your proposed solution and jointly present your case to management will make for a more effective pitch.
Effective AI solutions need user and organisational context
AI is a profoundly important technology because of the sheer variety of its potential applications across all functions within organisations. In order to increase AI’s impact across the economy, data science teams need to design and implement solutions that are either integrated into existing workflows or fundamentally change these business processes for the better.
End users are best placed to guide the design of these solutions – they will be the ones using the final product day in and day out. Nobody will better understand the intricacies of interacting with a fraud anomaly detection system than a financial analyst. The nuances of responding to customer queries from a triaged list are locked away in the experience of your customer service agents. And the pitfalls of integrating customer churn predictions with parallel marketing strategies are etched in the memory of your marketing team. There is no substitute for speaking to the end users to understand exactly what will make their lives easier and what the priority items in any solution should be.
Your data science team may not have dedicated UX expertise to draw on, so put on your design thinking hat and work with your end users to design a solution with their interests at the heart of it.
AI must deliver tangible results
We believe that data science and AI are only valuable so long as they’re creating change and delivering results in the real world – not in the lab. That means we have to do everything possible to ensure that our solutions deliver ROI.
We’ve talked about collaboration in the setup phase, but it’s equally vital in the execution phase to ensure that AI delivers results – and keeps on delivering them. That means data scientists and end users must be in near-constant contact throughout execution. End users must be able to highlight problems as they arise and give feedback, while data science teams need to have consistent, up-to-date information about the performance of their models.
It’s rare for a solution or strategy to be completely perfect the first time around; collaboration allows data scientists to do the kind of fine-tuning required to ensure that models perform to the best of their ability.
Collaboration is a word frequently thrown around in modern business interaction. Hopefully this quick introduction has shown you that, as technical experts, working closely with those in other departments and teams is not just a nice-to-have, but a necessary component of producing impactful results.
In the next part of this series, we’ll discuss why fostering collaboration within data science teams is vital to effective, efficient, and forward-thinking data science projects.
We built Faculty Platform, our data science workbench, to enable collaboration of all kinds – particularly the clear, rapid, and systematic transfer of data and insights between data scientists and external teams.