Graham Collins, FCMA, CGMA, is a vice-president of finance and economics for one of the world’s largest energy companies. He has worked there to create new roles that bridge the gap between finance and data science — and he is set to share lessons from his experience at the AICPA & CIMA Corporate Finance and Controllers Conference 27–29 October in Las Vegas .
“I’m looking at how we transform the finance function to take advantage of the massive progress we’ve made in digital and digitalisation, the simpler coding languages and cheaper cloud storage and compute power than we’ve seen in the past,” Collins said. “And that actually unlocks a huge opportunity for finance functions.”
His presentation is meant to guide finance professionals on their journey from “Excel to Python” — in other words, to help them embrace new data-science solutions that may be more automated and more customisable than legacy software.
“What we’re finding, more and more, is that there are digital solutions that you can actually put on top of your ERP system, which actually take that data in a much more powerful direction,” Collins said.
There’s a vast array of new options, from custom coding to powerful database tools. But finance professionals don’t have to become expert programmers or data scientists to succeed in this new environment, Collins said. Much of that expertise will still reside within IT and data-science teams.
Instead, he said, “the role that we’re particularly trying to evolve is the translator — someone who can translate between the business and the IT department.”
Data translators can help to identify a business problem, “and then they can work with the IT department to help identify and design the IT solution”, Collins said. “They understand the language of the IT people. They understand what code can and can’t do — what data can and can’t do.”
Follow these six strategies to develop data translators on your own team.
Build a broad base of knowledge
A translator is not a top-tier expert in data science. Instead, Collins said, their skills are anchored in finance, but they are broadly familiar with emerging technology and may have some direct experience with programming.
Collins suggests that experimenting with coding can help to build that familiarity for individuals. Self-directed online courses range from introductory to advanced. Progressing through those courses may not make you a master programmer, but it will teach you to speak the language and understand the capabilities of data science.
Collins suggests treating data and programming skills like a musical instrument, something to be developed over time with consistent practice. “You need to keep going with it,” he said. That could include developing projects at home, such as automating your personal finances.
As you become more familiar with the technology, it will become easier to discuss and navigate in the workplace setting.
Start with a problem
“Companies have made the mistake of setting up whole data science departments because ‘that’s the thing we need to get in’,” Collins said. But that’s a recipe for failure, he said.
A data-science effort, no matter the scale, is much more likely to succeed if it begins by focusing on a tangible problem.
Some common topics include:
- Automation: If your team frequently exchanges spreadsheet files between multiple people via email, it may be time to consider an automation effort. “As soon as you’re thinking of scale — is it going to be used by multiple people on a standard basis? Excel very quickly gets to a limit of what it’s able to do,” Collins said.
- Analysis: Economic modelling often can require painstaking gathering and collation of data — a process that can be vastly improved with automated sweeps. Once a solid foundation is established, companies can also layer on machine-learning solutions to deliver predictive analytics.
By understanding the capabilities of data science, translators can identify problems and inefficiencies that others do not.
Take an experimental mindset
Once you’ve identified a problem, start thinking like a scientist. There won’t be a single “correct” solution. Instead, you’re setting out to explore different possibilities and see what works.
“Data science is very much about experimentation. You’ve got to be very much prepared to fail,” Collins said. “You’ve got to recognise this is not going to work the first time — but by doing that you can head on the right path.”
The translator won’t be expected to know or implement every possible solution, but having familiarity with emerging technology — especially through outside training and practice — can help you understand the possibilities.
Find your allies
Data-science projects often require the cooperation of multiple departments, and a translator can build that accord.
It starts within the finance team, where Collins suggests searching out people with a “natural affiliation” who can be “allies and champions” for an automation or AI-based project.
“They may not be coders, but they have some background,” he said, whether it’s a knowledge of statistics or a programming hobby.
The translator also knows how to win support from “the technical side”, Collins said, whether that’s from data engineers, contract experts, or the IT department. Often, those teams’ resources are in high demand.
To win their support for a project, Collins said, go back to the original problem that defines the project. Clearly specifying the problem can help to convince others that a project won’t be an endless time sink. Crucially, it can help them understand why the project should be a priority.
“When you step back, make sure that you are addressing the issues that are most valuable to the company,” Collins said.
Learn about data culture
Starting individual projects is only the beginning. As you gain traction, you’ll encounter a question familiar to companies large and small: How will your team manage its new uses of data?
“In some companies, it’s not always clear who owns the data,” Collins said. In other words, it’s often unclear who is responsible for managing and maintaining a particular source of data. Who will ensure it is properly formatted, up to date, and accessible?
The rush to embrace data can grow increasingly risky, Collins said, if people begin to freely introduce customised software and solutions. Sometimes, downloading open-source packages of code from the internet can open new cybersecurity vulnerabilities.
“That introduces risk to the company,” he said.
These questions of data culture are often best addressed by policies set at the leadership level, but everyone working with data has a responsibility to consider them.
“As we all are upskilling ourselves and expanding our skillsets, we’re obviously stepping into new areas outside of traditional finance,” Collins said. “You may be an outstanding CPA or FP&A analyst, but you’re getting into spaces where we don’t have five, 10, 15, 20 years of experience. You need to make sure that you engage with the experts. It needs to be very much a collaborative approach.”
Hire and manage for data
Finally, Collins suggested ways of staffing these “translator” roles. With extreme demand for tech skills in the labour market, it may be easiest to look internally.
“Actually, upskilling your own people is a great way to do it,” he said. Whether hiring or training, look for those people who have shown an interest in data tools, including visualisation software like Tableau. “These are people who have an analytical mind and a kind of digital savvy.”
From there, encourage those team members to engage in the cycle of learning, experimentation, and collaboration described above. Allow them to try out new skills in automation and analytics, and give them permission to fail.
But he added a warning: “Remember, we’re not trying to make the finance function the IT function. The IT function will still have the data engineers or data scientists,” he said. Instead, finance should be looking for people “who are adding a tool to the toolbelt, but it’s to add to their finance qualifications, the finance skills that they have”, he said.
And when hiring is necessary for these new data-oriented roles, Collins said that a hiring plan is mandatory. Simply advertising a data-related job is unlikely to work, given the competitive market.
Instead, he suggested specifically identifying universities and building relationships, searching out those finance-anchored but data-savvy candidates. “You need to go that extra mile,” he said. “Just advertising on the web is probably not going to get you the quality of people that you want.”
— Andrew Kenney is a freelance writer based in the US. To comment on this article or to suggest an idea for another article, contact Neil Amato, an FM magazine senior editor, at Neil.Amato@aicpa-cima.com.