Companies know data analytics is powerful, but nailing the right approach from the get-go is the real challenge. How do you ensure you're set up for genuine results, and fast enough to keep everyone on board?
I recently sat down with Lydia Monnington for the Data & AI Podcast. Lydia’s worked in Data Analytics for more than a decade, so as well as talking about what data analytics actually look like in the real world, we discussed how automation is reshaping the role of the data analyst, plus why companies should prioritise analytics before jumping into AI.
Read on to find out some of the key takeaways from our chat and listen to the full podcast here.
Data analysis can be used to inform almost all decision-making, yet many organisations face challenges such as poor data quality, siloed systems and regional disparity.
Take the sign-up funnel, for example, where data can let you see exactly where people are dropping off. These invaluable insights can allow the software engineering or user research teams to look closely at every step of the journey to determine how it can be improved, helping companies to see value quickly.
Ideally, this should be someone relatively senior who has at least some experience of data engineering and analytics. It’s one thing to understand the data quality, but it’s just as important to have someone with the business context so they can get involved in the modelling and making the data ready for your use case. To move swiftly and effectively, they’ll need to know how you make money, what your big costs are, what the levers are and how these can be matched to the opportunities.
This needs to be well scoped out with a clear business value. It doesn’t need to be super simple, but it should be something where in three to six months’ time you can get a clear win that will go on to help you build momentum throughout the business. Having someone leading your project who’s a good communicator and can get support at a high level will help to set up the data team in partnership with the business, rather than operating as a support function.
Once you identify an opportunity you should prioritise looking at the quality of your data and determine whether this needs to be improved. Only then should you start to look at how you’ll pull in this data, clean it and collate it. Plus, you’ll want to think about how you’re going to measure and quantify your data analytics - what decisions have you made and why are they better?
If you’re keen to use AI in your project, be really intentional about why you’re using it and what the business case is. Ask yourself if it’s the best solution and, if so, how much you want to spend on it. While AI can allow you to get something up and running quickly, it can also mean that explainability and control may be quite low so think about what trade-off you’re happy to make based on the business and how it operates.
Historically, some analytics teams have ended up being help desks, dishing out profit numbers and updating dashboards while spending 60 - 70% of their time not working on the most high value tasks. Automation can be used to remove the most low value part of what analysts do. Non-technical users are now increasingly able to ask questions of data in plain text to get numbers and charts, freeing up analysts to focus on higher value tasks.
And while there can be challenges when it comes to using this technology to undertake more complex projects, utilising it to automate day-to-day tasks can significantly reduce the analysts’ workload. This frees up more time for critical thinking as they interrogate data to provide insights into some of the business’s more complex challenges.
Data definitions can vary between business divisions – whether that’s departments or countries - because they may think about e.g. profitability differently. At the start of your project, you should facilitate a discussion between the general manager of each relevant department/country etc. to align them.
But even if you align on the definitions, data quality can be a challenge: ideally you’d have automated quality checks in the data engineering pipeline, but data analysts are often the first people to look at data in depth and can spot things that other people have missed. If you can give someone with a bit of curiosity some time, they can get you more nuanced insights and help you to understand your data better.
To find out more about some of the help that is out there to get you started, take a listen to my full chat with Lydia on the Mesh-AI Data & AI Podcast.