What the middle-market business manager really needs to know
Data science is a big, exciting buzzword in business – and no wonder. With the increase in access to information and the advances in technology, data science is increasingly seen by some as a silver bullet to all management woes.
While data and insights can never replace excellent management, it’s true that they’re essential tools to the modern business. When done well, managers can use them to guide their instincts, make braver calls and convince naysayers.
For many leaders, this utopic view of data in business seems totally unattainable. With a seemingly endless stream of information, and a belief that they’ll be unable to turn it into useful business insights, they’re more likely to rely on their old faithfuls – trusting gut feel, copying competitors and following the advice of their buddies.
As we’ve seen over and over in the business world, making proper use of analytics is a key driver of ongoing success – and worth getting your head around.
1. Make data analysis your problem
According to Florian Zettelmeyer, a data analytics professor at the Kellogg School managers should see analytics as part of their job.
“The most important skills in analytics are not technical skills,” he says. “They’re thinking skills.”
To do that, says Zettelmeyer, doesn’t take mathematics genius. Instead, managers should build a working knowledge – to spot good data, challenge the source, turn the numbers into useful insight – and avoid making decisions based on bad ones.
As Zettelmeyer says, “if you don’t understand experiments, you don’t understand analytics.”
"The most important skills in analytics are not technical skills. They’re thinking skills."
2. Collect data to solve a business problem
Data is everywhere – we could measure everything. And so often, that’s what managers do. The result is a soup of data with no context and no use – and worse, it often muddies and obscures the data that actually could have an immediate business use.
While there’s basic data you should be collecting as standard, like site-traffic visits or sales by region, anything more in-depth must be connected to a real business problem. “What has the greatest effect on customer satisfaction?” “What are our optimal store staffing levels?” Or “How many marketing emails will net us the most sales?”
It’s a scientific exercise – you’re designing an experiment from scratch, to factor out variables and lead you to reliable conclusions. Only by purposefully choosing what data to measure, in order to answer a business problem, can you be sure your results will be unbiased and useful.
3. Build a working knowledge of statistical analysis
With so much of data analysis still being left in the hands of specialists, managers tend to defer to these ‘experts’, who present recommendations backed by seemingly complicated data analytics. While using your team’s expertise is important, managers must temper this with a working knowledge of analytics, in order to judge the quality of the insights for themselves. Without this, you’re essentially passing managerial responsibility down the chain.
A working knowledge begins with understanding how the data has been generated – how was it collected and what are the possible gaps or failings that could be present? Without understanding, you could be in danger of drawing incorrect assumptions, or relying on inaccurate data.
The quality of the analysis is important too – and requires at least some understanding of basic statistical concepts:
- Correlation: is likely the most well-known statistical tool. Comparing two variables is the simplest method, although identifying causation, is the difficult side of this tool. For example, “People with more education tend to have higher earnings, but does this imply that education results in higher earnings?”.
- Regression analysis: is used to strengthen the rigour behind correlation, preferably bringing a number of variables into the mix, to gain deeper insight. Regression analysis can also help to develop a better understanding of how closely related the chosen variables are.
- Time series analysis: Even if you do not use the full gambit of regression analysis, time series analysis can give you some solid insights into your company. Consider pulling together purchasing data across time periods, geographies, demographics… do any trends emerge?
- Pattern recognition: as more data becomes available, with more advanced data analysis tools, pattern recognition becomes a viable analysis tool. Extracting actionable insights from data will continue to develop, as Artificial Intelligence continues to advance. You don’t need AI to recognise patterns though, rather a strong understanding of your business and the key indicators that align with its performance.
4. Challenge your data
While data analysis can more accurately inform decision-making than gut instinct, it isn’t an objective truth. Somewhere along the line, a human had to be involved – deciding what problem to solve and what data points to collect – and inherent in that is a very real, and almost entirely unavoidable bias.
Microsoft researcher Kate Crawford points to the over 20 million twitter messages posted in response to Hurricane Sandy in the US as an example. While that sheer number of tweets suggests that the collected information would give you a reasonable gauge on the effects of the storm, this simply isn’t true. Twitter users tend to be younger, more affluent and to live in urban areas. “These were very privileged, urban stories,” Crawford said. This means the data source had inherent bias and, like most data, needed to be carefully challenged and considered within the context.
A quick way to check the validity of a data-driven insight is to pressure test it with a series of questions. These are:
- What is the source of the data?
- How representative is the data – is it a broad sample, narrow, or one with bias?
- Does data distribution include outliers? If so, have these affected results?
- What assumptions were used to reach the insights? Could any of them mean the insights are flawed?
- Why was the analytical approach taken? What alternatives were considered?
- What independent variables could be having an effect? How have these been factored for?
"It’s a scientific exercise. Only by purposefully choosing what data to measure, in order to answer a business problem, can you be sure your results will be unbiased and useful. "
5. Use your business expertise
While AI has come on by leaps and bounds in the last decade, nothing can replace a manager’s on-the-ground knowledge of the business to draw insights and gauge results. That real understanding of how a business works can add much needed perspective to data – especially when the results are unexpected.
For example, is the spike in sales a result of your recent promotion, or because your direct competitor has just suffered a PR scandal? Can you attribute greater staff retention stats to your management training, or is it because the slumping economy has made people more worried for their jobs?
There’s no substitute for this kind of pattern recognition, and that overarching insight only exists at management level.
6. Question your KPIs?
According to HBR, the modern mid-market business manager can expect data to have the most impact on three key areas – accurate pricing, predicting demand, and preventative maintenance for equipment.
But there is some doubt whether stats and analysis can sway behaviour at all.
Key performance indicators (KPIs), the most common people-management metric, are generally poorly applied. Many aren’t measuring the right thing at all – and those that are, say detractors, simply allow employees to create work-arounds. They hit their KPIs, and that’s about it. To make sure your KPIs are doing their job, you need to think carefully about what you’re measuring, and whether that measure truly represent ‘success’. Similarly, taking your unique business environment into consideration is critical – if your very dedicated, hardworking employees continue to miss KPIs because of challenges over which they have no control, they’ll quickly lose heart and dedication.
Don’t leave your business decisions to ‘experts’
If you’re a mid-market business manager, you might be daunted by the complexity of data analytics. But you know your business better than anyone, so it’s a mistake to abdicate your responsibility in favour of analysts, no matter how expert you think they might be.
Look for ways data analysis can solve your business problems, get to know what data you need and how it’s collected (and from whom), and choose only what will give you real answers. Learn about the different types and levels of analysis, and how to test that your data is valid and usable. It’s not rocket science – it’s data science. You don’t need to be a genius to figure it out, but you do need confidence in your own business expertise.
Above all, keep a close eye on what you’re measuring, so that your data analysis is accurate and relevant to your business, and not just a waste of everyone’s time.