How do you build a data-first company?



By Sean Pyott, MD of thryve

All good companies use data to make decisions. But it’s becoming apparent that a data-first organisation is a different animal. This realisation goes a long way to explain why many analytics ambitions fall flat. There are crucial differences between how we used to handle data to inform our businesses and how we’re meant to do it going forward.

The world is turning towards data analytics. Maybe it’s the rise of a new generation of people who instinctively appreciate the value of data analytics. We can also credit analytics technologies, such as Tableau CRM and Microsoft PowerBI, making it much easier and more affordable to use analytics on different granular levels. And then there is Big Data: practically every business has been collecting data in digital storage – yet they don’t do nearly enough with that. It’s not tough to guess why analytics is becoming so popular.

Yet for all the benefits, analytics projects can easily fail. Having the data – and technology to sift through that data – is perhaps the easiest part of the change. There are more substantial barriers to overcome at the operational level, something Google’s global head of customer analytics, Neil Hoyne, writes about in an article, Lessons from top companies on building a better data-first strategy.

Focus is essential to stay on the rails

Hoyne notes that people tend to look at their data in isolation and not as a larger picture. Rob Roy, the former chief digital officer at Sprint (which merged with T-Mobile this year), noted something similar in a 2018 interview with McKinsey. Warning that companies often get excited about data, they end up being overwhelmed. Ironically, it’s because they bite off too much but then create pockets of analytics information.

Sprint eventually sidestepped this problem by focusing on a more narrow question – in this case, understanding customer behaviour – and tailoring additional demographic elements as the data revealed more questions. This approach inevitably made the data more useful to other parts of the company as well:

“We then overlay those insights with data from digital properties: website, mobile app, stores, and call centers. And we started to understand better our customers’ journeys across the web, as they called us, tweeted about us, etc. We’re now starting to teach our “bots” to learn more about contextually relevant interactions with the customer.”

Think beyond metrics and answers to problems

Focus checked Sprint from being overwhelmed, and yet it didn’t lock those insights into a particular silo. The data stops becoming a means to an end and more resembles a potential roadmap of discovery. To encourage that line of thinking, Hoyne asks three questions: Do I know what this metric truly means?; What could influence this metric, and how?; and, Am I limiting what I can learn from my metrics?.

“I’ve seen that successful companies don’t just look at their metrics as numbers,” Hoyne writes. “They look at their metrics as opportunities to ask more questions: Where is the market headed? What should we be aware of? That way, a single metric becomes part of a larger story, not the whole picture.”

Lesson one is thus: don’t over-consume analytics, at least at the start, and don’t use that information to purely answer specific questions, such as measuring underperformance.

Allow employees to think and fail

The second lesson I draw from Hoyne and Roy, as well as this O’Reilly article, is that you need to figure out how to empower employees with data.

All three sources give different yet complementary advice. Hoyne argues that you must allow your teams to appreciate failure and see it as the first step in growth. This point relates to the earlier one that the questions which data creates are as important as the answers it can provide, so an open mind is essential.

If you’re only hunting for efficiencies, you’ll overlook other considerations the data is telling you about – and sometimes risking failure is the only way to test new assumptions. If every data-related question is about fixing something, then failure doesn’t seem like an option. While data will certainly reveal bottlenecks and logjams, modern analytics is as much about ‘what if’ as ‘what’s wrong.’

A culture that decides using data

Another consideration, this time from O’Reilly, is to make sure your company’s decision-making culture is data-driven. Most likely, it is not – there is still a tendency to defer decisions to the most senior person, not the person with the best insights. This is classic top-down decision-making or HIPPA (the highest-paid person’s opinion). Changing the direction requires your employees to use data as part of their decision processes:

“To succeed at becoming a data-driven organisation, your employees should always use data to start, continue, or conclude every single business decision, no matter how major or minor. This kind of inquisitive culture should drive everyone on the data team—including IT, data engineers, data scientists, and data analysts—to continually enhance and refine the tools that business users need to inform their decisions.”

The right technology environment

This point relates to technology deployment as well. Roy says it best: “Having a data-first mentality is a crucial first step, but then you need to put in place the processes and capabilities to be able to use the data. We had to first collapse data into one or two locales so we could easily extract it.”

Technology is the easiest part of creating a data-first company. You likely have the data, and you can conveniently access a platform such as Tableau or PowerBI. But some considerations will impact how well the technology works: is your data pooled and accessible to different groups?; is it easy for people to capture relevant data, especially if business systems don’t prioritise data capturing?; and, do you have processes that make access easy?

In summary, how do you create a data-first company?

  • Start small, focus, and then build on those successes
  • Look beyond metrics for possibilities, not just answers
  • Encourage questions, experiments and failure
  • Push employees to always include data in their decision-making
  • Don’t base decisions on seniority, but on insight – look for the trickle-up effect
  • Create processes and a technology environment for easily accessing and capturing data