page titel

10 Dos & Don’ts of Creating a Data-Driven Culture7 min read

Data is the new oil. We have been hearing this over and over in the past few years. While the analogy may not be perfect (I would personally take gold rush over oil boom!), it does convey the point that a new sort of Wild West is happening. A Wild West in which people and organizations can become very rich very quickly. As humans, we have always produced (big) data. For most of our history, we simply did not have the means to keep it alive.

Today technology allows unprecedented storage capabilities. For instance, a significant fraction of what we communicate to each other every day is captured in email, social media or messaging apps data. In this context, our ability to make sense of the flood of information coming at us defines our success. This is true not only in business but in every walk of life – from politics to economy to education or medicine.

The data deluge requires a data-driven culture: one in which decisions are taken based on quantitative insights about people and situations. This post gives a few rules of thumb about driving this culture shift in your organization – and more precisely how to build it up from zero. As such, it is intended primarily for junior data scientists and managers starting on this journey. This article reads as a series of five steps, from starting to scale up and eventually tackling the culture challenge.

“Our ability to make sense of the flood of information coming at us defines our success. This is true not only in business but in every walk of life – from politics to economy, education or medicine.”


What does success look like?

A bit counter-intuitive, but sometimes it is useful to start with the end-goal in mind.


DO: Have a clear picture of what a successful outcome looks like

You can keep the timeline open or limit it to, say, 3 years from now. There are two main benefits to this. First, you expand your knowledge in your field, getting a better understanding of your organization, its competitors and the industry overall. Second, reading about successful people and enterprises is often motivating and inspiring, providing a needed shot of excitement on days when nothing works.


DON’T: Stop trying to refine your notion of success

The reality is that success is a moving target which needs regular tweaking as you move along. New technologies, occupations and trends are sure to emerge. Staying informed helps you recognize potential opportunities to evolve.


How should you start?

Data professionals and teams exist in many flavors – so many that it is easy to get lost trying to categorize all of them. In a large company, data teams may be associated with specific departments (marketing, finance, IT, sales), specific software (IBM, Oracle, SAP) or specific activities (reporting, marketing analytics, predictive analysis). At the beginning, it is a good idea to focus less on the classification or internal environment and more on the problem you will be looking to solve.


DO: Start small and look for quick-wins beyond reporting

Small here means using your laptop, quick-wins means 3 months at most, and beyond reporting means not by summing or counting. For example, if you are charting monthly revenue over the last 10 months, it is reporting. If you are predicting these revenues for the next 6 months, it is analytics. The point here is to do something new, different from the classical reporting exercise of summing rows in an Excel file, which is standard in most companies by now.
How to identify which questions to answer and which problems to work on?
From my own experience, anything involving business predictions (especially financial ones) is very likely to become visible. So are segmentation projects that promise to offer insights into customer behavior, especially in relation to purchasing patterns or product usage. Depending on your industry, many other interesting use cases are possible.


DON’T: Worry about financial investment or lack thereof

During early stages, you will need to give a lot more than you take. You may have your “day job” and work on the side on your analytics project. This is fine and you will have the opportunity to scale up and ask for funding once you have obtained your first meaningful results.


How to maintain your focus?

Once you have decided on a list of interesting questions to answer or use cases to pursue, just find the data. Some use cases may require that you obtain external data (customer survey, economic indicators, etc) in addition to the internal data existing in your company.


DO: Follow the process to get access to data (and be nice to your IT department)

Depending on the company and the type of data you need, the process may require considerable effort and time to complete. Patience and resilience are key attributes in this phase, as is having your manager’s backing. Trivial as it may seem, access to data is a common cause of failure for analytics projects.


DON’T: Be distracted while building a trail of quick-wins success

Distraction may come in the form of easy data questions coming your way, which tend to multiply over time. Once it is known you have access to data, it may become a full-time job to answer reporting-type questions such as:

• How many customers with this profile do I have in country X?
• How much revenue is contributed by client Y?
• Has it been decreasing recently?
• etc.

Each of these questions only takes minutes to answer but, if left unchecked, may easily spiral into hours or even days, leaving no time for the actual analytics work.


How to scale up technology and people?

With sufficient quick-win material, you will be able to take on a larger project (6 months to a year) which carries more risk and more impact. This may imply hiring one or few people and asking for a small amount of money to invest in technology – enough to move away from using your laptop.


DO: Search for curiosity when hiring data scientists

The classical data scientist profile combines science skills with programming and communication. A Ph.D. in a hard discipline usually embodies these traits. However, not every Ph.D. hired will turn into a successful data scientist (and conversely every successful data scientist does not have a Ph.D.). The extra ingredient to look for is curiosity. Is the candidate genuinely interested in understanding all the aspects of your business inside-out? Are they curious about the industry as a whole? If they are, you will have new hires with potential to make a transversal impact and to inspire a data-driven mindset across the organization on a larger scale.


DON’T: Invest heavily in tools

Ideally, people will come first and technology second. In the 1980s and 90s, a company’s sales pitch might have included something like: “We have the most expensive tools on the market”. Declining technology costs have reduced the barriers to entry, and the sales pitch has evolved into “We have the best experts on the market”.


How to tackle culture change?

DO: Copy the masters

Picasso once said: “Good artists copy, great artists steal”. Many companies have figured out how to create outstanding corporate cultures inspired by the likes of Apple, Google or Netflix. By attending conferences and joining various professional groups you will meet people who have tackled the same culture and organizational problems as the ones that you are facing today. Steal from them and always give credit when credit is due.


DON’T: Underestimate the power of culture

Beyond the value created directly for a business, a great culture can be a means in itself (think of Zappos). It is also one of the top reasons a superstar will join your organization. Strong data cultures are easy to spot in the market. Some identifying marks include: abundant personal development opportunities, external communication of technical results via papers or patents, domain-specific conference attendance, partnerships on collaboration projects – often with start-ups or universities, etc.


The final thought concerns the time required to create this culture shift. Depending on the industry, company size and available budget, the change may take anywhere from 2 to 5 years. However long it takes, remember this is a definite “must” in the Wild West of opportunities created by data.


Many thanks to Benjamin Perrin for helping me write this article! 🙂

Hire a data scientist freelancer here.


Catalin Ciobanu

Catalin is co-founder and CTO of, a French startup focusing on the science and importance of skills in the job market. Besides entrepreneurship, Catalin’s experience includes business travel and experimental particle physics. He is passionate about education and sports.


How fashion retailers use AI to accelerate benchmarking5 min read

Mar 3, 2018

Over the past decade, the fashion industry has been shaken by the adventn

Why you should consider building your own CRM solution7 min read

Fév 2, 2018

All companies are eager to send emails and notifications with the right messagen

Leave A Comment