Most executives already agree that the insights from data science can lead to better business decisions. Many businesses have the data ready to be analysed along with the right data systems and tools to solve their problems. The sticking point is in finding data scientists who are going to implement this project and how this project will directly benefit the business.

In many cases, we've seen that the hesitation on investing in data science ultimately stems from the belief that its actual cost and benefits seem unmeasurable.

While we're not in possession of a magic formula that spits out the right answer to this question (there isn't one), we firmly believe there are standards and best practices to set your data science initiatives up for success.

Pitfall #1: Investing in infrastructure without project management

It all starts with project management. Most businesses begin with the most obvious: constructing data infrastructure and hiring data scientists.

Many companies are misinformed that data infrastructure is the key investment that they need to make, believing that their employees can solve data science problems once they are equipped with those tools. Only after they find that there are more data and graphs, but not more actionable insights nor innovative data-driven products, do they realise that it's about attaining a balance of the right talent and concepts for gaining insights from data, rather than a matter of tools.

What should you do instead?

  • Don't rush into data infrastructure
  • Plan your data science initiative: identify the needs; define objectives, requirements and timelines; allocate resources.

Pitfall #2: Pushing forward without suitable talent

Some businesses will look to build a data science team right away and spend potentially months looking for suitable candidates. These companies may already have a data analytics team who is doing constructive, but separate, data work. While searching for the new candidates, this analytics team is often put to work on data science in the meantime. But they may not be capable of creating and using advanced data science algorithms, and could be put under a lot of stress due to working well beyond their areas of expertise when tasked with conducting a full data science project.

What should you do instead?

  • Don't just wait for the right candidate
  • Identify the skills needed for the initiatives
  • Assemble a team according to the skill-sets required

Pitfall #3: No culture for converting data science insights into actionable outcomes

There are companies who have successfully avoided the previous two pitfalls but are still unable to gain the actionable insights of data science. Unfortunately these are not a rare cases, and it is probably the deepest reason why executives hesitate to dive into data science.

The third pitfall is the most common cause for failure in data science initiatives and often precedes the previous two - incompatible culture.

Successful data science initiatives need to end with action. It may be an uncomfortable truth, but while most young data scientists are great in programming and machine learning, they are less experienced in formulating their insights into substantial executable actions. After all, business nous is acquired through hard-earned experience rather than a quick bootcamp or school.

In fact, even very experienced data scientists might be frustrated when they are unable to push their insights through to action because of administrative processes or organisation politics. A data scientist withers in incompatible culture, along with the data science project.

So how does one create a 'data science friendly' culture? It is much more than data scientists and data infrastructure. It is about embracing the mindset of collecting data, respecting data, collaborating with a data research and development workflow, making data-driven decisions, and focusing on creating products and services based on data.

Building a companywide data culture is often the key to breaking down barriers and nurturing innovation, and at the same time, it is often the most expansive, and expensive, transformation among all the things related to data science.

What should you do instead?

  • Foster a data science friendly culture in the company
  • Introduce data-driven decisions from a top-down approach
  • Empower specialists and project managers to conduct change management

Take it progressively

There are solutions for the pitfalls, but they are not easy tasks either. Can a company start a data science initiative in a more progressive way such that the risk will be contained? Hiring an experienced data science consultant would be a good solution. By partnering with a data science consultant, there won't be the need to put down substantial investment in data infrastructure. The company can now take the time to hire candidates that suit the initiative as it grows. The organisation is also better able to absorb and integrate the data science workflow, all while having the initiative improved on and delivered.

Besides solving the three main problems, working with data science consultants has a few additional advantages:

  • they usually work in a team with diverse expertise
  • they understand business needs and business process
  • they are good at project management
  • they are able to communicate cogently with senior management on pushing forward specific actions based on insights

In the long run, companies who are building a data-driven culture should plan on employing in-house data scientists. They will be able to spend more time learning the company business in depth, building close relationships with all business units, planning and implementing the data policy and participating in the whole development of data products. Nonetheless, it's often wiser to take it step by step - learn how to run data science projects from experienced consultants before forming your own data science team.

Lastly, one quick reminder for developing data science initiatives in your business: begin by finding your pain points and end with change management!

We would love to know more about your data science experiences and projects, so please share any questions or comments below!