Today, we hear a lot about corporate data-driven cultures, but what do we actually mean by this? In my opinion, corporate culture can be called data-driven when all stakeholders agree that data can provide an answer or solution to the most critical business dilemmas and/or issues. Such an agreement implies there’s a shared vision with regard to business data usage across the entire organization and all stakeholders accept the following manifesto:
- We are ready to invest time and money in data extraction, storage, analysis, interpretation, visualization, etc.;
- We’re ready to listen to our data. That is, when we need to make a business decision, we stop and tell ourselves – let’s look at the numbers;
- We know how to understand the data. Indeed, there’s nothing more daunting than making wrong decisions with all the necessary data at hand.
- We trust the data and use it to make decisions.
When managers look at a well-crafted analytical report and say that they’ll make their decision based on previous experience and not the provided report, they aren’t necessarily wrong. What if the analysts didn’t take into account the seasonality or results of the upcoming elections? It’s crucial to have a well-established and seamless communication and trust between analysts and managers.
Admittedly, it’s much easier to build a data-driven culture within an organization when founders clearly understand the value of data analytics and data science. Using data for informed decision-making is pretty expensive and time-consuming. If there’s a resistance to change in particular departments or if not all senior managers are on the same page regarding the value that will be unleashed with the help of big data, the risk is high that any pilot data science project will be suspended or canceled just because top management will anticipate a very quick turnover. There should be a strong commitment at the organizational level to develop a data-driven culture and the business leader him/herself should be the catalyst and driver of this transformation.
First steps to getting business value from your data
The first thing you’ll discover on your way to data-driven decision-making is that you do not have enough data. In general, you will always be missing some data for objective reasons, but you have to start somewhere.
First, you need to build an infrastructure to collect and store metrics. In the overwhelming majority of back-end data science projects (e.g. that collect information about customers, their credits and payment transactions, etc.), data infrastructure starts with the replica of the production database with a relatively simple structure, as well as the questions you’ll want to ask about the data. So it’s a starting point and investing in something more complex doesn’t make sense.
For front-end data points (e.g., web page views, user interaction with controls, scrolling, clicks, input, etc.) you can use classical tools like Google Analytics or HotJar to record sessions. The basic functionality is sufficient for marketing tasks and sales-related tasks.
Having built the basic infrastructure and started collecting basic statistics, you need to make sure that your digital product will evolve in line with its metrics.
That is, each time you want to implement a new feature in the product, you need to answer these questions:
- Which key business metrics will be affected by feature?
- What changes in customer journey or backend algorithms will be introduced? And how will this affect the existing metrics?
- What stages can I decompose the new functionality into, so that I can look inside and analyze the impact by collecting the right data for each stage?
Now think about whether it is part of your problem statement to be able to collect all of the above metrics. And how exactly will you collect them after the functionality has been implemented?
Next, you need to make sure that your product development team understands the importance of data collection and treats your data storage system (or subsystem) the same way it treats production. This will ensure that your data science function is embedded in product design from the very beginning to ensure the scalability and efficiency of your data analytics efforts. To reinforce collaboration, you’ll have to create shared collateral like common libraries and APIs, QA guidelines, etc.
Analytics for data analysts
The availability of data does not necessarily mean that it can be used effectively. The following issues usually arise:
- Where can I find a metric?
- Is data collected appropriately for this metric?
- What kind of data visualization will make the most sense when it comes to making conclusions?
- Is there statistical significance?
- Is it possible to derive more data to better understand what is going on or to check data collected by other metrics?
It turns out to be a fairly extensive job that requires special skills and abilities, and, most importantly, time. So it makes sense to build a full-fledged data science or data analytics team in-house or offshore to make sure you squeeze maximum value from your endeavor.
Hiring the right people for your data science team
As the company grows, it turns out that not all employees understand the importance of data and know how to work with it. Two questions arise: how to promote a data-driven culture internally and how to hire the right people right away.
As for internal promotion, as mentioned above, if the company’s senior leadership is a proponent of a data culture, then it has to be promoted top-down to top management, middle/linear management, and so on. For example, as a leader, you can demand from your product managers to calculate how a newly added feature will affect the product’s profitability or how a slight change in a key metric can influence user loyalty.
At 8allocate, we approach the planting of data-driven culture on our client-tailored dedicated development teams from both sides. Each PM requires that a team lead clearly describes the metrics by which they will track the results of implemented changes, and PM must ensure that these metrics are implemented in the right way.
When we interview software developer candidates who want to join our client teams, we screen them for data analytics skills or at least for the understanding of big data potential.
The following questions help verify such skills and experience:
- Did you calculate the commercial effect of that feature’s implementation? If yes, how? OR
- Why do you think a positive outcome was attributed to that particular feature?
The right candidate will always be able to justify this or that action in terms of data and value.
With the growth of business and data volumes, it makes sense to use more advanced statistical methods and more advanced data libraries – something that is now commonly referred to as data science.
If we talk about data science in a broader sense (than neural networks and machine learning), transitioning from classical statistical analysis system (SAS) packages for building a logistic regression to a DIY solution on Python can be regarded as successful experience with data science. For instance, on one of our client projects, it helped reduce the time to develop a credit scoring solution by 5 times!
Check out 8allocate’s Python development case stories:
To sum up, a data-driven culture is based on the following principles that you should follow to embed it into your business DNA as early as possible:
- The expected return on investment (e.g. in terms of time-saving, improved accuracy/speed of decision-making, etc.) is adequate to the resources involved;
- When creating and developing a data infrastructure, all stakeholders’ feedback should be analyzed carefully and taken into account when making the final decisions;
- Data Infrastructure development should keep pace with the development of your internal processes and methodologies. It should neither lag behind nor outpace your brand/product evolution in terms of its business insights needs.