The intoxicating love for data

We all love data — it is present in pretty much all things digital nowadays. It allows us to understand our customers better, sell more effectively, develop our services faster — and just about a million things more. It’s then easy to see how falling in love with this mysterious term of large amounts of information is easy. However, when companies race to build data-driven services nowadays, several fundamental things seem to be forgotten in the process — a typical falling in love process. Consider the next aspects, at a minimum, to improve your chances for a successful data project.

Ignorance of limited range

Not all data, or information, is usable. Just having large quantities of seemingly “valuable” data is not going to take you anywhere. Consider data as a source to an argumentative essay — if you aimed to write a convincing article, would you just turn to a single extensive writing where you would source all background to your arguments? For sure that would lead to an unconvincing piece of content.

Data is no different. Making decisions based on limited or bad quality data is going to lead you to make limited and bad quality decisions, ignoring many elements that intuitively would have been considered. Datasets rarely cover all the information from the process at hand.

Let’s consider an example by Pinterest, an online platform for discovering mainly visual content on the web, that happened some time ago. To improve their email communication effectiveness, they started generating personalized emails for users based on their preferences on the platform. Personalized communication, sounds superb — right? Well, people following wedding-themed content received emails declaring “Congratulations on getting married soon! Here are some cool designs for wedding invitations for you.” Needless to say, not all people following wedding-themed content had plans to get married soon. This slightly awkward mishap shows that reliance on a limited data source can produce highly distorted results.

Lack of data enrichment

Closely related to the previous point, single data sets often offer relatively limited information. This is not to drive you away from using these data sets — but rather enrich the existing data. Data enrichment is by definition enhancing, refining or improving data. This enrichment can be connecting other (relevant) data sets to your existing one and finding correlations between them, manually adding information (dangerous, due to subjective views and errors), and so forth. Data enrichment allows for a more multi-dimensional view of given data sets — improving understanding and reliability.

Forgetting the human element

Likely the most central thing in building data-driven digital services. While data analytics offer better guidelines for informed decision-making, humans are needed to ensure the correct course of action is taken — at least before the introduction of Broad (or at least broader) AI. While basing decisions on conclusions drawn solely from nice, logical data sets may feel very tempting, the “soft” elements with human interaction should not be forgotten. Talking about data, there are two central elements crucial to success, from the human viewpoint.

Firstly, data visualization. As with many emerging technologies, data-driven applications can easily be too engineer-y — to put it simply. In most cases, your users are not data analysts. The core aim of data visualization is to make information accessible and easy to understand — a point that can quickly fade away when data masses grow and new dimensions are found. Developing data-driven services is no different to developing any other services — always build together with your users. Data visualization is a topic we could ramble on about endlessly (which we will, in another post dedicated solely to that). But for now, to achieve the most from your data, you want to keep information accessible, usable and limited enough to your users.

Secondly, human decision-making. While data is vital for business automation, decision-making will remain largely a human activity — by its very nature, as analytical automation tools can only make decisions based on the data they read (again, point 1: datasets rarely cover all aspects). Data and analytics are a superb tool for humans to make informed decisions, by adjoining them with emotional intelligence.

Coming back to the wedding-themed mishap of Pinterest, human oversight and data enrichment could have just prevented this communicational error from happening. Emotional intelligence and human decision-making combined with their vast user data could have been able to produce more accurate and appropriate messages. The automation, reading only data it was given, did not know these people were not necessarily going to get married — but humans would have.

To conclude, data, analytics, machine learning, and AI are all tools inevitable for modern businesses to thrive — but like any other tools, they have their limitations. By acknowledging these limitations, these tools can and will disrupt the automation capabilities of your business.

Kalle Kyyrö


+358 50 304 7409