The world’s leading publication for data science, AI, and ML professionals.

4 Paradoxes on the Way to Data-Driven Culture

What skills are required from an analyst when building a data-driven system?

source: Natalia Kiseleva via Analyst's comics (CC BY-NC-ND)
source: Natalia Kiseleva via Analyst’s comics (CC BY-NC-ND)

Managing a business by KPIs, making decisions based on operational data and forecasts, finding business insights with the help of artificial intelligence is a dream every manager and business owner has. However, building such a system is not easy – there’s an entire zoo of IT services, and each requires support. In the end, everything comes down to people, those data analysts who know how to ‘communicate with machines’ and translate the results of their work into the language of business.

The process of turning big data into business solutions, an information product, consists of several stages and requires the teamwork of different people. There is even such a term – a Data-Driven Organization. It is a company where management makes decisions based on analytics, not just experience, opinion, or intuition.

In order to become such a company, it is not enough to simply hire a team of top-notch programmers or buy an advanced CRM-system, you need to change the culture of corporate communications, understand the psychology of the participants of these processes.

An analyst is not necessarily a position, it is a role, above all: many employees participate in the preparation of management reports. But oftentimes customers and managers receive numerous tables and slides from their analysts, rather than information appropriate for decision-making.

In my 10 years of integrating corporate reporting systems I have observed many cases of expectations that the reality didn’t live up to, and described them in this article. This will help you understand where your company has flaws in communication and how to remove barriers on your way to Data-Driven culture.

3 Types of Analysts

Staff working with data and reports can be divided into 3 groups: analysts, picture people, and technicians. They have different roles, tasks, and requirements for data, their processing, and results. They all do joint work using different approaches. This is where the paradoxes of their attitude to work arise.

  1. Analysts are finance specialists, economists, marketing specialists that look for answers to specific questions. For this purpose, they conduct research, collect information, and draw conclusions. Their task is to analyze data, identify cause-and-effect relationships, and study the impact of various factors on business processes. For example, an economist identifies the factors that influenced the growth of costs, and a web analyst builds the semantic core of the site out of search queries keywords.
  2. Technicians are database developers, mathematicians, and Data Science specialists. They derive information processing algorithms, design data warehouses, and automate reports. It is important for them how the system works, what affects it, and how to make the process error-free and uninterrupted, that is, to create a perfect algorithm.
  3. Picture people make the final product, know how to turn tables into clear diagrams and slides or even dashboards. Usually, they don’t hold the position of a designer, but a manager who understands what the business needs and can set tasks for designers and developers. An analyst can also play the role of a visualizer, think in categories of usability and have a sense of beauty.

Each of them considers their part of the work to be the most important. Analysts think it is the process of finding answers to questions, picture people consider it to be the beauty and accessibility of presentation, technicians believe it is models and algorithms. And the result is paradoxes that are not obvious to the end customer.

Paradox 1. Analysts do not need visualization

Perfect tools for analysts are Excel spreadsheets, OLAP cubes, or data showcases that can be connected to via Qlik or Power BI. For them, the result is in the intersection between several tables or arrays. They feel comfortable and convenient with them, they understand them. Not for nothing have they sliced and diced everything and organized by cells.

Simple diagrams or presentation slides are not interesting for analysts, to them, they are just ‘beautiful pictures’ that are useless without immersion in the context. After all, you can always delve into tables and get to the bottom of it. They understand the value of dashboards, but make them too complex, with many bookmarks where filters take up half of the screen.

Multi-filter dashboard that only analysts can understand
Multi-filter dashboard that only analysts can understand

Here’s a common problem –

an analyst doesn’t know how to delegate access to information.

They have set up different reports for themselves, and they will collect data and prepare the necessary sample upon your request. But this way they themselves become the constructor of summary tables, closing information flows on themselves. Such an employee will spend half a day creating the same reports, only from different angles, complaining that they don’t have enough time for analytics.

Paradox 2. Picture people do not care about data quality

When I say ‘picture people’, I mean analysts with developed visual thinking, not designers. They shape the Analysis results, and in pursuit of harmony and aesthetics the reliability of information pales into insignificance for them. Then a picture person can miss important details that an analyst-researcher would definitely pay attention to. For example, in this factor analysis diagram colors have been selected incorrectly.

Factors of the actual payroll deviation from the planned one
Factors of the actual payroll deviation from the planned one

According to the idea of the diagram, those factors that go with a minus are shown in red, and the increase is colored green. But here we analyze the reasons for payroll overspending, and reduction, economizing are good, whereas the increase in expenses should, on the contrary, be colored red.

But the picture person did not pay attention to the context, because the diagram itself looks clear and logical.

Paradox 3. Programmers do not care about business results

Data-Scientist, Machine-Learner, it all sounds cool, but in fact it is the profession of a mathematician, a programmer that works with algorithms. For them, the technical side is important, the nature of the relationship described by the model of machine learning, but what to do with this information further does not bother them. They will build you a model of an ideal employee’s professional competencies, but will not tell you how to find one and how to manage them.

I am not criticizing developers for their concern about the database structure integrity instead of how to increase customer loyalty. Rather, I am talking about the fact that it is naive to expect ready solutions for business development from them.

Visualization is also of little interest to a programmer, if it is not a database communications scheme. They are not responsible for data quality either.

Their area of responsibility is stable work, system performance and absence of errors.

It is also very important, though.

Paradox 4. Business needs text

In the beginning of the article I talked about three roles, but there is another participant in the process – the business customer, the executive who makes decisions. The most surprising thing is that they don’t care about visualization, depth of analytical research, or the algorithm workability.

Business needs conclusions, clearly defined options for solutions or strategies, as well as forecasts about what they can lead to. Ideally, they need a solid, clear text containing simple sentences without any participle clauses: what prices will ensure maximum sales volume, why investments in this particular project will pay off.

Of course, if an executive is provided with a straight text of an analytical note, they will not be happy.

They need text, visualization, and conclusions, all on one screen.

I began to see more and more dashboards which contain a block with written conclusions, and even BI-systems add-ins that automatically generate text from Dashboard.

Management dashboard with a block of written conclusions
Management dashboard with a block of written conclusions

At first, I thought such requirements were whimsical, but now I admit that it is an additional stage of validity control, where an analyst makes sense of the conclusions. It is not about manual tedious copying of data from a table to a chart, but about understanding the reasons for deviations from plans, about preparation for a meeting.


What’s next?

So, analysts look for answers to questions and structure data, picture people create a harmonious picture, technicians derive algorithms, and monitor the continuity of their work. And business waits for simple and clear conclusions that will make it possible to make specific decisions.

You should not expect one single expert, even if they share your opinion and your perception, to set up your system of end-to-end analysis of all business processes. Here you need teamwork where people complement each other and understand the customer’s expectations correctly.

If you order the project of a house or apartment from a designer, it is naive to expect that they will design a layout for communications as well. In the same way, going to a hairdresser you do not hope to receive a manicure or back massage, too. Although you can get all these services in the complex, in the former case in a construction company, and in the latter in a beauty salon.

You can argue that hiring a whole team of analysts of all kinds is too expensive, and ordering a project from a systems integrator is even more expensive. Of course, you can somehow cope on your own if you have a simple task to optimize the current process or perform a cosmetic report refresh. It’s just that it will not be that digital transformation, a transition to Data-Driven culture.

There is an option where you can get the result without increasing human resources. In order to achieve a balance, you will have to involve the management team, immersing them in the operational activities, process metrics, taking on roles of technicians, or at least analysts and researchers or picture people. This is the ‘rightest’ option when the director becomes chief analyst and leader of the culture of data-driven decision making.


Related Articles