In my 8 years in analytics, I have interviewed and hired hundreds of people – I have a good idea of how the analytics market works.
Key knowledge about data analytics – this market is practically non-existent. In 2019, I hired 34 analysts to my team, of which 23 (68%) are interns and juniors. We would be happy to hire more mature guys, but they simply do not exist, so we have to hire people with potential and grow them.
For comparison, we hired 23% of juniors to the team of data engineers (5 people out of 22) – there is a market here. This discipline is developed in banks, telecoms, and retail, which means that there are more ready-made specialists on the market.
Skills and requirements for data analysts at different levels in Yandex. Data analyst profession
This article has two goals:
The first is to share with the market the terms in which we at Yandex think about the levels of analysts. This will reduce entropy in the market, where today an arbitrary set of expectations and skills can be hidden behind a vacancy and resume of an analyst: from project management and system analysis to automation of routine business operations. The prefixes junior, senior, leading do not carry any information at all.
The second is to give a clear plan for growth and development in the role of a data analyst. At Yandex.Taxi, we are forced to build a pipeline for the growth of employees, because otherwise we simply won’t be able to cope. The very formalization of analyst levels is a consequence of this approach. But not everyone works in large companies, not everyone has a strong mentor or mentor nearby. This article is intended to help such people look at their points of growth and work on them.
What is the Data Analyst Role?
A data analyst is a person who helps a team:
Make decisions more objectively based on facts and data (as opposed to opinion, intuition and experience);
Look for product and business growth points.
The work of an analyst involves working with data through the use of SQL, Python and other programming languages, creating dashboards, and automating processes. But these are only tools to achieve the two goals described above. If a person is engaged only in this, then he should not be classified as an analyst. Perhaps he is a data engineer, an automation developer, but these are completely different roles with different requirements.
The analyst concentrates on studying data, structuring complex systems, understanding processes in order to benefit the business. The product of analytics is the answers to asked and unasked questions, the creation of thought models and frameworks, and the recommendations derived from them that lead to the growth of business performance.
The way to achieve this in most cases is to work with data, but not necessarily. Analytics helps to make decisions, leads to actions in the product and business. In some cases, you can make a decision without analyzing the data: just formalize all possible situations and decision forks, and then discard most of the options, based on what the team already knows.
↓ To gain a deeper understanding of how products are created, developed and scaled, take training in GoPractice simulators.
→ The Data Driven Product Management Simulator will help you learn how to make decisions using data and research when creating a product.
→ "Product Growth Management Simulator" will help you find ways to manage product growth and scaling. You will build a growth model and develop a product development strategy.
→ "SQL Simulator for Product Analytics" will help you master SQL and apply it to solve product and marketing problems.
→ Don't know where to start? Take a free test to assess your product management skills. You will identify your strengths and blind spots, get a professional development plan.
→ Even more valuable materials and insights are in the GoPractice telegram channel.
Basic data analyst skills
In order to effectively cope with the described tasks, the analyst needs to:
Have an excellent mathematical background. The analyst does not want to double-check the calculations and formulas;
Understand basic probability theory and mathematical statistics. You need to be able to test hypotheses, understand errors of various kinds, dependence and independence of tests, and so on;
Have a mathematical culture. If the analyst uses a method or algorithm, he must know the scope of its applicability;
Possess critical thinking. Less likely than other people to fall into the traps of cognitive distortions;
Have a product mindset. Be able to digitize user experience in metrics, as well as see behind the metrics users trying to solve a specific problem;
Have a business mindset. Be able to digitize the company's business processes and market changes, link it together with the product and users;
Be a techie. An analyst is not required to program like a developer (efficient, fault-tolerant and scalable), but he should not have technical blockers in order to solve business problems: study the documentation, go to some new database and pull out the necessary data, write a parser, use some -some API for automation and so on.
An important property of a good analyst is objectivity in relation to oneself. The analyst must control his own desire to appear better than he is. Even the strongest analysts make mistakes. This is normal, development is impossible without mistakes. It is very important to be able to track your mistakes and quickly communicate them to the team, especially if this can change a decision made earlier. Hiding errors is a flag of the analyst's unsuitability.
In some startups, analysts are assigned to some teams, but we do not attach much importance to the subject area: marketing, product or operating system – the requirements for the level of thinking described above allow you to switch from one to another (of course, it takes time to dive in).
Trainee data analyst: entry level jobs
A little higher I described the requirements for the analyst. These are screening requirements, meaning they must all be met even at trainee level. Most often, a trainee analyst does not have any relevant experience other than a university one. At the same time, he has the necessary qualities to grow into a strong analyst.
And here you may be surprised – isn't it too much? The answer is no, and here's why.
The trainee will spend all 3 months of his internship taking the time and energy of the manager and colleagues (and not adding a resource, as it may seem), so it makes sense to hire only guys with noticeable potential as interns in order to maximize the return on investment in training.
The most important criteria when hiring a person for a trainee analyst position are academic background, mindset, and programming experience at a university or on some projects.
The analyst works with statistical data. The intern may not have practical experience, but during the internship he will figure out why he had this course at the university. If there was no course, the amount of trainee training becomes too large. We at Yandex, most likely, will not take such a person to the team.
Way of thinking
This is a hard to formalize thing, but we are trying to come up with entry tests that do not depend on the candidate's work experience. For example, take the process that every resident of the city faces on a daily basis and formalize it in the form of product metrics.
Thus, we test the ability of a person to put himself in the place of the consumer and highlight the most important thing in his consumer experience. This is not something special that is taught – this is a way to look at the world, the level of empathy, a critical look.
Example tasks: come up with custom metrics for a traffic light, digitize the process of heating water in an electric kettle, and so on.
It's not about industrial programming. In the work of a data analyst, you need to be able to read documentation, quickly understand and use data tools, and automate your routine. These are SQL of different implementations, Python, Pandas libraries, visualization libraries, the ability to use the API.
People with a technical warehouse, as a rule, are faced with programming much earlier than they go to work, they pump technical skills easier and faster – this is what we want from people who have entered the profession.
Without organization and responsibility, the ability to communicate constructively with colleagues, it is difficult to achieve any result in a team. There are exceptions to this rule, but they are very rare.
The trainee analyst works at the level of precisely formulated and well-formalized tasks. Tasks for the trainee are set exclusively by the senior comrade-analyst (mentor or leader). He also checks the results of these tasks before giving this data somewhere else. The functions of checking, comprehending data, visualizing them and communicating the conclusions still lie with the leader, although the trainee may make his first approaches to these parts of the work.
Junior data analyst jobs
A junior analyst is a trainee who has mastered data processing tools. For him, there are no longer any restrictions in the tasks of transforming the available data to the required form. He can still do something non-optimally: unnecessarily heat the computing cluster with inefficient calculations, spend a lot of time on simple tasks, but he will solve the problem.
Data distrust and data validation
The experience of a junior analyst is already enough to not trust the data. When a junior analyst starts working with some data, he learns the nature of this data and makes checks: he makes sure that the data is exactly what is expected. For example, that the indicators are in the right dimension, the distribution of values looks reasonable, there are no strange outliers, the data reflects the real picture of the phenomenon under study, and so on.
The junior analyst lacks experience in a real product and business, so he tends to solve problems the way they come:
Do you need to upload an excel file with these columns? – Hold on.
Need to make a dashboard? – Draw on a piece of paper how it should look.
And so on.
As a rule, when setting a task for a junior analyst, a detailed algorithm is discussed with a description of the data to be used, a method for transforming this data (filtering, grouping, joins) and a literal description of how the result should look (if it is a graph, then what should be the axes , normalizations, signatures, rendering method).
A business customer can set simple tasks for a junior analyst directly, but this is not recommended. Ideally, all tasks should still be controlled by his manager.
As experience grows, the junior analyst can work more confidently with familiar data and with less task detail from the mentor. In this case, the flag for determining the level is how it works with unfamiliar data.
Implementation and application of the results of the work
As a rule, a junior analyst is not able to prepare an analytical report or study in the format of recommendations with a clear rationale. And even more so, he is not able to bring his recommendations and conclusions to the level of real changes at the level of the product, processes or business.
Junior analysts still provide little value to the business, so it's critical that they grow as quickly as possible. The principle that pumps a junior analyst is "do as I do." An older analyst (in the presence of a younger one) discusses a new problem with the project team, asks clarifying questions, dives into the context of the problem, and as a result of this dialogue, a solution approach emerges. Further, the senior comrade decomposes this approach into data processing tasks for a junior analyst. This allows you to observe how business problems turn into tasks for writing code and drawing graphs, and find their solution in them. Over time, the junior analyst should learn to do this on his own.
Why it is important to discuss the problem statement in depth with junior analysts
Typically, when running a project, there are questions and problems that analysts can help with. Along with the idea of going to an analyst, a business customer, as a rule, also comes up with an approximate solution, with a request for which he, most likely, will come (“make me such a schedule?”). But a project manager, designer or product may simply not have enough information about what other data and tools are available to get an answer. Perhaps there is a more precise way to give an answer, or an easier one. Perhaps, within the framework of the hypothesis that has arisen, it is not necessary to process the data at all, but it is enough to look at the dashboard, where there will be, if not ideal, but a graph that allows, with some assumptions, to answer the question asked.