Getting started as a data analyst
In very simple terms, a data analyst job is to translate numbers into plain English so the business can gain a better understanding of performance. However, in terms of their actual day to day routine, it’s a bit more complex than that. In this article, we will be discussing the 5 + 1 things that a data analyst does in his or her day job and the skills required for each step to be successful. Additionally, we will touch on the kind of tools and data, analysts use in each step.1. Reporting
As a data analyst, you will be spending a lot of your time either creating, refreshing or running reports. These reports will likely be in the form of Excel or PowerPoint with commentary, and use visualisation tools like Power BI, Tableau & Qlik Sense. If you are creating a new report, then you will probably be using SQL, Excel or any other database; as it’s the place where the raw data will be saved after you collect it from other people or data sources. When I started working as an analyst, I was given responsibility for 5 different reports. I had to make sure that all the reports were refreshed on time, with the correct data and that they were sent out to the right stakeholders at the right time. For example, I had a marketing report, a retail report, a product report, etc and each report had a different group of stakeholders. This is actually a good way to start your career as you get exposure to the wider businesses and start to understand how different teams use their reports. This will give you the confidence to start creating your own reports. The kind of skills you need in this initial step are the following:- data gathering skills
- data cleaning skills
- data transformation skills
- data storing skills
- technical Skills – SQL, Excel, database knowledge, BI/Visualisation
- organisational skills & attention to detail
- punctuality – ensuring the reports are out on time
2. Analysing the data
After refreshing the reports, you will be spending time looking at the data and trying to understand it, looking at the patterns and the performance of what you are reporting. In some cases, the reports alone will not be sufficient for your analysis and you will have to spend extra time pulling and analysing more data from the database in order to complete your analysis. After the analysis, you will be writing your commentary/insights – think of this as a story of your findings. While going through the process described above you will then to need to check you work for any obvious errors/mistakes in the data/reports, having business knowledge is critical here. The kind of skills you need in this step are the following:- problem solving – no solution will be the same hence your problem-solving skills will be challenged
- ability to ask the right questions
- business knowledge
- fixing mistakes
3. Presentation of results
After finishing your analysis, you will have to present the results back to the business. This will usually be in the form of a PowerPoint presentation. This is often the skill that the business values the most in a data analyst. It does not matter how good your coding/ modelling/predicting skills are if you are not able to turn your findings into insights and business recommendations. By mastering the ability to turn data into insights and being able to communicate this to the business will see your career start to move in a positive direction. The kind of skills you need in this step are the following:- insights
- storytelling
- communication skills
- agile thinking
- strategic thinking – align your analysis with the business’s strategic objectives
4. Ad hoc analysis
The next thing you will likely be doing is ad hoc analysis. Ad hoc is usually a specific request that crops up on the back of your initial presentation or frankly any meeting with stakeholders in general. It’s typically just a one-off analysis on a subject of interest. For example, you have identified from your analysis that category X had an amazing week last week. Maybe your manager will ask you to do an ad hoc analysis on why category X performed so well? What are the drivers? Ad-hoc analysis is a good way to show your creative thinking and problem-solving skills as it’s different from the standardised reports that the business has. In this step, you can also demonstrate your machine learning (ML) skills (if you have some); that is if there are applications of ML that would benefit the problem you are trying to solve verses what can happen which is trying to force ML in because you think it will impress. The kind of skills you need in this step are the following:- problem solving
- data gathering
- asking the right questions
- time management
5. Automating all processes
As an analyst you will have to learn to work smart and not just hard. The majority, if not all reporting should be automated using SQL (or similar) to Excel or SQL to one of the visualisation tools. There is a smart way of creating reports so that as soon as there is new data in the database, all you have to do is press “run” and “refresh” and within 10 secs all the data is cleaned, transformed, modelled and refreshed. The obvious benefit of this, is that you can save a significant amount of time which you can then spend analysing the reports and adding more value to the business. The kind of skills you need in this first step are the following:- technical skills
- methodical thinking – you’ll need to think hard about the order of doing each task and when and why.
- ability to see the bigger picture – if you recognise the benefits of automation then you can really add value to the business.
6. Build machine learning models
This step is mostly for advanced data analysts that know how to use machine learning. It’s not something that an analyst will spend much time doing but at some point in your career, you will build or be involved in building a machine learning model. Personally, I believe that the urge to learn will come naturally in all good data analysts after mastering the basics discussed above. The kind of skills you need in this first step are the following:- ML knowledge – start by learning some basic models like linear regression, logistic regression, naïve Bayes & decision trees for supervised learning & k-means for unsupervised learning.
- Python or R skills – I recommend Python but R is equally good.
- statistics – correlation analysis will be very useful
- advanced ML libraries like TensorFlow, Keras and PyTorch (GPUs or VMs)
