How to get into data analysis

content data analysis

There is a clear appreciation for data analytics among accountants, and it’s not hard to see why. Data helps us look back in time, spot patterns and ask deeper questions.

From auditors using analytical techniques to test processes and detect anomalies, tax accountants carrying out scenario testing to manage tax strategies, and financial planners anticipating margin shifts for future budgeting, data analysis enables accountants to measure rates of growth, and shrinkage, with flexibility.

So, what are the key skills you need to add to your stack to become the go-to data analyst?

Prep and cleansing

Most often, raw data is rich but messy. Data requires cleansing, transforming and classifying to achieve the appropriate level of counting, summing, averaging and ultimately insight.

Data engineers use a series of tools to consolidate data into a neat, analysable state. Common examples include coded (Python, SQL, Java), no-code (Airtable, Zapier, Bubble, Databricks Lite) or low-code solutions (Fabric, Azure Data Factory, Databricks, Python and SSIS).

An engineer’s role is to use algorithm-based operations to clean data. Engineers locate the data into a flat-file or database-table format for an analyst to pick up. In most companies, analysts perform data-cleansing activities to ensure the data is in a neater, analysable state.

Contextual knowledge

Analysts understand the business environment and can identify how the data represents business activity. They typically answer questions about what processes generate the data, how information affects activity and decisions, and what impact it has on performance measures.

Developing technical data skills enhances an analyst’s ability to quickly develop automated reports, set sensitivity alerts and scenario plan upcoming changes.

Technical skills

Excel should not be dismissed; it is a very versatile data analytics tool. Modern data-gathering, however, renders it unable to analyse large data sets.

Low-code platforms are very good at enabling the compilation and presentation of data. The more complex the business is – for example, a bank – the more useful coded techniques become. Popular languages in the data space are SQL, T-SQL, Python, R, JSON and C#.

As for data visualisation platforms, it is simply a matter of picking one and getting familiar with it (Power BI, Tableau, Qlik, Google Charts, Zoho, IBM Watson, SASS, SAP, Matplotlib and R are some of the more commonly used tools). Power BI has an embedded query language of its own. M-Query and DAX have something similar.


 

Get ACCA Careers working for you...

Search hundreds of roles from all over the world on ACCA Careers

Sign up for a job alert tailored to your desired location and role


Data visualisation

Storytelling with data is crucial to ensure stakeholders understand the information presented and can deep dive when needed. Understanding the purpose of the information empowers the stakeholder. An analyst must therefore stack the information in a way that enables business users to act on the information provided – for example, trend lines to illustrate growth and shrinkage, barcharts to show side-by-side differences between products or departments, boxplots for distribution comparison, etc.

Data validation

Data represents an event, process or population. Learning statistics techniques will help you understand how best to use the correct data or the correct level of sampling.

Taking things back to the core principles of accounting in reliability; carrying out validation checks to ensure the data is reliable through additional data tests like duplication checks; ensuring completeness or representation (how the sampled data preresents the whole population); looking for error traps (to ensure the code has not manipulated the data to a point that it no longer represents the activity being analysed); learning how to design good tests – all these help an analyst answer accuracy and reliability questions from stakeholders.

Where to start

So, which of the tools should you learn first? Excel is still a versatile data analytics tool, providing the ability to present quick-fire analytics and mock-up solutions. Now, with the added Python capability, it’s even more powerful.

Visualisation platforms, and low-code and no-code environments, can achieve a lot without a detailed coding knowledge. However, investing time in coding and understanding database principles helps an analyst develop efficient systems.

Database platforms can be interrogated with SQL and NoSQL platforms. NoSQL works in the same way as a low-code environment, where you can select fields and apply filters to get the answers needed. When embarking on a coding journey, SQL is a crucial language to learn.

Libraries are catalogues of pre-coded functions which allow you to call a name and it will perform an action – for example, ggplot in R will take the data you feed it and draw the chosen graph for you. In Python, NumPy and Pandas libraries support data importing and cleansing, whereas Scikit and Matplotlib libraries work similarly to R and MATLAB in performing advanced mathematical computations for AI and machine learning.

If the analytical task is pattern recognition or time-series analysis, then Pythos, R and MATLAB will create graphs and plots for insights. The advanced work will be in applying models for predictive ability.

Start with SQL and a visualisation tool; 80% of what you are presenting can be done using these two elements. Excel will still be there as you transition to SQL. You will still need that familiar space to cross-check your SQL coding. When advancing into the predictive analysis space, get familiar with regression techniques and statistics methodologies as you stretch your skills into predictive analysis.

Finally, consider joining forums. These can be very helpful in uncovering shortcuts, tips and tricks.


More information

Watch our webinar, ‘Skills for the future: data analytics in audit and the finance function’

This article was first published in AB magazine June 2024

Back to listing