Data analysis and management: friend or foe?
Tim Bottelbergs, Partner BDO Digital
Steven Cauwenberghs, Partner BDO Risk Advisory
Walter Vanherle, Partner BDO Digital
Every organisation produces data. But not everyone actually does something with them. Yet this goldmine of information can be the compass your organisation needs to navigate the increasingly digital corporate landscape. Good data analyses and management are indispensable in this regard. How? Asking the question is answering it…
I’m at the beginning of my data analysis journey. What is the best way to start?
You don’t have to open the big box of tricks straight away. Start small and refine as you go along. For example, start by centralising data from operational systems in a data warehouse – which you can use to build reports and dashboards to obtain the first insights. Important: guard data quality, privacy and security from day 1. Make sure that the data are documented and that it is clear how these data are processed before reporting on them. As you progress, you can experiment with all kinds of data science techniques and machine learning models (in the form of proofs of concept) in order to gain ever deeper insight into your business and the value of your data. You can then use these insights operationally in processes to support your staff in their daily activities (perhaps via Robotic Process Automation – also read our free white paper ‘Robotic Process Automation’).
Are data analyses expensive?
A data analysis project does not have to cost a great deal of money. You can, for example, centralise data in a cloud solution – whereby you only pay for what you actually use, in terms of both storage and the tools necessary for extracting data from operational systems and transforming them. Moreover, many analysis and visualisation techniques are already embedded in open-source tools such as PowerBI and Python.
Do I need to bring in ‘deep learning’ experts to carry out the data analyses?
Large companies, such as Google, Facebook and Uber, rely on top-shelf models such as neural (‘deep learning’) networks. However, there are less complex models available in the market that don’t require you to know all of those algorithms. Linear regression and decision trees often go a long way. Still need more advanced techniques? Then these solutions are often available (ready for use) via an API (Application Programming Interface), with or without a fee. If you envisage something bigger and want, for example, to train up on and maintain machine learning models, it’s best to seek advice from specialised parties.
Why is data quality becoming increasingly important?
As you make use of more data analyses, you will increasingly use the results of visualisations and models as the basis for business decisions. So, it’s essential that what is shown in these dashboards and reports is correct and accurate. Also, to avoid erroneous outcomes, document where the information comes from and how it must be interpreted.
How do I handle data privacy?
Always ask for the individual’s explicit consent before storing data. Make an inventory of all of the personal data processed within the organisation and check that the processing is warranted. Keep and process only the data that are required, and anonymise as much personal data as possible in order to limit the impact of possible data leaks. Data are valuable, so protect them as if they are your most precious possession. In the event of a data leak, inform the persons concerned. Finally, make your employees aware of the importance of data privacy and cyber security by regularly organising events such as phishing campaigns, in which you make clear to them the dangers of fraudulent e-mails.
How do I get the most out of a machine learning algorithm?
A cost-benefit analysis provides insight into various scenarios with a detailed multi-year or quarterly time axis, the anticipated costs and revenues, and the risks and expectations. The costs include: the time of your own, and possibly external, employees (converted into euro), and the IT costs such as infrastructure and software. You enter this as either one-off or recurring costs. The revenues are expressed in so-called Financial Impact Templates (FIT) that estimate the savings, time in euro and higher revenues for each process, as well as the impact in time, expressed in percentages. The FITs also indicate the expected realisations that you must follow up.
Is the IT department responsible for data management?
IT is not solely responsible. Everyone in the organisation must become more aware of how the company handles data, what their impact is, and how that should be interpreted. From input through processing to output – you must set up the correct processes, embedded in a data policy, whether or not monitored by a Data Privacy Officer. The DP Officer also indicates what is, and what is not, possible with regard to data quality, privacy, etc.
Questions about data management?
Are you looking for help with the analysis of your situation? If so, please do not hesitate to contact our specialists: firstname.lastname@example.org, email@example.com or firstname.lastname@example.org
Also, read our white paper ‘From waste to fuel – take data-driven decisions with AI and Machine Learning’.