Why data can lie

This is why you should not fully trust your data

Why you shouldn’t fully trust data

We make decisions based on different data every day so that we don’t have to base these decisions on intuition. By doing this, we assume that we are making the right decision. Since we rely so heavily on the underlying data in this way, it is important to be aware that data can lie – especially when we make important decisions that can have far-reaching consequences for our business. If the results of the analysis are inaccurate or wrong, they can lead to incorrect decisions, which in turn can lead to financial losses, a bad reputation, customer loss, or even legal disputes. In addition, faulty data analysis results can also undermine confidence in a company’s ability to analyse data and can thus negatively impact business outcomes. Therefore, it is crucial to be aware that the data can sometimes be misleading and to take appropriate measures to ensure that the analysis results are always reliable and accurate.

Why data can lie

There are numerous reasons why data can be misleading:

1. Faulty Data:

If the data fed into the BI solution is faulty, it can lead to incorrect conclusions. Various factors can be responsible for this:

• Data Quality: If data quality is not high, the data used may be incorrect, incomplete, or inconsistent. To learn why this can happen and how good Data Quality Management helps to ensure high data quality, read the article “Data Quality Management“.
• Timeliness: If the data on which the analysis is based is outdated, the analysis cannot reflect current market conditions or changes in the industry, which can lead to false results and reduce the relevance of previous analyses and data.
• Data Source: If the data sources used are not good, the analyses in your BI system can also be wrong. For example, data that has been manually entered or combined when from different sources can be incorrect, incomplete, or inconsistent. Furthermore, if data sources are not correctly linked, analyses can be biased. This happens, for example, if you want to analyse which countries most of your customers come from, but do not consistently maintain the country in your CRM system when creating customer data, so that the data source is inadequate. Therefore, it is important to ensure that the data comes from a reliable source.

2. Poor Data Collection:

If the analysis techniques used in data collection are incorrect or not applied correctly, the results can be biased. For example, a poor choice of statistical methods or an inadequate interpretation of results can lead to false conclusions. Lack of validation can also be problematic as without it, it is not ensured that the analysis methods or algorithms were correctly applied and that the results are plausible and understandable. This can happen, for example, if you conduct a survey but do not take all relevant factors into account, ask the wrong questions, or do not clearly define the answer options. This influences the results of the survey, so the result is not correct. It is also possible that the analysis method is influenced by unconscious biases beforehand, so that the results of the analysis are incorrect.

3. Selection Bias:

If the data used for analysis is not representative, this can lead to bias. The same applies if the data is not sufficient to adequately represent the question or problem. This is the case, for example, if you collect product reviews on a specific platform that is mainly used by a special segment of your customers. In this case, the reviews are not representative of the opinion of all your customers. The picture you get through the reviews is therefore distorted.

4. Processing Bias:

If the data is not processed correctly, this can also lead to bias. For example, if you collect data to capture the income level of your customers but define the upper and lower limits of income incorrectly, the results of the analysis can reflect a distorted picture.

5. Faulty Interpretation:

Interpreting the available data correctly and drawing correct conclusions is a difficult task. If the person performing the analysis does not have sufficient expertise or understanding of the underlying data, mistakes can quickly occur. For example, most customers are more likely to be tempted to give an online review of your company or product if they are dissatisfied with the performance. If you are not aware of this, you could conclude from negative reviews that the majority of your customers are dissatisfied, even if this is not the case.

6. False assumptions:

If the assumptions underlying the analysis are incorrect, the results can also be biased. For example, assuming incorrect market conditions or customer needs may lead you to believe that customers would not buy your product if a certain feature were missing. However, customers may find this particular feature unnecessary or may buy your product for other reasons.

7. Disconnected representation:

If data that are related to each other are analysed in a disconnected way, this can also lead to a distorted view of reality. For example, it is well known that chocolate Easter bunnies are popular two months before Easter. However, since Easter does not occur on the same date every year, a pure view of chocolate bunny sales disconnected from the dates of the holidays could be misleading. The data would show an increase in sales in spring and a sudden drop in sales at some time, but the reason for this would not be clear.

8. Faulty application:

If the BI solution is not used for the purpose for which it was developed, this can also lead to incorrect results. For example, if you use a solution developed for financial analysis to evaluate the performance of your production, not all relevant data may be reflected, resulting in erroneous analysis results.

What can be done to avoid these errors?

To avoid the above-mentioned errors when using BI solutions, appropriate measures can be taken:

1. High data quality:

Companies need to ensure that the data fed into the BI solution is clean and accurate. This can be achieved by conducting quality controls and continuously reviewing data for completeness and accuracy. Also, avoid using outdated data or data from unreliable sources.

2. Proper data collection:

It is also important that the data used for analysis is representative. Therefore, think carefully about where and how you want to collect the data and which data or factors are relevant. Also, validate the data, for example, by using representative samples. When selecting data, make sure they are relevant to your question or problem.

3. Correct processing:

Correct data processing can be achieved through standardized procedures as well as verifying results for logical errors.

4. Proper interpretation:

Proper interpretation of the available data can be very difficult. Therefore, ensure that anyone who analyses and interprets your data is sufficiently trained. Additionally, the assumptions made during analysis should be carefully considered and documented so that they can be traced at any time. If faulty assumptions are made, this will be quickly identified, and the analysis can be corrected. Also, avoid viewing related datasets in a disconnected manner.

5. Intended use:

A BI solution must be used solely for the purpose for which it was developed. To ensure this, users should be adequately trained. If you initially decide to use your BI solution for a specific purpose but want to use it for other purposes later, the software should be adapted to accurately reflect all relevant data.

Conclusion

Following the motto “never trust a statistic you haven’t manipulated yourself”, data in business intelligence solutions can sometimes be misleading. Since decisions based on such data can have far-reaching consequences, it is important to be aware of this issue and take measures to ensure that analysis results are reliable. Despite the possible challenges, analysing data and using a BI software such as myPARM BI^act is very useful and an important part of a successful business strategy.
It is also possible and sensible to use a BI software like myPARM BI^act to verify and visualize the quality of existing data. For example, incomplete or incorrect data, as well as inconsistencies regarding logical connections and dependencies, can be displayed directly in a dashboard.

Would you like to get to know myPARM BI^act in a demo presentation? Then make an appointment with us right away!