Guide To Ensure Data Quality

Watch Charlotte Forbes (nee Penn), Associate Research Director of Explori's guide to ensure data quality

Data quality is critical to any data analysis and reporting to access insights. To ensure you’ll be working with the best quality data, there are various important factors to consider depending on the source of the data, such as surveys, registration forms, and online platforms. Let’s explore how to ensure data quality in three stages. 

Stage 1: Research design for data quality 

Surveys and registration forms offer great flexibility over what data you can collect. A few quality assurances from the outset of the research design process should be in place all the way through to the analysis stage. 

Question design

As a rule of thumb, you should ensure your participants are able to answer every question truthfully. Including options such as “Other”, “Not applicable”, or “Don’t know”. Attaching a “Please Specify” text box to collect random answers but leaving it optional can avoid hindering participants from completing the survey, where they can offer valuable opinions. Understanding that something is unknown or isn't applicable can be as insightful as other answer options. 

Keeping rating scales user-friendly and consistent 

Logical and consistent scales make it easier for you to compare the results of different questions against each other. There are two main types of rating scales, namely a numbered scale and a Likert scale.  

A numbered scale is used where you ask respondents to select a number that reflects their view. Both ends of the scale should be anchored by the least positive and most positive sentiment, then the numbers are used to grade the options in between.  Learn more about numeric values.

A Likert scale is used to represent people’s attitudes to a topic, and respondents choose the descriptive option on the scale that best aligns with their sentiment. When using a Likert scale, make sure there is a clear but gradual distinction between the sentiment of each option on your scale. 

A few more tips when using a scale

Ensure the scale options mirror the question. For example, for a question about event satisfaction, the options should relate specifically to satisfaction, not about usefulness or anything else.  

Consistent direction. No matter whichever scale you’re using, the direction of the scale should be consistent with either the most negative option first or the most positive option first. It doesn’t matter which direction you choose, but stick with one approach to avoid confusion. 

Unform point scales matters as inconsistent length of scales will make it more difficult to compare the results of different questions against each other. This saves you the trouble of re-weighting the scales if you use multiple scale lengths. By keeping things consistent, you’ll be able to focus on comparing and contrasting different results directly instead of spending your effort on analyzing your data on a question-by-question basis. 

Net Promoter Score is an exception. NPS should always be asked on an 11-point scale due to how an NPS is calculated.   

Minimizing potential errors

To ensure you’ll have the best quality data, you need the relevant answers coming from your respondents. you can limit outliers and irrelevant responses by a few simple measures: 

1. Use screening questions to filter out irrelevant participants.  
2. Use appropriate question types to reduce errors in data. For example, if you need a respondent to provide a number, set up the question for them to provide only a number, not a letter.  
3. Make open-ended questions optional. 

Stage 2: Data Collection and Sample Size 

We’ve covered measures to consider during research design. Let’s dive into the process of collecting data and sample size for some quality analysis. 

A good sample size 

Depending on the size of the universe you sought to collect data from, the response level tends to be in line with engagement, and it’s typically higher when it’s related to feedback on a specific experience or benefits the responder in some way. While a large sample is great, a representative sample is what’s important - that it reflects your overall universe. But if your sample size doesn’t mirror your universe, it’s possible to re-weight samples with the right software to have your dataset mimic the demographics or psychographic sample of your universe. 

Maximize the size of your data set 

Integrating data from multiple sources to avoid collecting the same information multiple times. You may even make certain types of data submission mandatory to ensure you have a complete set of responses from each participant.  

Motivate survey responses with incentives 

A small gesture may go a long way to encourage participation. The more engaging the universe, the higher the response level you may have to achieve a greater sample size. 

Stage 3: Optimizing data quality for analysis

There are two steps you can take to prepare the data for analysis is in the best quality. 

1. Data cleaning – removing any outliners, where the data provided is incorrect. And remove any duplicated responses that will skew the results. Filter out irrelevant respondents if screening questions weren’t in place in the research design.  
2. Numerical values for rating scales – assigning numerical values to rating scale questions gives you flexibility in reporting KPIs like satisfaction score. 

Take the steps at each stage to ensure data quality allows you to get the most accurate and insightful analyses to make the most out of your data.  

Want to explore more about data?

Check out our courses available to level up your event measurement skills today.