Data Analysis in User Studies: Extracting Valuable Insights

You have spent weeks distributing questionnaires, conducting interviews, and pulling circulation logs from your library management system. Now you have a pile of responses and rows of raw numbers staring back at you. This is the moment where user studies actually earn their value. Raw data on its own tells you nothing. The skill lies in turning hundreds of scattered answers into clear patterns that explain who uses your library, what they need, and where your services fall short. This is what data analysis in user studies is all about: moving from collected facts to decisions you can defend.

What data analysis means in user studies 🔗

A user study is a research method designed to understand how people interact with library services, resources, and facilities. The aim is to understand user needs, preferences, and behaviours so that the library can improve what it offers. But collecting data through surveys, interviews, and usage logs is only half the job. Analysis is the stage where you systematically describe, aggregate, and interpret that data to answer the questions you set out to study.

Analysis works on two broad types of data. Quantitative data consists of numbers: how many respondents, what percentage prefer e-resources, how often a section is visited. Qualitative data consists of words: open-ended comments, suggestions, and complaints. Most serious user studies combine both, because numbers tell you what is happening while words often explain why. The investigator usually makes use of more than one method, particularly in survey research, and analysis has to bring these strands together.

Preparing the data before you analyse it 🔗

You cannot analyse messy data. Before any meaningful breakdown happens, raw responses must be cleaned and organised. This preparation stage is often skipped by students, and it shows in their results.

Editing the responses 🔗

Editing is the first check. The process involves removing unnecessary or inaccurate data and organising it so you can focus on the pieces that matter. Field editing happens right after collection, while details are still fresh in the investigator’s mind. Central editing is a more thorough review done later to ensure consistency across every questionnaire. Together, these checks prevent contaminated data from flowing into the analysis stage. Incomplete forms, contradictory answers, and obviously careless responses are flagged here.

Coding the answers 🔗

Coding turns responses into a form that software can process. It works by assigning numeric or alphanumeric codes to responses, which is what makes large volumes of data manageable. Closed-ended questions are easy because the answer choices are already fixed; a response of “postgraduate” might simply become the code “3”. Open-ended responses are harder. Here you use post-coding, where you read a sample of answers, identify recurring themes, build a category scheme, and then code the full set accordingly. A codebook records every variable along with the codes assigned, so anyone working with the data later understands what each number represents.

Analysing library user data through demographics 🔗

Once the data is clean and coded, the first real analysis is almost always demographic. You break responses down by characteristics like age, sex, and course of study. This tells you who is actually responding, and whether different groups behave differently.

Frequency distribution: the starting point 🔗

The simplest tool is the frequency distribution, a count of how many respondents fall into each category. You might find that 60 percent of respondents are undergraduates and 40 percent postgraduates, or that science students make up the largest share of library visitors. These counts and percentages are descriptive statistics, and you should always start with descriptive statistics, since they form the foundation on which everything else rests. Frequencies summarise a single variable at a time and give you the overall shape of your sample.

Cross-tabulation: connecting two variables 🔗

Frequencies alone cannot tell you how groups differ. For that you need cross-tabulation, also called a contingency table. A crosstab is a table showing the relationship between two categorical variables, where a plain frequency table can only describe one. This is where demographic analysis becomes powerful. Instead of knowing only how many male and female respondents there are, and separately how many prefer print versus digital resources, a crosstab shows you the intersection: how many male postgraduates prefer digital, how many female undergraduates prefer print, and so on.

This is exactly what makes cross-tabulation valuable in user studies. Frequency analysis examines the distribution of individual items, while cross-tabulation explores associations between demographic variables and behaviour. If course of study turns out to be linked with preferred service, that finding directly shapes how you allocate resources.

Checking whether a pattern is real 🔗

A crosstab might show a difference between groups, but you need to know whether that difference is genuine or just chance variation in your particular sample. This is where inferential statistics come in. Inferential analysis tests hypotheses and uses sample data to make statements about the larger population. The chi-square test is the standard tool for two categorical variables. It compares the counts you actually observed against the counts you would expect if there were no relationship at all. The key figure is the p-value, which estimates the likelihood of seeing a pattern this extreme if there truly was no relationship between the two variables. A small p-value, conventionally below 0.05, suggests the pattern is real.

One condition matters here. Inferential statistics are only appropriate when your sample was gathered using probability sampling. If you handed out forms only to whoever happened to be in the reading room, you cannot legitimately generalise to all users, no matter how large your sample is. A study on academic libraries, for instance, used a t-test to examine differences in awareness and perception based on gender at a significance level below 0.05, which is the kind of group comparison demographic analysis often calls for.

Identifying trends in user behaviour 🔗

Beyond who your users are, analysis reveals what they do. Libraries collect usage statistics precisely to support collection management, decision-making, and the identification of trends. The data here comes less from questionnaires and more from circulation records, gate counts, and electronic resource logs.

Spotting popular and neglected resources 🔗

Circulation data is a goldmine for understanding what gets used. By analysing loan records you can find which subject categories move and which sit untouched. One study mining patron book-loan data found that circulation data helps libraries decide which book categories to keep on site or withdraw, and even calculated the useful lifespan of books to decide when to relocate them to storage. The same logic applies to digital resources: download and access counts show which databases justify their subscription cost and which do not. Surveying users about e-resources can reveal that they prefer more online databases or find certain platforms hard to navigate, and usage logs confirm whether that preference shows up in actual behaviour.

Tracking change over time 🔗

A single snapshot is useful, but the real insight comes from comparing data across time. This is longitudinal analysis. Broad surveys provide baseline data, and if the same questions are repeated in later surveys, they offer longitudinal data to track changing patterns of use. Repeating a study annually lets you see whether footfall is rising or falling, whether digital use is overtaking print, or whether a new service is gaining traction. For numerical usage data, statistical methods can go further. One academic library project used linear regression and Pearson correlation to model electronic resource usage, quantify the trend, and predict the future, turning past records into a forecast that guides collection decisions.

Making sense of comments and suggestions 🔗

Open-ended responses need a different approach. Content analysis is the standard method, referring to the process of categorising text to classify, summarise, and tabulate the data. You read through the comments, assign codes to recurring ideas, and group those codes into broader themes. A complaint about “no quiet space” and another about “too noisy near the entrance” both fall under a single theme of study environment. Counting how often each theme appears turns soft feedback into something you can actually act on, and lets you display frequencies and patterns across many responses.

Tools for analysing library user data 🔗

The right tool depends on the size of your dataset and the complexity of your analysis. Libraries conducting user studies typically use spreadsheet, database, or statistical analysis software to manage and analyse their data.

Spreadsheets for the basics 🔗

Microsoft Excel is the workhorse for most small to medium studies. It is good for the simplest descriptive statistics and is convenient for data entry and reshaping. You can build frequency tables, create pivot tables for quick cross-tabulations, calculate means and percentages, and produce charts. For a typical college-level user study, Excel handles the bulk of the work comfortably.

Statistical packages for deeper analysis 🔗

When you need inferential tests like chi-square, t-tests, or regression, a dedicated package is better. SPSS, which stands for Statistical Package for the Social Sciences, is among the most widely used in academic research and performs statistical analysis on quantitative data through a graphical interface that handles even complex models without programming. Its menu-driven design makes frequencies, crosstabs, and significance tests accessible to those without a coding background. For students who want a free alternative, PSPP offers a similar look and feel and can perform descriptive statistics, t-tests, regression, and more.

For larger or more advanced projects, R is a strong open-source choice. It is both a programming language and an environment for statistical computing, and the College of New Jersey library project mentioned earlier chose R precisely because it is freely available with a rich set of packages for statistics and graphics. The trade-off is a steeper learning curve than Excel or SPSS.

Tools for qualitative data 🔗

Coding open-ended responses can be done in a spreadsheet, but specialised qualitative software helps with larger volumes. NVivo is a qualitative data analysis package that helps researchers organise and analyse complex non-numerical data, letting users classify, sort, and tag thousands of pieces of text. For modest studies, structured use of Word and Excel to group similar comments into themes works perfectly well and keeps the process transparent.

Bringing the analysis together 🔗

Good analysis follows an order. You clean and code the raw data, run frequencies to describe your sample, use cross-tabulations to compare groups, apply inferential tests where your sampling allows it, and analyse usage records and open comments to round out the picture. Each layer builds on the one before. The point is never the statistics themselves but the decision they support: which databases to renew, where to add seating, which user group is being underserved. A user study only matters when its analysis feeds directly into how the library is run, and the results should reach the people empowered to decide how those results will be applied.

What do you think? If you ran a survey in your own college library tomorrow, which single demographic breakdown, by course, by year, or by frequency of visit, would reveal the most useful insight about who is being underserved? And when a usage trend and an open-ended complaint point in opposite directions, which would you trust more to guide a decision?

References

1. https://www.lisedunetwork.com/user-study-in-library/
2. https://ebooks.inflibnet.ac.in/lisp4/chapter/user-studies-users-education/
3. https://www.lrs.org/library-user-surveys-on-the-web/
4. https://sociology.institute/research-methodologies-methods/data-analysis-survey-research-coding-tabulation/
5. https://medium.com/@xiaohaaa97/descriptive-vs-inferential-statistics-for-survey-data-analysis-fc745c354520
6. https://libguides.baylor.edu/c.php?g=1351162&p=9972302
7. https://www.researchgate.net/publication/245086225_A_Student's_Guide_to_Statistics_for_Analysis_of_Cross_Tabulations
8. https://viva.pressbooks.pub/sociology-research-methods/chapter/14-3-bivariate-data-analysis-crosstabulations-and-chi-square/
9. https://arxiv.org/pdf/2508.00906
10. https://arxiv.org/pdf/2501.03632
11. https://arxiv.org/pdf/2008.03545
12. https://www.clir.org/pubs/reports/pub105/section2/
13. https://www.tandfonline.com/doi/full/10.1080/0361526X.2019.1590774
14. https://www.geopoll.com/blog/coding-qualitative-data/
15. https://guides.library.ucla.edu/c.php?g=180764&p=1185385
16. https://slulibrary.saintleo.edu/c.php?g=1104741