| n (%) | |
|---|---|
| Design, Diversity and New Commons | 3 (7%) |
| Digital Economy, Internet, Ecosystems and Internet Policy | 3 (7%) |
| Digital News Dynamics | 3 (7%) |
| Digitalization and Networked Security | 3 (7%) |
| Digitalization and Opening up Science | 1 (2%) |
| Digitalization, Sustainability and Participation | 2 (5%) |
| Dynamics of Digital Mobilization | 5 (11%) |
| Methods Lab | 5 (11%) |
| Norm Setting and Decision Processes | 3 (7%) |
| Platform Algorithms and Digital Propaganda | 2 (5%) |
| Reorganization of Knowledge Practices | 8 (18%) |
| Research Syntheses | 1 (2%) |
| Security and Transparency of Digital Processes | 1 (2%) |
| Technology, Power and Domination | 1 (2%) |
| Weizenbaum Panel | 1 (2%) |
| Well-being in the Digital World | 3 (7%) |
| Working with Artificial Intelligence | 2 (5%) |
Weizenbaum Institute Methods Survey 2025
Introduction
Between November 18 and December 8, 2025, the Methods Lab conducted an internal survey on experience and needs regarding research methods, research software, and tools at the Weizenbaum Institute. Building on the first survey carried out in late 2022, it provides an updated picture of the current state and challenges in a research context that has evolved considerably since then. While the earlier survey was conducted at the outset of a new research phase, the Institute is now in the midst of ongoing projects at different stages, with corresponding shifts in methodological demands. In addition, reflecting recent developments, this year’s survey also addressed the emerging topic of AI in research practice. It brings together perspectives from all positions at the Institute (Principal Investigators, group leads, doctoral researchers, post-docs, and student assistants) and is intended to inform concrete next steps in developing training, consulting, and research infrastructure.
The survey was carried out by Roland Toth (Data Scientist), Christian Strippel (Lead), and Martin Emmer (PI). Items on the coordination and assessment of AI-related needs were developed by Jochen Knaus (Research Information Officer) as part of the AI Working Group at the Weizenbaum Institute.
Sample
A total of 44 people participated in the survey. With an estimated population of 117 researchers at the time, this results in a response rate of 38%.
Research group/unit
The sample covers almost all existing research groups and units at the Weizenbaum Institute.
Position
With the exception of PIs, all positions are represented among the respondents.
| n (%) | |
|---|---|
| Doctoral researcher | 19 (43%) |
| Research group/unit lead | 9 (20%) |
| Postdoctoral researcher | 7 (16%) |
| Student assistant | 6 (14%) |
| Other | 3 (7%) |
- Researcher
- Associated researcher
- Associate Researcher (PhD-level)
PhD stage
Out of the 19 doctoral researchers who participated in the survey, most see themselves in the research phase or final phase of their dissertation projects.
| n (%) | |
|---|---|
| Research phase (execution, adjustment, publishing of partial results) | 9 (20%) |
| Final phase (completion, disputation, publication, career planning) | 7 (16%) |
| Orientation phase (decisions regarding topic, supervision, financing | 2 (5%) |
| Initiation phase (writing a synopsis, familiarization with the research topic) | 1 (2%) |
Disciplines
Participants are predominantly coming from the social sciences. Large shares reported backgrounds in communication studies (34%), sociology (25%), and computational social science (23%). Other well-represented disciplines include science and technology studies (20%), psychology, business administration, and business informatics (each 11%). As respondents could select multiple disciplines, the percentages add up to more than 100.
| n (%) | |
|---|---|
| Business administration | 5 (11%) |
| Business informatics | 5 (11%) |
| Cognitive science | 2 (5%) |
| Communication science | 15 (34%) |
| Computational social science | 10 (23%) |
| Computer science | 4 (9%) |
| Data science | 4 (9%) |
| Design studies | 3 (7%) |
| Economics | 4 (9%) |
| Geography | 2 (5%) |
| Information science | 1 (2%) |
| Law / legal studies | 2 (5%) |
| Media studies | 3 (7%) |
| Pedagogy / educational research | 1 (2%) |
| Philosophy | 1 (2%) |
| Political science | 2 (5%) |
| Psychology | 5 (11%) |
| Science and technology studies | 9 (20%) |
| Science communication | 3 (7%) |
| Sociology | 11 (25%) |
| Visual and digital studies | 3 (7%) |
Other
- Gender Studies
Methods
Methodological experience and needs regarding training and consulting were queried in five different areas: IT tools, data collection, data analysis, software/tools, and AI tools. Below, you can switch between experience and needs in all areas. The results show that there is both expertise and needs for all methods and tools at the institute. Most participants with prior knowledge in methods estimate their experience level as rather experienced, indicating that the workshops and training provided by the Methods Lab in the past years had a considerable impact. The most-requested forms of support are training and learning material.
Tools provided by WI
This area was first introduced in this survey, focusing on research tools and services already provided by the IT department at the Weizenbaum Institute.
Out of the listed tools and services, participants indicated that they had most experience with LimeSurvey and the qualitative data analysis software MAXQDA (45%). GitLab (18%) and JupyterHub (5%) were only used by a smaller fraction of participants.
| Category | n (%) | Rather inexperienced | Rather experienced | Very experienced |
|---|---|---|---|---|
| LimeSurvey | 20 (45%) | 5 | 14 | 1 |
| MAXQDA | 20 (45%) | 3 | 14 | 3 |
| GitLab | 8 (18%) | 3 | 4 | 1 |
| JupyterHub | 2 (5%) | 0 | 2 | 0 |
| Other | ||||
| SPSS | 1 (2%) | 0 | 1 | 0 |
| virtual machine | 1 (2%) | 0 | 1 | 0 |
Despite the high experience with MAXQDA, participants still reported most needs for this software (25%). In contrast, only a few require further training and support regarding LimeSurvey (9%). Gitlab was the second-most indicated option (16%). This corresponds to the open-ended responses, where hands-on training for Git as well as computational reproducible research (using Quarto, Overleaf/LaTex), and scripting for HPC, were highlighted. Participants prefer hands-on methods, learning material, and exchange over consulting.
| Category | n (%) | Training | Learning material | Exchange | Consulting |
|---|---|---|---|---|---|
| MAXQDA | 11 (25%) | 8 | 4 | 6 | 4 |
| GitLab | 7 (16%) | 7 | 2 | 2 | 0 |
| JupyterHub | 4 (9%) | 2 | 2 | 1 | 0 |
| LimeSurvey (to create surveys) | 4 (9%) | 4 | 2 | 3 | 2 |
| Other | |||||
| SPSS | 1 (2%) | 0 | 0 | 0 | 1 |
| Overleaf | 1 (2%) | 1 | 1 | 0 | 0 |
Other kinds of training/support
- 1) Adivise on how to set up a workflow to cooperatively edit Quarto/Markdown-generated documents would be very helpful (best practices to bring non-coding colleages on board of a code-savvy project, in which manuscripts are edited from Quarto; how to integrate Overleaf (for writing) with Quarto (for code) etc.
2) Hands on Git-related training would be appreciated; with a particular focus on cooperative workflows
3) Training on bash basics would be helpful to make working with virtual machines more accessible to those new to Linux
4) Related, but not WI-specific: Hands on training on how to write batch scripts (e.g. to be sent to a HPC) would be helpful
Data collection
Consistent with the 2022 findings, participants reported most experience with established research methods: qualitative content analysis (70%), quantitative surveys (66%), qualitative interviews (57%), and quantitative content analysis (52%). By contrast, experience remains limited for more specialized approaches, including eye-tracking (2%), automated (9%) and non-automated (7%) quantitative observation, as well as experience sampling methods (9%) and diary studies (7%).
| Category | n (%) | Rather inexperienced | Rather experienced | Very experienced |
|---|---|---|---|---|
| Qualitative content analysis (e.g., text annotation, discourse analysis, hermeneutics) | 31 (70%) | 7 | 20 | 4 |
| Quantitative surveys | 29 (66%) | 10 | 10 | 9 |
| Qualitative interviews/focus groups | 25 (57%) | 2 | 11 | 12 |
| Quantitative content analysis (e.g., content coding) | 23 (52%) | 4 | 12 | 7 |
| API-based data collection | 20 (45%) | 6 | 6 | 8 |
| Qualitative observation / Ethnography | 17 (39%) | 3 | 13 | 1 |
| Automated content analysis (e.g., automated labeling, dictionaries) | 13 (30%) | 4 | 4 | 5 |
| Experiment | 13 (30%) | 4 | 8 | 1 |
| Web Scraping | 13 (30%) | 2 | 8 | 3 |
| Automated observation (logging, behavioral tracking) | 4 (9%) | 1 | 2 | 1 |
| Experience Sampling Method (ESM) | 4 (9%) | 1 | 2 | 1 |
| Diary method | 3 (7%) | 0 | 3 | 0 |
| Non-automated quantitative observation | 3 (7%) | 1 | 2 | 0 |
| Eye-tracking | 1 (2%) | 1 | 0 | 0 |
| Other | ||||
| Digital Ethnography | 1 (2%) | 0 | 1 | 0 |
Compared to the 2022 survey, there was a shift in reported training needs. Needs regarding technical data collection methods have declined, with the share of participants citing API-based data collection dropping from 27% to 16% and web scraping from 42% to 25%. Despite this decrease, web scraping remains one of the most frequently mentioned needs. Training in automated content analysis is still in demand (39% in 2022 compared to 30% now), and this is echoed in open-ended responses that call for more advanced NLP methods, highlighting the ongoing relevance of computational approaches at the institute. At the same time, qualitative methods remain important: needs in the area of qualitative content analysis (30% in 2022 vs. 23% now) and qualitative interviews or focus groups (18% in 2022 vs. 23% now) point to their continued role in digital and data-driven research. Across all areas, hands-on training is clearly preferred again, followed by learning materials.
| Category | n (%) | Training | Learning material | Exchange | Consulting |
|---|---|---|---|---|---|
| Automated content analysis (e.g., automated labeling, dictionaries) | 13 (30%) | 10 | 9 | 4 | 4 |
| Web Scraping | 11 (25%) | 5 | 8 | 3 | 3 |
| Qualitative content analysis (e.g., text annotation, discourse analysis, hermeneutics) | 10 (23%) | 5 | 3 | 4 | 3 |
| Qualitative interviews/focus groups | 10 (23%) | 8 | 3 | 2 | 3 |
| Quantitative surveys | 10 (23%) | 10 | 3 | 3 | 4 |
| Qualitative observation / Ethnography | 8 (18%) | 6 | 0 | 4 | 3 |
| API-based data collection | 7 (16%) | 4 | 5 | 2 | 2 |
| Experiment | 6 (14%) | 4 | 2 | 2 | 0 |
| Automated observation (logging, behavioral tracking) | 5 (11%) | 3 | 4 | 1 | 1 |
| Quantitative content analysis (e.g., content coding) | 5 (11%) | 4 | 0 | 2 | 0 |
| Experience Sampling Method (ESM) | 4 (9%) | 3 | 1 | 2 | 0 |
| Diary method | 2 (5%) | 2 | 1 | 0 | 0 |
| Eye-tracking | 2 (5%) | 2 | 0 | 0 | 0 |
| Other | |||||
| Advanced NLP (RAG, Agentic Workflows, Mechanistic Interpretability, ...) | 1 (2%) | 1 | 1 | 1 | 1 |
| Dynamic modeling of social media networks | 1 (2%) | 1 | 0 | 0 | 0 |
Other kinds of training/support
- I would be interested in learning more about the possibilities and challenges of analysing text data with generative AI. I would also need help with wht is best to translate speech into text.
Data analysis
Most participants reported experience with core data analysis methods, particularly data visualization (80%), statistical tests (66%), and regression analysis (59%). By contrast, fewer participants have experience with more advanced modeling approaches such as Structural Equation Modeling, multilevel, or mixed-effects modeling (20%), and deep or transfer learning (14%). Notably, among the small group with experience in deep or transfer learning (six participants), five describe themselves as very experienced, suggesting potential for targeted exchange and consulting formats in this area.
| Category | n (%) | Rather inexperienced | Rather experienced | Very experienced |
|---|---|---|---|---|
| Data visualization | 35 (80%) | 8 | 23 | 4 |
| Statistical tests (e.g., chi-squared test, t-test, ANOVA) | 29 (66%) | 7 | 13 | 9 |
| Regression analysis | 26 (59%) | 6 | 10 | 10 |
| Classification (e.g., Cluster analysis, LCA, LPA, Topic modeling, Naive Bayes Models) | 21 (48%) | 9 | 4 | 8 |
| Network analysis | 17 (39%) | 4 | 8 | 5 |
| Principal Component Analysis (PCA) | 13 (30%) | 7 | 4 | 2 |
| Natural Language Processing | 12 (27%) | 2 | 3 | 7 |
| Exploratory Factor Analysis (EFA) | 11 (25%) | 5 | 4 | 2 |
| Structural equation modeling (SEM) / Multilevel / Mixed-effects modeling | 9 (20%) | 2 | 5 | 2 |
| Deep / Transfer Learning | 6 (14%) | 0 | 1 | 5 |
| Other | ||||
| Qualitative content analysis (inductive, deductive) | 1 (2%) | 0 | 0 | 1 |
| various kinds of qualitative data analysis: situational analysis (Clarke); grounded theory (Glaser & Strauss), thematic analysis (Braun & Clarke), multimodal and semiotic analysis, etc. pp. | 1 (2%) | 0 | 1 | 0 |
| Grounded Theory | 1 (2%) | 0 | 1 | 0 |
| Exploratory Content Analysis | 1 (2%) | 0 | 1 | 0 |
Participants reported substantial training needs in areas where experience is already comparatively high, most notably data visualization (30%) and regression analysis (25%). This suggests a demand for more advanced or applied training rather than introductory support. At the same time, methods with more limited expertise at the Institute, such as natural language processing (23%) and Structural Equation Modeling, multilevel or mixed-effects modeling (20%), are also frequently identified as training needs, indicating a demand for introductory training. Like in the previous sections, participants indicated that they prefer hands-on training and learning materials over other resources. In open-ended responses, participants expressed interest in learning about the possibilities and challenges of analyzing text data with generative AI and guidance on best practices for translating speech into text, as well as qualitative approaches such as intersectional analysis and grounded theory. These requests mirror the broader demand for computational methods and qualitative analysis methods also observed in the data collection section.
| Category | n (%) | Training | Learning material | Exchange | Consulting |
|---|---|---|---|---|---|
| Data visualization | 13 (30%) | 13 | 6 | 4 | 1 |
| Regression analysis | 11 (25%) | 10 | 6 | 0 | 3 |
| Natural Language Processing | 10 (23%) | 7 | 4 | 3 | 1 |
| Network analysis | 9 (20%) | 8 | 2 | 3 | 1 |
| Structural equation modeling (SEM) / Multilevel / Mixed-effects modeling | 9 (20%) | 8 | 6 | 2 | 1 |
| Statistical tests (e.g., chi-squared test, t-test, ANOVA) | 9 (20%) | 9 | 5 | 0 | 2 |
| Classification (e.g., Cluster analysis, LCA, LPA, Topic modeling, Naive Bayes Models) | 7 (16%) | 7 | 5 | 1 | 2 |
| Deep / Transfer Learning | 6 (14%) | 5 | 2 | 2 | 2 |
| Principal Component Analysis (PCA) | 4 (9%) | 3 | 2 | 0 | 1 |
| Exploratory Factor Analysis (EFA) | 3 (7%) | 2 | 2 | 0 | 1 |
| Other | |||||
| Mechanistic Interpretability, Visualization of Activation Steering | 1 (2%) | 1 | 1 | 1 | 1 |
| Quantitative content analysis | 1 (2%) | 0 | 0 | 1 | 0 |
| Intersectional analysis | 1 (2%) | 1 | 1 | 1 | 0 |
| Grounded Theory | 1 (2%) | 1 | 0 | 0 | 0 |
| Qualitative Content Analysis | 1 (2%) | 1 | 0 | 0 | 1 |
| Bayesian Regression | 1 (2%) | 1 | 0 | 1 | 0 |
Software/tools
Overall, participants reported a high level of experience with research software and tools. Most have worked with survey tools (70%), R (64%), and qualitative or quantitative data analysis software (61%). Notably, experience with R has increased substantially since 2022, rising from 46% to 64%, and familiarity with markup languages has also grown, from 36% in 2022 to 52% in the current survey. This, again, indicates a considerable impact of the corresponding R workshops provided by the Methods Lab in the past years.
| Category | n (%) | Rather inexperienced | Rather experienced | Very experienced |
|---|---|---|---|---|
| Survey tools (e.g., Unipark, Limesurvey) | 31 (70%) | 6 | 16 | 9 |
| R | 28 (64%) | 12 | 9 | 7 |
| Qualitative data analysis software (e.g., MAXQDA, Atlas.ti, NVivo) | 27 (61%) | 4 | 19 | 4 |
| Quantitative data analysis or modeling software (e.g., SPSS, Stata, AMOS, MPlus) | 27 (61%) | 6 | 17 | 4 |
| Markup languages (e.g., LaTeX, Markdown) | 23 (52%) | 3 | 10 | 10 |
| Python | 21 (48%) | 8 | 5 | 8 |
| Network analysis software (e.g., NodeXL, Gephy) | 16 (36%) | 7 | 7 | 2 |
| Other | ||||
| Cpp | 1 (2%) | 1 | 0 | 0 |
Despite this generally high level of experience, participants continued to report strong training needs for programming languages, particularly Python (25%) and R (23%), as well as for qualitative (23%) and quantitative (20%) data analysis software. This pattern mirrors the 2022 results and points to the ongoing importance of these tools for (digitalization) research. It also reflects the broader demand for computational and qualitative methods training observed in the data collection and data analysis sections. Across all software-related needs, hands-on training is the preferred format again, while exchange and consulting were mentioned less frequently. In the open-ended responses, participants indicated the need for guidance for sharing code and data in open science repositories, and AMOS and MPlus training.
| Category | n (%) | Training | Learning material | Exchange | Consulting |
|---|---|---|---|---|---|
| Python | 11 (25%) | 8 | 6 | 4 | 1 |
| Qualitative data analysis software (e.g., MAXQDA, Atlas.ti, NVivo) | 10 (23%) | 7 | 4 | 4 | 3 |
| R | 10 (23%) | 7 | 7 | 1 | 1 |
| Quantitative data analysis or modeling software (e.g., SPSS, Stata, AMOS, MPlus) | 9 (20%) | 8 | 4 | 2 | 0 |
| Network analysis software (e.g., NodeXL, Gephy) | 6 (14%) | 5 | 4 | 1 | 0 |
| Survey tools (e.g., Unipark, Limesurvey) | 4 (9%) | 4 | 2 | 2 | 2 |
| Markup languages (e.g., LaTeX, Markdown) | 2 (5%) | 2 | 2 | 1 | 1 |
| Other | |||||
| LLM Agent Toolkits | 1 (2%) | 1 | 1 | 1 | 1 |
| Open Science Repositories (OSF, Harvard Dataverse, etc.) | 1 (2%) | 0 | 1 | 1 | 1 |
Other kinds of training/support
- Regarding the Open Science repos: Would be greate if the WI would nudge and support researchers in making their code and data available to other researchers + supporting them in handling privacy and copyright concerns
- A AMOS or MPlus crashcourse would help me to reactivate my skills (have used both programmes before, but haven’t used them now for a long time). But other learning materials for self-study also help.
AI tools
This area was first introduced in this survey, focusing on software using AI technology (e.g., chatbots), which has become increasingly relevant over the past years in everyday life as well as academic work. The corresponding survey items were developed by Jochen Knaus (Research Information Officer) as part of the AI Working Group at the Weizenbaum Institute.
Participants reported particularly high levels of experience with AI tools for translation (89%), text production (59%), as well as for programming, simulation, and information retrieval (45% each). By contrast, fewer participants reported experience with AI-supported speech processing (16%), the development of teaching materials (11%), or exam preparation (7%). When asked about specific tools they regularly use for work, participants most frequently mentioned ChatGPT and DeepL.
| Category | n (%) | Rather inexperienced | Rather experienced | Very experienced |
|---|---|---|---|---|
| Translation | 39 (89%) | 3 | 20 | 16 |
| Text production | 26 (59%) | 5 | 15 | 6 |
| Programming and simulation | 20 (45%) | 5 | 9 | 6 |
| Researching information | 20 (45%) | 6 | 10 | 4 |
| Text analyses and text processing | 19 (43%) | 4 | 11 | 4 |
| Literature search and study | 18 (41%) | 4 | 14 | 0 |
| Concept development | 14 (32%) | 4 | 8 | 2 |
| Data analysis, data visualization and modeling | 11 (25%) | 4 | 6 | 1 |
| Communication with students/colleagues | 9 (20%) | 3 | 3 | 3 |
| Problem solving and decision-making | 9 (20%) | 3 | 4 | 2 |
| Speech processing (acoustic, speech-to-text) | 7 (16%) | 3 | 3 | 1 |
| Development of teaching content | 5 (11%) | 0 | 4 | 1 |
| Exam preparation | 3 (7%) | 2 | 1 | 0 |
Specific tools
- DeepL, ChatAI
- ChatGPT, Mistral, Gemini, Grammarly
- Coursor, Whisper
- DEEPL
- I prefer to use my own brain as my job is boring enough as it is. I have used AI to translate German text from a co-author into English and have received an AI-translated text for the English version of a textbook (what the publisher pushed and the editors wanted to try). In both cases, the translation needed a lot of work to get it ready for publication, so even this rather tedious work is not satisfactorily done via AI (at least not in 2025).
- ChatGPT, DeepL, ResearchRabbit
- deepl, chat ai, noscribe
- deepl, chatgpt
- Google Translate; DeepL
- DeepL, Google Translate, ChatGPT, ChatAI, ResearchRabbit, LitMaps
- ChatGPT Edu
- DeepL, Google Search AI summary, Trint
- ChatGPT, ChatAI, deepL, elicit
- ChatGPT
- deepl, chat gpt, chat ai
- mistral chat, chat AI and voice AI (by academic cloud), chat pdf
- Chat AI, ChatGPT
- Mistral, OpenAI (Academic Cloud)
- ChatGPT, Deep
- ChatGPT, Claude via Cursor, Spark
- Grammarly, ChatGPT Pro
- ChatGPT, ollama, rollama, llama-family of models, Research Rabbit, DeepL, transformers models, huggingface
- Notebook LLM
- Deepl
- Grammarly, DeepL, ChatGPT (rarely)
- DeepL, Claude AI, ChatGPT, Afroféminas GPT
Reflecting the high experience level, only few participants indicated a need for further training in AI-based translation (2%). In contrast, substantial training needs are reported for text analysis and text processing (27%), for data analysis, data visualization, and modeling (20%), as well as literature search and review (20%). Similar to the other areas, hands-on training is clearly the preferred learning format here.
| Category | n (%) | Training | Learning material | Exchange | Consulting |
|---|---|---|---|---|---|
| Text analyses and text processing | 12 (27%) | 8 | 6 | 2 | 3 |
| Data analysis, data visualization and modeling | 9 (20%) | 9 | 4 | 2 | 1 |
| Literature search and study | 9 (20%) | 7 | 7 | 1 | 2 |
| Concept development | 8 (18%) | 6 | 5 | 1 | 1 |
| Text production | 5 (11%) | 4 | 2 | 2 | 2 |
| Programming and simulation | 4 (9%) | 3 | 2 | 3 | 0 |
| Problem solving and decision-making | 3 (7%) | 2 | 1 | 1 | 2 |
| Researching information | 3 (7%) | 3 | 2 | 0 | 0 |
| Speech processing (acoustic, speech-to-text) | 2 (5%) | 1 | 1 | 2 | 0 |
| Development of teaching content | 2 (5%) | 1 | 1 | 0 | 0 |
| Translation | 1 (2%) | 0 | 0 | 0 | 1 |
Timing
To help plan future workshops and training, we asked participants which months and days of the week typically work best for them.
Month
March is clearly the most popular month for workshops and training, with 80% of participants indicating their preference. It is followed by February (64%), November (59%), and April (57%). Like in 2022, this indicates that spring and fall/winter are the best seasons for workshops and similar events, while the summer months are least suitable: July (18%) and August (14%) are rarely preferred, likely due to holiday periods.
| n (%) | |
|---|---|
| January | 17 (39%) |
| February | 28 (64%) |
| March | 35 (80%) |
| April | 25 (57%) |
| May | 23 (52%) |
| June | 12 (27%) |
| July | 8 (18%) |
| August | 6 (14%) |
| September | 16 (36%) |
| October | 20 (45%) |
| November | 26 (59%) |
| December | 21 (48%) |
Day
Like in 2022, Thursday stands out as the most convenient day for workshops and training (70%), followed by Wednesday (64%). Monday was indicated least frequently (52%).
| n (%) | |
|---|---|
| Monday | 23 (52%) |
| Tuesday | 24 (55%) |
| Wednesday | 28 (64%) |
| Thursday | 31 (70%) |
| Friday | 24 (55%) |
