Data analysis steps – A basic guide

If you need to do some research to find answers to a business problem or a research question, you need to collect some data.
Joerg Hecker
Jörg Hecker
  1. Step 1: Planning the research
  2. Step 2: Preparing the data
  3. Step 3: Exploratory data analysis
  4. Step 4: Build a code system
  5. Step 5: Query your coded data and write up the analysis
  6. Step 6: Data visualization
  7. Step 7: Data presentation

If you need to do some research to find answers to a business problem or a research question, you need to collect some data. There are many ways of collecting data: You can collect primary data yourself by conducting interviews, focus groups, or a survey. Another option is to use secondary data sources. These are data previously collected for other projects, historical records, reports, statistics – basically everything that exists already and can be relevant for your research. The data you collect should always be a good fit for your research question or business problem. For example, if you are interested in how many people in your target population like your brand compared to others, it is no use to conduct interviews or a few focus groups. The sample will be too small to get a representative picture of the population. If your questions are about "how many….", "what is the spread…", etc., you need to conduct quantitative research. If you are interested in why people like your or another brand, their motives, and their experience, then conducting interviews can provide you with the answers, you are looking for.

Here are the seven steps for preparing a qualitative research study!

Step 1: Planning the research

As the saying goes: "Garbage in, garbage out." Suppose you find out after you have collected data that
• you talked to the wrong people
• asked the wrong questions
• a couple of focus groups sessions would have yielded better results because of the group interaction, or
• a survey including a few open-ended questions sent to a larger group of people would have been sufficient and required less effort.

Then it is too late, and you wasted a lot of resources. So think thoroughly about sampling, the questions you will be asking, and in which form. If you conduct a focus group or an interview, you are the research instrument, and your data collection will only be as good as you are. If you have never done it before, seek some training and practice. If you have other people do it, make sure they have the skills.

Figure 1: Memos offer the perfect space to plan and write about your research

Step 2: Preparing the data

When you conduct focus groups or interviews, think about how to transcribe them. Do you want to run them online or offline? If online, check out tools that deliver automatic transcriptions to save you time and money. The automatically generated transcripts still need to be checked. If you don't do this yourself, make sure that you instruct the person doing it on how to prepare the data.
• How should the final transcript be formatted for the later analysis?
• Which names and locations should be anonymized?
• What kind of speaker IDs to use?

Check out ATLAS.ti. It is a powerful tool to help you along the way. See the video below on how to import automated transcripts.

Some programs will immediately provide you with some basic descriptive level analysis if you have collected survey data. ATLAS.ti will support you with the analysis of the open-ended questions. For this, you need to export your data as an Excel file. ATLAS.ti's survey import wizard will guide you through the process. However, it can be helpful to clean up your Excel file first to optimize the data input.

Please allow marketing cookies to watch the video.
Manage cookies

Step 3: Exploratory data analysis

You can run a few simple exploratory analyses to get to know your data. For instance, create a word list or word cloud of all your data or compare and contrast the words in different documents. Let ATLAS.ti find find relevant concepts for you. Then, auto code the data and begin to review and refine the coding. Get a feeling for the sentiment in the data. Who is more optimistic, pessimistic, or neutral in their responses? Let ATLAS.ti auto code your data and then review and fine-tune. Another option is to browse through the data using the mentioned exploratory tools without getting a feel for the data and then begin to read and manually code.

Step 4: Build a code system

Whether you start with auto coding or manually coding, after having generated some first ideas, you need to get some order in your code system by building categories and subcodes, using folders to sort and order them. As this process requires reading and re-reading your data, you will become very familiar with your data. Therefore, you must have a good tool like ATLAS.ti qualitative data analysis software that supports you in the process and makes it easy to review your data, modify codings if necessary, change code labels, and write code definitions that explain what the code means and how to apply them.

Figure 2: Build a code system

Step 5: Query your coded data and write up the analysis

Once you have coded that data, it is time to take the analysis a step further. When using software for qualitative data analysis, it will be easy to compare and contrast groups of respondents. For instance, you can query the various opinions of female vs. male respondents? Is there a difference between consumers from rural or urban areas or among different age groups or educational levels? Which codes occur together throughout the data set? Are there relationships between various concepts, and if so, why? Check out the following article to see how you can analyze data using specialized software like ATLAS.ti: Code Co-occurrences with ATLAS.ti. Parallel to querying the data, you will begin to write summaries and interpretations in memos. Those will be the building blocks for the report.

Step 6: Data visualization

Data visualization brings your data to life. It is a powerful way of seeing patterns and relationships in your data. For instance, diagrams allow you to see how your codes are distributed across documents or specific subpopulations in your data. Exploring coded data on a canvas, moving around code labels in a virtual space, linking codes and other elements of your data set and thinking about how they are related and why – will all advance your analysis and give you further insights. Visuals are also great for communicating results.

Step 7: Data presentation

The final step is to summarize the analysis in a written report. You can now put together the memos you have written about the various topics, select some good quotes that support your writing and add tables and diagrams. If you followed the steps above, creating a report or presentation should be done in a breeze as all of the building blocks already exist, and you just have to put them together.

When preparing a report or a presentation, keep your audience in mind. Does your audience better understand numbers than long sections of detailed interpretations? If so, add more tables, bar or pie charts, and short supportive statements to your report or presentation. If your audience loves a good interpretation, add your full-length memos and visualize your analysis using conceptual networks and maps.

Figure 3: Visualize your data in Sankey diagrams (and more) with ATLAS.ti