Editor’s note: Tim Macer, managing director of U.K. consulting firm meaning ltd., writes as an independent software analyst and advisor.

The Internet has become a very natural place to carry out qualitative research, as it is an environment where many harder-to-reach groups like senior professionals or the under 25s are to be found, not just in abundance but often in a conveniently loquacious mood. And therein lies the hidden cost of online qual - it can be easy and inexpensive to collect large amounts of textual data, without the normal burden of room hire or transcription fees. But the text arrives unobserved, often in a silent, relentless avalanche that can overwhelm the researcher. It takes time for the researcher to read and organize and assimilate all this data - there is no equivalent to being in on the groups and being able to get thinking right away on what to put into the report. The only way to the findings is by reading and reading and reading.

Coping with the silent avalanche

While some researchers experiment with using sentiment analysis or text mining, one overlooked remedy is that stalwart of the academic social researcher - NVivo - as it still places the researcher at the center of the interpretation while leaving plenty of room for her or his skill and intuition.

With the arrival of NVivo 9 at the end of last year, offering capabilities to work with many other kinds of data, I though it was time to see how it would shape up in the context of commercial market research - a sector where the software currently has few users.

On first opening the software, the interface can be bewildering - but think back to the first time you opened Word or Excel, if you can remember that far back. In NVivo, a number of sections and tree-structures in a panel on the left allow you to organize your work and a varying, context-specific set of tools are presented at the top, organized into tabs. The rest of the space is organized into two horizontal panels. The upper one displays selectable lists or catalogs from which you can select items. The lower panel is a tabbed window where source documents, recent reports or any other content appears.

Working with a tool like this you are likely to follow a rough workflow of discovery and exploration, classification and gathering together and then working through what you have gathered in order to extract meaning and summarize the findings. However, NVivo is not at all prescriptive about how you work: You can start with all your analysis frameworks fully developed or you can let them fall out organically from the data or steer a course somewhere between those two extremes.

The stepping stones

There are a few key concepts in the software it is worth understanding, as these are the stepping stones over which you will make your journey through the data.

“Sources” are the documents or transcripts you will work on, which NVivo will import into its internal database. Sources can also include other notes, reports or stimulus materials - and a wide range of formats are supported such as Word files, PDFs, tables in Excel and even photos or video. NVivo offers excellent support for video, making it a great tool to use to analyze conventional focus groups or vox pops too.

The next key concept is the “node,” which is your principal means of categorizing information. If you have an interview guide, this could be imported and used as a set of nodes - nodes can be hierarchical to any depth and they are easily reorganized simply by dragging and dropping. A common way to use NVivo is to run through your transcripts, marking up sections and dropping them into the appropriate node. The interface makes that task pretty efficient - just a few mouse-clicks to post words or phrases into the right node.

You can also create node structures on the fly, adding categories or subcategories when they start to emerge. But there are some other smart ways to create nodes which are particularly helpful in dealing with very large volumes of text. One way is to use “autocoding,” which probably does not do what you think it might. If your documents are already structured in some way (e.g., the heading structure denotes the question, the topic or the speaker in a group) autocoding can be used to create nodes from the heading structure and then populate these with all the relevant passages. It makes a lot of sense if there are several transcripts that share a similar structure.

Word frequency counting

Perhaps more interesting to those handling Web transcripts is another autocoding method based on word frequency counting. NVivo identifies words and phrases that occur multiple times across your sources, creates nodes for them and then posts sections of the text around that word into the node along with all the context, such as the person and their demographics. There are a total of seven different query tools for creating advanced searches based on other attributes such as demographics and the results of any of these can also be used to create and populate nodes. All of these tools can provide useful alternative perspectives to the ones you may start out with.

An increasingly common feature of Web-based qual, particularly projects that have an auto-ethnographic side to them, is the use of photographs taken by the participant and even video or audio recordings. This is an area very well-supported by NVivo. Files in any of these formats can be imported and then these too can be categorized.

The third and more obvious concept is that of the “classification,” which makes most sense when used to apply demographics or other fixed categorizations such as location. Classifications tend to apply to the whole document or source element, rather than specific passages within the source.

Once you have categorized and coded your source material, it is time to start pulling out the findings - and there are two vehicles to use here. There is a series of built-in reports that reveal all of your coded snippets in, for example, node or category order. A report wizard also lets you create your own reports based on any organizing principle you can construct. You can view these or export them to work into your final report.

An explore tab in the tools area at the top also provides you with a broad selection of ad hoc query tools that you can use to interrogate your data. Alongside these, there are also tools to visualize data that nudge you in the direction of text mining. For example, you can perform cluster analysis on the word similarities between different documents or sections, with a resulting visualization showing word proximity. Cluster analysis can also be performed on nodes, with clustering revealing the similarity of the different coded fragments within each node. None of these tools on their own provide an automated means of analyzing the data but they provide very useful diagnostics and can also help to reveal connections you could easily overlook. However, this is not a text-mining tool - the options on offer are limited in scope.

A multi-user version of the software is also available, which I did not test for this review. This allows researchers to collaborate on a shared set of documents, so they can work on coding and analysis in parallel - another practical way to speed up getting the results to the client.

NVivo in the hands of a researcher

Marleen Morris & Associates is a research and strategic planning consultancy in Vancouver, B.C., that uses NVivo to analyze conventional qualitative research transcripts and also, increasingly, unstructured comments provided through the Web from members of the public invited to engage in consultation processes.

“We work with clients to develop strategic plans,” says Marleen Morris, the firm’s principal. “We will often design a quantitative survey instrument as well as use focus groups, interviews and input received through a purpose-designed Web site in order to develop a body of evidence for planning and decision-making purposes.

“NVivo can make a huge difference for organizations and researchers who want to invite input through a Web site. When you open up a site and invite comments, you quickly receive a very large amount of input. If we were to try to analyze that volume of information without NVivo, we would be facing a long and difficult process. NVivo makes the whole task manageable,” she says.

Asked how NVivo achieves this, Morris explains: “The software makes the process of analyzing this kind of data very easy. As you start to code your data to the different nodes, around the common themes, you quickly start to see the patterns. When there are large volumes of data, you can start to lose track of what you have read. NVivo keeps track of everything in a very efficient way.”

Morris is aware that qualitative researchers can be skeptical of software playing any part in the analysis qualitative data, for the risk this brings in turning a creative activity into a mechanical one that leaves no place for skill or intuition. “I’d say NVivo has helped me be more creative and intuitive as it helps me to see the patterns more easily. Using NVivo means that you do not have to try to hold all the data in your head. If anything, it frees you to become more creative,” she says.

Asked which features she finds particularly useful when analyzing research data, Morris highlights the query functions for the way they help you understand the relationships between the themes in the data. “For example, you can query the data to discover how many times a specific theme was mentioned in concert with other themes. You can also do a word search, which gives you a different perspective on the data.”

She emphasized that people should not be put off by the quantitative aspects of these query reports - in her view, the query function simply helps you explore and understand the data, which in turn facilitates the discovery of meaning. These are just tools that serve to sharpen the qualitative researchers’ intuition.

Morris is also impressed with the software’s capabilities in handling rich media and other sources of data. “If you are working on a project wherein you are asking people to think creatively about the future, you want to allow people to be creative in the way they provide their input as well. NVivo facilitates that.”

Morris notes that NVivo easily deals with video and audio input. “We had one project recently in which young kids did collages for us. We were able to upload photos of these collages and describe them.”
She estimates that a new user familiar with PC office software would be able to master the basics of the NVivo software in just a few hours. “We use it for the qualitative aspects of most of our research and I have not yet run into anything that it hasn’t been suited for. So far, I have not found any limiting factors.”

A role to play

NVivo clearly has a role to play in market research. Not everyone will find it as easy to get started as Marleen Morris - anyone used to working directly from transcripts and other sources on paper may feel the software provides a less direct route to getting to the findings in the initial stages. It is definitely worth persevering. The return on that early investment of time comes when you have complete mastery of all of the data. And all the evidence to back your skilled observations - and your intuition - is only ever a few mouse-clicks away.