Building Mental Model Diagrams
How to collaboratively make mental model diagrams in your organisation
In our previous article we introduced mental model diagrams, some of the theory behind them, and the kind of value we believe they offer organisations. In this more hands-on article, we’ll take you through our process of collaboratively building a comprehensive mental model diagram.
Mental model diagrams are a great way to capture how the people we design for experience a specific area. As an approach, they allow us to clarify people’s intents, their feelings, and what attitudes they bring to the table as they move towards their goals.
Building a mental model diagram is an iterative process that is composed of continuous rounds of research, analysis, synthesis and refinement. We can break down this process into 6 discrete steps (Fig. 1).
Step 1 — Scope the study
As with any research study, start by defining your scope. To do this, make sure you have a clear research goal and questions. If you have existing insights, then you might be able to leverage those to kickstart things. If you don’t have any research, but have strong assumptions about people and what they’re trying to achieve, use these to give your research an initial direction.
A study to build a mental model diagram is exploratory in nature. You want to be in the problem space and study people, not observe users of your system. Consider these 3 factors before running a study:
- The breadth of knowledge that you’re aiming for.
- Your budget and time restrictions.
- Your level of comfort doing qualitative research, especially exploratory.
For example, SEEK is a marketplace that connects job seekers and hirers. By building a mental model diagram, our goal was to create a unified view of job seekers across different countries that we could scale with ease. To help us get there, we guided our study with the following high-level research question:
What is it like for job seekers who are looking to improve their professional working lives?
The purpose was to get insights that had enough breadth and depth. This included understanding people’s motivations, how they go about the job-seeking process, what they value as they assess jobs, and how they feel throughout the overall experience, including when their application is rejected.
We knew that this would take more than just a handful of interviews, so we crafted a plan to build our insights cumulatively and across different points of research.
Step 2 — Do the research
Typically for a mental model diagram you’ll run 1-on-1 interviews to get the data. To find people to talk with, recruit them as you would normally do. For example, at SEEK we commonly use a recruitment company to source participants based on certain criteria.
To improve the richness of your insights, get a good mix of participants across different segments. Aim for breadth and not just depth². For example, you might want to get a good balance between people aggressively looking for a job, versus people waiting for the right opportunity to come along.
When building our own mental model diagram, we repurposed data from other projects to get it off the ground. For example, our research team already had learnings about job seekers for specialised roles that we leveraged. We also tapped into research other teams were running as part of their Continuous Discovery process. Just make sure that any research you use or tap into explores the problem space and doesn’t focus solely on evaluation of solutions.
Create script templates
When working with other teams, we found that using script templates was helpful. This helped different people and teams run their sessions and ask questions to participants consistently. For example, we had 1 template for teams doing research purely in our problem space. For teams doing other research, we had other templates they could use to slot either 20 or 40 minutes of our problem space exploration into their script.
In order to explore areas of interest effectively, use open-ended questions and make sure that you’re engaged. Indi Young uses the concept of active listening. It emphasises that as a researcher you want to be in the moment, carefully listening to what people are saying and minimising your biases.
Use either unstructured or semi-structured interviews. With unstructured interviews, you get added flexibility, because as a facilitator you’re using broad “fire-starter” questions that might lead you down unexpected but useful pathways.
However, make sure that you’re listening and don’t become overly attached to a script. Be flexible.
For our study however, we used semi-structured interviews. We had general areas we wanted to explore, and defined a structure around those. This worked well because we already had a solid understanding of the domain. Additionally, this helped us get added consistency across researchers and teams. However, make sure that you’re listening and don’t become overly attached to a script. Be flexible. Explore areas of interest that aren’t in the script if the situation calls for it.
Facilitating the sessions
Make sure that facilitators are skilled at interviewing people. At SEEK our designers and researchers predominately facilitate the research. However as part of the Continuous Discovery process, we also had product managers facilitating. Our UX research team was able to help out by providing them with training. This made people more at ease and empowered them to be actively engaged with the research process.
Additionally, our script templates helped less experienced facilitators not worry too much about what to ask, because they had something to fall back on.
Creating the transcripts
The best way for you to analyse the data is through accurate transcripts. Creating transcripts is time consuming, so we recommend using an automated tool or a third-party service or company.
We used a tool called Otter.ai, but there are other options such as Google’s recorder tool present in some Android devices. Sometimes you might also want to use a third party for transcriptions across languages.
If you’re using an automated tool, it’s likely that you’ll need to clean your transcript. To improve the accuracy, make sure you get a good microphone (e.g., directional) and that it’s well positioned. Avoid interrupting people when they’re talking, such as minimising “Mmm, mmm” affirmations. Seeing how many times you’re mentioned in a transcript really brings to the forefront how you ask questions, and how effectively you’re communicating with participants.
If you can, ask others to help out with reviewing and cleaning transcripts. We found that this significantly increased how quickly we were able to move into the analysis stage.
As a facilitator, it’s fine to note down some comments, but keep your focus on what people are saying. Let someone else do the note-taking, because it helps them be actively involved and adds another potential layer to your insights. At SEEK, we usually have different people helping out with note-taking and to support them, our research team offers training with some hands-on practice sessions.
Step 3 — Summarise the data
After you collect the data and have your transcripts, it’s time to start extracting insights. The first thing you want to do is to summarise your data, as seen in Fig. 2. These summaries are useful for 2 reasons:
- They help you create descriptions close to what people said.
- They help you familiarise and empathise with the content and reflect on whether you should include it in your model or not.
You want your summaries to be close to the quotes. Use a clear, simple format that fosters empathy with whoever reads it (see Fig. 3). Consider the following guidelines:
- Start the summary with a verb in the present tense, as this grounds you closer to what the person is conveying through the quote and promotes empathy¹.
- Extract the key insights from the quote using a verb + key point format. Focus on a single concept as conveyed by the person. If you find yourself using “and” conjunctions, it’s probably a sign that you need to break down a quote into 2 or more pieces.
- Avoid inferring too much meaning or overly synthesising the data. That’s better left to the next stages, when you start clustering summaries together. Going up the hierarchy too soon leads you to lose detail in favour of abstraction and simplicity.
Summarising quotes was also a great way for others to get close to the data and build empathy for the people we wanted to learn about. So we involved fellow researchers, designers and product managers in the process of crafting these summaries.
Including others at this stage can also help you get added buy-in. As people get more involved and understand the people you’re studying, they become more aware of the benefits of the process and are more likely to champion it. Here’s what we did:
- The first time we asked people to take part, we ran a workshop where we explained the process (see Fig. 1).
- We prepared the transcripts up front and put them onto a shared Google Spreadsheet that everyone could access.
- We paired and assigned people to a transcript. This way they had someone to brainstorm and work with. It also meant we could work on multiple summaries at once.
- We shared examples of good and bad summaries. These were crucial so people had something to refer back to.
- After the team completed the summaries, we reviewed them to check for consistency and that they were composed correctly.
After you have the team contributing to the summaries — and you’ve reviewed the content — you’re now ready to start shaping the hierarchy of the mental model diagram.
Step 4 — Synthesise insights
When synthesising, it’s a good idea to first look back at your research goal and questions. Use this information to frame the mental model diagram and choose a title. At SEEK the title for our mental model diagram was Improving my professional working life. Having a clear title helps you keep focused on what the context of your work is.
Synthesis is mostly a bottom-up process. Your summaries are grouped together through affinity to create the basic building blocks of the problem space. If your study is focused or small, your summaries are your boxes (i.e., the lower level of the hierarchy).
For more complex studies it’s a good idea to first group together summaries into boxes, like those in Fig. 4. Doing this helps you avoid overwhelming models, especially when you move into the visualisation stage. After you get a set of boxes, you’ll then group these together in order to create towers. Similarly, you’ll group towers so as to create mental spaces.
In order to make this process systematic, we recommend the following:
- Make sure you’re using a tool that allows you to easily move around your quotes and summaries. This tool should also allow you to cross-reference other content, such as participant details. In our case, we used a spreadsheet.
- Create the basic structure for your hierarchy so you have a consistent way of adding content. If you’re using a spreadsheet, it will look something like Fig. 4.
- Create unique references in your mental model diagram. For example Fig. 4 shows how we used codes like G, G01 and G01A to reference mental spaces, towers and boxes, respectively. This is useful when you have your whole diagram and want to connect between the problem and solution spaces in your diagram. It’s also useful when you connect with other frameworks, such as with Jobs-to-be-Done, or when you want to refer back to the raw data.
As your mental model diagram starts to take shape, you’ll get a better understanding of the problem space. One of the key advantages of using mental model diagrams is that it makes adding new content easy. You’ll find that as your model grows, you’ll move boxes between towers and move and create new towers. There’ll be a point where new data fits nicely into the structure you’ve built.
Step 5 — Assess the results
After you have a structure and insights in place, you should pause and reflect. If you find you have gaps in your knowledge after the initial sessions, consider running additional research and iterating through the process again.
Identifying gaps in your mental model diagram is crucial, because it helps you decide whether you should visualise the model or not. The main reason to wait is that you’ll lose flexibility when you transition out of a tool such as Google Spreadsheets or Excel. In order to help you decide, think about the following:
- Do your insights give you enough clarity into the problem space?
- Do you have boxes based on only 1 or 2 summaries? This could be a sign that you need additional data to add to their credibility.
- Are there any gaps in areas that the business wants to know more about?
- Do your insights have enough breadth as well as depth? For example, you might want to talk to people who are looking for jobs in different countries, or with people who struggle to land a job no matter what they do.
- Does the organisation support additional research? You might first need to show how you’re using existing insights in order to get buy-in for further research.
Step 6 — Visualise the model
The last step is to visualise your mental model diagram. Having a visual representation makes it easier for others to understand and relate to the insights. When those insights are in a spreadsheet, it can be hard to see patterns emerge.
The diagram itself is simple to build and is made up of basic shapes to represent the boxes and towers. Fig. 5 is a slice of the actual diagram with a zoomed-in section that shows how the shapes came together to create a visual representation of our research.
To create your diagram you can use a tool such as Sketch, but there are some limitations. For example, your diagram will quickly become disconnected from your data source, meaning that any changes to the original data won’t be reflected in your diagram.
To avoid having to continually update your diagram to reflect the underlying data, make sure your insights are refined enough before you move into your visualisation. You can also try to automatically generate the diagram from your data. Indi Young offers a free-to-use auto generator tool that you can leverage.
In order for you to communicate different aspects of your insights, you can decorate your diagram. Indeed, a key advantage of using mental model diagrams is that they allow you to customise them to suit the context you’re working in. Here are a few examples for inspiration.
Fig. 6 shows a diagram where the authors added coloured bars inside each of the boxes. These colours correspond to their user segments, meaning that some boxes are only relevant to some of the segments.
In the last example, Fig. 7 the authors used the outline weight of the box to denote an insight. In this case a heavy outline shows that the insight came from more than 4 people, and a lighter outline shows the research came from less.
Creating a print out
We decided to print out our mental model diagram. It was big. We used a local printing shop and chose something that could be drawn on (some paper is shiny so isn’t good for that). We also put the diagram where it was visible to everyone, encouraging people to add to it. We saw that as a good sign if people wanted to engage with it.
Printing isn’t cheap so we used three B1 size images on light weight paper and stuck them together. Our mental model diagram ended up being about 2m long and 1m high (80x40 inches).
We also added the images to a Miro board. We found this was especially good for those working remotely. The nature of the tool also made it easy to zoom-in and out. This ability to move between bird’s eye view and detailed perspectives made it easier to explore and identify patterns within the diagram more easily.
Learnings & Conclusion
Building a solid mental model diagram does take work, but we believe it’s worthwhile given the level of depth you gain on how people experience the domain that you and your organisation are working in.
Our 6-step process offered us a solid foundation to move from research into a detailed visualisation in a systematic way. We believe that our success was based on 3 major principles: planning, collaboration, and efficiency.
- Planning — Make sure you get the scope of your study right and that you’re exploring the problem space. Reuse existing knowledge to help you kickstart your mental model diagram. Create a plan to gather data from different points in time and adapt to your circumstances. Create script templates if working across teams to ensure you’re asking the right questions consistently.
- Collaboration — If you can, work together with others. This helps everyone foster empathy for the people you’re learning about, as well as helping others see the benefits of the process. Working collaboratively on the analysis of your insights is also a good way to check the validity of your insights by being transparent on how your interpretations ladder up from the raw data. Make sure however that you’re giving your colleagues the tools and training they need. Always strive to ensure that your insights are of high quality. It’s fine to only collaborate with others at certain points of the process rather than at every step.
- Efficiency — Be focused when interviewing people and let others do the note-taking. Use automated tools for transcripts, and ask others to help correct them to speed up the process. Make sure that your insights are refined and strong enough before you visualise your diagram. Use references in the problem space so you can connect with other content and can relate back to the raw data that support them.
In our next article, we’ll get into more detail on how we’ve used the insights from the problem space to help us guide product design and strategy. Additionally, we’ll go into detail on how we’ve built and decorated the solution space. Finally, we’ll show how we’ve bridged our mental model diagram with frameworks such as the Jobs-to-be-Done.
Thanks for reading.
If you enjoyed this article, check out our previous one where we talk about the basics of mental model diagrams.
- Gair, S. (2012). Feeling their stories: Contemplating empathy, insider/outsider positionings, and enriching qualitative research. Qualitative health research, 22(1):134–143.
- Russell, C. K. and Gregory, D. M. (2003). Evaluation of qualitative research studies. Evidence-Based Nursing, 6(2):36–40.