Before you start working on a visualisation, there are a few questions you should ask yourself to ensure that you create the most effective visualisation possible.
And above all, make sure your data are accurate and clean before undertaking a visualisation!
Thinking about the intended audience, or what the purpose of your visualisation is, before you start making it will help you to create the most appropriate visualisation for your needs. It can also help you save time when visualising your data, as you'll only include elements that'll be useful for your intended audience. You could be making a visualisation for:
Yourself - Visualising your data can be an important part of data analysis during research, allowing you to discover trends in your data, elucidate relationships, or even plan out future data collection. In making this kind of visualisation, you don't need to be producing a polished end product. Instead, you should develop a workflow that allows you to visualise your data quickly, and in a repeatable manner. Make sure that you document your visualisation workflow to allow you to re-use it in future analyses, and to ensure that you know how you produced a particular visualisation. Documentation may seem dull, but in a few months, you'll be very glad you did it!
Others in your field - Visualisations can be an important part of disseminating your results to researchers in your field, through figures in papers or posters, or as part of conference presentations. These visualisations are showing off your work, and so need to be polished, clear, and well annotated. You should consider employing figure types that are commonly used in your field, as familiarity with the layout will help your audience to quickly grasp your results. Common conventions for visualisations should be adhered to if possible, as not doing so can introduce unnecessary confusion. For example, maps are normally drawn with north to the top of the page; a map that's oriented in a different direction stands a higher chance of being misunderstood by readers.
Public outreach - Visualisations can be an excellent way to engage the interest of non-specialists in the results of a research project. These visualisations should be kept very simple and clear - don't try to display too much competing information on a single visualisation! Clear annotations and descriptions of what's being displayed should be provided, but make sure to avoid using any specialist terminology in your descriptions. It's a good idea to highlight any key points on your visualisation, as a public audience might lack the specialist knowledge to pick out the main messages without prompting.
Accessibility considerations - It's a good idea when creating a visualisation to consider whether your audience might include people with visual impairments, such as colour blindness. To keep your visualisation accessible, ensure that colour isn't the sole way of conveying information by using patterns and labels to help differentiate the information. You can also find colour hues and saturation that will still be differentiable to colour blind vision. Resources like ColorBrewer allow you to come up with colour blind safe (as well as printing in grey-scale safe!) colour palettes.
The kind of visualisation that you choose to create will be largely dependent upon what kind of data you have. Visualisations will each be suited to specific functions, and trying to force your data to display in an unsuitable form will result in an ineffective or misleading visualisation. Think about the function that you want your visualisation to perform. For instance, you might be trying to: show relationships between things (a network diagram might be appropriate), identify correlations (scatterplot), follow a trend through time (line chart), or map a distribution (choropleth map), to name but a few. The Data Visualisation Catalogue and The Data Viz Project are useful resources to help you identify which type of visualisation is most appropriate to your data and the message that you want to communicate.
Some types of visualisations are encountered more frequently than others in everyday life. These visualisations will have an advantage in that most viewers will know how to interpret them without needing any additional instructions. More novel visualisations can be eye-catching and help to engage the interest of your audience, but it's important to keep in mind that they'll also take longer for your audience to understand, and may require an explanation. And always keep the goal of your visualisation in mind and assess whether the visualisation you've chosen is the most effective way of achieving that goal. There are fancy visualisations out there that look really exciting, but a simple bar chart might display the information far more effectively. If your viewers are struggling to figure out what's going on in your visualisation, then it's not doing its job!
The medium that the visualisation will be presented in will influence what kind of visualisation you can choose to create. Presenting information accurately is the driving motivation behind data visualisation, and choosing the right visualisation format can help you avoid presenting a misleading view of your data. An important consideration is whether a visualisation will be static, or if the medium it's presented in allows for animation or interaction.
Visualisations that'll be printed, such as figures in papers, or on posters, will be static. In these cases you might wish to use several different visualisations: one that provides a clear overview of the story that your data is telling, and subsequent secondary visualisations that allow viewers to drill deeper into your data, or explore it from another angle.
Other formats for displaying visualisations, for instance presentations or websites, allow for dynamic visualisations, such as animations or interactive visualisations. Animations are an effective way of displaying data that changes through time, or with some other variable. Websites are particularly suited for presenting interactive visualisations, as they allow viewers to explore aspects of the data that interest them most. Interactive visualisations may also avoid the need to present multiple visualisations of different aspects of the data, as these different views can be incorporated into a single updatable visualisation. Both animations and interactive visualisations can be effective ways to visualise three dimensional data, as they provide the ability to view the visualisation from multiple angles. However, it's important to remember that a static visualisation may still be the most appropriate way to display your data, even when using formats that would allow for dynamic visualisations.
Check out Visualisation Tips for some pointers and guidelines for creating specific visualisations.
Deciding how you want to visualise your data will also depend on the visualisation tools that you have experience in using. Learning a new piece of software, or even a new programming language, to visualise your data can represent a significant time investment, one that might not always be worthwhile.
It's worth assessing whether you'll continue to use the skills and knowledge that you acquire in learning the new tool in the future, in which case time spent learning will be amply paid back by repeated future use. On the other hand, if you're only likely to need the tool for the visualisation you're currently producing, then it's probably not a worthwhile investment. Instead, see if you can adapt some of your existing skills to produce the visualisation you want, or try to select a different kind of visualisation that you already have the knowledge to create to represent your data.
It's also a good idea to consider the time frame that you have to create the visualisation. Even if it might be a good idea in the long-term to acquire some new skills, sometimes there just isn't the necessary time before a deadline!