How do you assess the impact of data visualizations in different domains?

Active participation in the initial discussions is crucial to understand the goals and underlying purpose which provides an opportunity to offer your insights and help to define clear objectives, in addition to engaging with business users in the same domain through surveys

Do you want people to like your visual, or do you want them to gain new insight from it? And if you opt for the latter, what if the insight provokes displeasure (e.g. "Our sales went down")? Is that a good outcome? If it leads to corrective actions that will make sales go back up, I would say "yes". But if that is the case, can you really put a "Like" button next to that visual? Because if you do, what is the meaning of "Like"? Does it mean "This visual was great to help me understand what I had to do (to bring sales back up)", or does it mean "I like it when sales go down?" Bottomline: be mindful of your primary goal, and track it carefully.

In situations where defining the goal can be hard this is a tactic I found work in house in public sector. Gaining insight from stakeholders. This should be a given however some times it can be worth while having surveys from independent experts to weigh their opinion on how the organization is being ran. This way when results are released not one person is put at blame for their answers. For example, having a survey circulating on what each member values within the organization does not necessarily mean the truth will be said. But if you allow an outside observer (firm conducting analysis) stakeholders can be more incline to share real goals to achieve a clear objective.

I have always found, thinking in terms of story telling is very helpful. Figuring out what kind of story my audience is anticipating is very crucial when deciding on certain visual(s). Measuring success of a visual is how much effort you put in explaining the visual. The more you struggle to explain your ideology behind the visuals the more you have strayed from what your audience was anticipating. The goal is always to tell a compelling story using the dataset at our disposal while making sure we check as many boxes as we can.

Data visualisation alone rarely tells us exactly "why" something is happening, and is better used to test for and quantify "what" is occurring (or likely to occur). In this regard, data visualisation is often focused on a question. An important first step therefore is not just to identify the needs of your target audience, but to consult with them on their understanding of what insights they are looking to generate, the delivery timelines, what actions are within their remit to deliver, and what specific domain definitions need to be calibrated when collecting the data in order to ensure that we are: 1) Creating visuals that will be understood within our own domain contexts, and 2) Are visualisations likely to be actioned upon/useful.

Qualitative data technically cannot be measured, at least in the traditional sense. At the point you can measure qualitative data, begin re-coding qualitative data to make it measurable, or even look at a larger collection of qualitative data to get an overview, your measurement methodology just turned quantitative. When looking at metrics, there is a clear distinction to make. Both qualitative and quantitative data can feed in, but if the outcome metric is a measurement, what's feeding out is going to be a quantitative metric. A quantitative metric may indicate qualitative aspects of the data, but is still quantitative as a metric.

One thing to keep in mind that we're not living in a perfect world. Always perform basic anomaly checks on the dataset to assure that you're analysis is as valid as possible . Check for missing values, inconsistencies, or unexpected data types within your dataset. You can use descriptive statistics or visualization techniques to identify potential anomalies. Identify outliers, which are data points that significantly deviate from the rest of the dataset. Outliers can distort analysis and model performance.

I would add - you may need a combination of these methods tied together to produce one methodological result. (For instance, visualisation (or, exploratory) analysis is key in the first step in a lot of statistical analysis, because you'll tend to use basic graphs to see what your data is telling you and see whether there are any outliers, likely issues with normality, skew, kurtosis, homoscedacity, etc, before launching into your main statistical testing).

Data dashboarding platforms such as Power BI and Tableau are also good standalone options here. I have found in my field a lot of analysts to just dump as much data on to a dashboard and hope it does something. If this is the kind of data visualisation being constructed, by all means go to powerpoint. However, if you have constructed a data visualisation dashboard in such a targeted way so as to tell a story properly, that story should speak for itself. Data literacy or not - if my target audience cannot understand the overall dashboard I've put together, I've failed as an Analyst to tell the story in a clear and concise way in line with my target audience's capabilities.

Data visualization is a way of trying to help solve or identify the potential problem. In order to to give a meaningful full insights it's imperative to understand the business problem we are trying to solve. There are many ways the Data Visualization (DV) can be defined and every business user can have its own definition, before we implement and make an impacting and insight full visualization we need to come up with the definition of it. Once we have it defined, Need to ensure that the BI tool has the ability to provide the required visualization and ability to slice and dice. With the new age technologies, can use the ability of NLP and api for real time or near real time data for fast and latest updates to the users.