The Essential Guide to Error-Checking and Reviewing Presentations
Everyone who works in Data Science is aware that data storytelling is a crucial skill. There are thousands of books, articles, and posts about it. It is common to have a section devoted to it in a Data Science course. We are told that when presenting data, we should tell a story using the right graphs, gently guiding our audience.
What we are not told, probably because it is obvious, is that our presentation should be correct! And since nobody talks about this, we are never taught to error-check a presentation. I believe that this is a crucial skill that is overlooked.
Letting errors slip into a presentation can have dire consequences. Failing to make the audience understand the data analysis results is the least severe one. A presentation with mistakes will make the presenter appear sloppy at least, ignorant or deceitful at worst. It will eat away his or her credibility.
Please note that when I say a presentation should be "correct" I am not just referring to using the proper statistical analysis tools, or the appropriate machine learning algorithms. Maybe a better word is that a presentation should be "consistent" with itself. An element in a slide should not contradict an element in another or, worse, in the same slide. Also, note that consistency isn't just about the data itself. It extends to the narrative, design, and even the language one uses throughout the presentation.
One way to classify errors is based on whether they have to do with design, narrative, or logic. Let's examine each category.
Design errors
I have to admit that design errors are covered in various courses. There are tons of literature on what type of graph is best for each situation. For example, a line plot to depict the evolution of a continuous variable in time, or a bar plot for showing the distribution of a categorical variable, etc. Furthermore, there are guiding principles for avoiding 3D graphs, do not clutter, use an appropriate color palette…
When discussing design I would like to emphasize the importance of consistency between slides. It goes without saying that the same color palette should be used for all slides. Similarly, the same font type and, ideally, size should be used for the same elements. This holds especially for key elements like slide titles, section headers, graph titles, and footnotes. Also, key elements should be of the same size and in the same place. For example, a company logo should not be at the top right corner in one slide and at the bottom right corner in the next. Or, the title of each slide should be at the same height. Or, if there is a horizontal line that separates the title from the content, then it should be the same color, same height, same width, and same length. The same applies to text boxes, section titles and footers.
A good test I use is to quickly flicker from one slide to the other and watch whether those elements move or not. Using a template is very useful in maintaining consistency. And yes, I am fully aware that sometimes, consistency has to be broken, but this should not happen often and without a reason.
All those design errors might sound like minor things, but having them gives the impression that the presentation was created in a hurry and sloppily.
Narrative errors
Regarding narrative, there should be a smooth storyline in the presentation. Each slide should have a specific key message that, ideally, is clear from the slide's title. The order of the slides should be carefully chosen to tell the story in the most appropriate way.
One common mistake is to have slides that are out of order, especially if more than one person has worked on the presentation. It's like watching a movie whose scenes change abruptly and/or are out of order. (One might argue that there are such movies that are considered highly artistic. That is true, but in our job usually what we want is to have blockbuster success and not artistic legacy).
Having slides out of order is quite easy to catch. More difficult is to detect issues in the story itself. One common example is emphasizing the great importance of one thing and then not dealing with it or even downplaying its importance by putting emphasis on something else. Or leaving out key elements, for example, a presentation on the sales strategy of a company should have reference to all major products of the company.
Here, I would suggest carefully rehearsing the delivery of the presentation and making certain that the order of the slides is aligned with the story that you want to tell. Having a second opinion will help you create a clear, understandable narrative.
Logical errors
Logical errors range from obvious ones to difficult to spot. A common case of the first type is having wrong titles in sections or graphs. For example, having a bar chart titled "top 5" and displaying six bars. Embarrassing but, almost all of us make silly mistakes from time to time.
Slightly more difficult are errors like having brackets for a continuous variable that do not cover its whole range. I.e we might have binned age to create a graph and have bins that ≤21, (21,45], (50,65] and >65. Or saying in someplace that our customer base is x and in another place providing a different number. You should be extra cautious about this in segmentation projects where the task is to split the customer base into different groups. The size of the groups should add up to the total population.
Similarly, in the financial or retail sector, we usually state that the total income from one group of n customers is an x amount. If at a later point, we state that the average income for this group is y then obviously, y should equal the division of x by n.
Lastly, some errors are domain-specific. For example, having a customer group that is characterized as affluent and consists mainly of young, blue-collar workers is at least strange in most scenarios.
How to spot errors
As discussed above, there are a few things that you can do to find correct errors in a presentation. The number one rule is to carefully go through it many times. And ideally, let some time pass between reviews. Revisiting something with fresh eyes is always helpful. When you make the review, be on the lookout for the errors we discussed above. It might even help to think of it like a game: "What in this presentation is inconsistent or does not make sense?".
Also, rehearse the delivery of the presentation. Devote time to it and make the rehearsal as close to what you like to deliver.
And if you can find another pair of eyes to review the presentation and/or attend the rehearsals then all the better.
Practice
Time for some practice! Let's suppose that you are an executive of the largest Greek bank. And you are reviewing a presentation about the creation of a customer segmentation model for the private/individual customers of the bank. The bank is called Eureka Module Bank (a tribute to my failed start-up that I might write about in another article). Your task is to spot the mistakes or strange things in the presentation. All you need to have is common sense and just know that Greece has a population of around 10 million. Can you spot what is wrong in the slide bellow?
First of all a customer base of 15 million while the country has just 10 million residents is strange. Then, the gender distribution is highly unbalanced. In the table with the age distributions, age of 19 is missing and the percentages do not add up to 100%. Finally, the graph with top 5 occupations has "Agriculture" spelled wrong and, more importantly, has just four occupations.
Editing and reviewing your presentation may seem like a tedious task, but it's a crucial step in ensuring your data story is not just compelling, but also credible. Remember, your goal is to inform and persuade, not confuse or mislead. By prioritizing accuracy, consistency, and clarity, you're not just protecting your reputation – you're also respecting your audience. So, next time you're finalizing a presentation, take the time to double-check everything. It's an investment in your success.
