Original video;

Sequential palettes shouldn't be used with non-sequential categories such as "Canada, Mexico, USA". This makes it look like the categories with higher-intensity colors (Canada, in this case) have a higher quantity of "something" when, in fact, higher or lower color intensity doesn’t mean anything in this chart. A categorical (“rainbow”) palette should have been used here.

This color palette isn't balanced, i.e., some colors draw far more attention than others (in this case, Carbs draws far more attention than Proteins). This would only be an effective palette if one deliberately wanted to draw attention to Carbs. Also, because a sequential palette was used for non-sequential categories, it looks like Carbs have a higher “something” than Fat, which has a higher “something” than Protein, which is not the case .

On a dashboard, bigger/darker red shapes should indicate bigger, badder problems. At first, "Sleep" looks more problematic than "Calories Burned" because it has a larger red shape but, in fact, the opposite is true. I was also misled at first because I thought that the colored areas were ranges of “good” or poor” and the black lines were the current values since this is the way that most gauges work (e.g., car tachometers and speedometers).

It looks like the main message of this chart is that Cuba is a problem and Guatemala is “caution”, but this is presumably not the message of this chart. Also, because the same hue (blue) is used for both North and South America, it looks like these categories are “grouped” as being distinct from the others, but no such grouping exists (presumably). It also makes it look like Canada is being highlighted among the other “blue” countries, but it’s not.

The original "countries by population" chart was, in fact, colorblind-safe for all major forms of colorblindness in the first place, and so didn’t need to be made safe for colorblind users. The "safe" version in the video (right-hand chart above), on the other hand, is actually NOT colorblind safe for red-blind protanopia (one of the most common forms of colorblindness).

This isn’t how to decide when to use a diverging color palette. A diverging color palette should be used when there’s a dividing point in the data and when readers feel categorically different about values that fall on one side of the dividing point or the other (profitable vs. unprofitable, above average vs. below average, etc.). This also isn’t how to decide when to use a sequential palette. A sequential palette should be used for categories that represent different quantities of something (risk, profitability, population, etc.) or for quantitative values (sales, customer satisfaction, etc.)

It took me at least 30 seconds to correctly figure out what the green shading in this table meant. The obvious meaning is that some values in the right-hand column were “a little bit good” and other values were “very good”. As it turns out, that’s not what the green shading means at all (I assume), and the purpose of the green shading is, in fact, to highlight the entire right-hand column. Green and red should only be used to highlight “good” or “bad” data (at least, in Western cultures), not for general highlighting. Having several intensities of green REALLY makes it look like the green is being used to highlight “good” data. There are several ways that one could use to highlight a column in this table that would be much less confusing (extra spacing, enclosure, a vertical border, etc.)

A minor issue, but worth mentioning in a video about effective use of color: Since the “Last 7 days” values and the “Previous days” values are part of the same data series, they, should have the same hue (blue, gray, purple, etc.). The “Last 7 days” values could then be highlighted by making them a higher intensity (“darker”) shade of the same hue. The fact that different hues are used in the original chart makes it look like the “Previous days” and “Last 7 days” values are categorically different sets of values, but they’re not (they’re all part of the same series).