There is a failure mode I have watched destroy good work more times than I can count. Someone spends three weeks on a genuinely careful analysis, produces a deck with eleven charts on it, presents it to people who make decisions, and nothing happens. The analysis was right. The audience walked out with nothing.
Then a competitor's much worse analysis, presented with one clear number and a plain recommendation, gets funded. Technical people usually diagnose this as the audience being unserious. It is nearly always a communication failure, and this is the book that fixes it. Knaflic's central move is to start before the chart.
Who is the audience, what do they need to decide, and what do you want them to do. That sounds like a platitude until you notice how few analyses are built that way. Most start from the data and work forward, producing whatever the data can support, then present all of it because leaving something out feels dishonest. The reframing here is that including everything is not neutrality, it is abdication.
You did the work, you know what it means, and refusing to say so pushes the interpretation onto people with less context than you. That argument alone justifies the book for a lot of readers. The most immediately useful chapters are on clutter and attention. Every element on a chart competes for a fixed budget of attention, so anything not carrying meaning is actively costing you.
Gridlines, borders, legends that could be direct labels, decorative colour, default axis furniture, three dimensional effects on a two dimensional relationship. The before and after examples are the strongest part of the whole book, because you can see the same data go from unreadable to obvious with nothing removed except noise. I applied this the week I read it and the difference in how people responded to my charts was immediate and slightly embarrassing, in that it revealed how much of my previous output had been quietly unreadable. The material on preattentive attributes is the part that gives you a mechanism rather than a preference.
Size, colour, position and enclosure are processed before conscious attention, which means you can direct a reader's eye deliberately. One element in a saturated colour against a grey field will be seen first, every time, by everyone. This turns chart design from a matter of taste into something closer to engineering, and it is the section I return to most. The writing is clear and short and free of jargon, which matters more than it sounds.
I have given this book to engineers and analysts who do not read books, and they finish it. That is a real property. Now the criticisms, and they are substantial enough to keep this well below the top of my ratings. It is repetitive.
The core content is maybe half the page count, and by the later chapters you are watching the same principles applied to a new example without learning anything new. This is a common shape for business books and it does not become less irritating for being common. Read the first five chapters carefully and skim the rest without guilt. The scope is narrower than the title implies.
This is a book about business communication, specifically about presenting a finding to stakeholders. It has very little for exploratory analysis, where you are looking at data to understand it rather than to explain it, and almost nothing for scientific visualisation where conventions and expectations are entirely different. Someone doing research who applies this advice literally will produce figures that annoy reviewers. The most serious gap is statistical.
There is extensive treatment of charts that mislead through clutter or poor design, and very little on charts that mislead through truncated axes, dubious aggregation, cherry picked windows, or the correlation that survives only because of an outlier. A book on communicating data honestly ought to spend real time on the ways a technically clean chart can still be a lie. It does not, and that is a genuine omission rather than a scope decision. The tool agnosticism cuts both ways.
It keeps the book relevant regardless of what you use, which is why it has lasted, and it means you close it knowing exactly what you want and not how to make your software do it. Anybody who has fought a default charting library to remove its decorations knows this gap is not trivial. It is also from 2015 and dashboards, interactive visualisation and web native charts have grown enormously since. The principles largely transfer.
The absence of any treatment of interaction, filtering or a reader controlling their own view is a real hole in 2026. My 4.2 reflects a book with an unusually high ratio of practical value to page count in its first half, undercut by repetition, a narrow business framing and a missing chapter on statistical honesty. If your work is technically strong and keeps failing to land, this is the cheapest fix available to you.