In a chart, what is the difference between a colour set once for every mark and a colour bound to a column?
answer
- one of them carries data
- mapping against decoration
- does the picture change with the column?
- a bound channel owns a scale and guide
basics
~20 sA bound colour is a mapping: the column's values decide each mark's colour, and a scale plus a legend let a reader read them back. A fixed colour is decoration — every mark gets it, and it carries no data.
solid answer
~40 sBinding means stating that a **visual channel** — a property of a drawn mark such as position, length, area, colour or marker shape — varies with a column's values. A bound channel comes with a scale, the rule that turns values into colours, and a guide such as a legend that lets a reader run the scale backwards and recover the value. A fixed appearance is the opposite: one colour, size or shape used for every mark, chosen for looks and carrying nothing decodable. The practical test is whether the picture changes when that column changes; if it does not, the colour was never carrying data. The distinction also decides what happens to next month's data: a bound channel extends its scale, a fixed one quietly does not.
go deeper
Recall the two roles a colour can play: varying with a column, or set once for every mark. Be able to look at a chart and say which one it is using, and how you can tell.
Explain what a bound channel brings with it — a scale turning values into colours and a guide that lets a reader read them back — and why a fixed colour brings neither of those.
Show the maintenance consequence. When a new category arrives, a bound channel extends itself, while a hand-set colour and a hand-written legend drift apart and start mislabelling the picture.
Turn it into a review rule the team can apply without running anything: whatever varies across marks must be a stated binding, so a chart's meaning survives new data and a second author.
## A chart is a set of statements, not a picture you decorate A **visual channel** is a property of a drawn mark — position along an axis, length, area, colour, marker shape — that a column's values can be carried by. A **binding** is the statement that a given channel varies with a given column: this column's values are what decides where each mark sits, how long it is, or which colour it takes. The **mark** is the thing actually drawn for each row or each group — a point, a bar, a line, a box, a tile. Read strictly, a chart is a small set of bindings plus the marks that render them. A **fixed appearance** is the other thing you can do to a mark's properties: choose one colour, one size or one shape and use it for every mark. In the finished picture the two look identical — both produce coloured marks — but they are different kinds of statement. A binding says *this varies, and here is what it varies with*. A fixed appearance says *this does not vary at all*. ## What a binding drags in behind it A bound channel is never just a colour. It brings two more things with it: - **A scale** — the rule turning the column's values into positions, lengths or colours. A scale has a set of values it must cover and a set of display values available to cover them with, and it is what makes two marks comparable to each other. - **A guide** — an axis or a legend, the thing that lets a reader run the scale backwards and recover a value from the picture. Without one, a bound channel is a private code. A fixed appearance has neither, because there is nothing to map and nothing to read back. | | colour bound to a column | colour set for every mark | |---|---|---| | Varies across marks | Yes, with the column's values | No, identical everywhere | | Has a scale behind it | Yes — values to colours | No | | Can be read back by a reader | Yes, through the legend | Nothing to read | | Reacts to next month's data | New values get colours | Unchanged | | Carries information | Yes | No — decoration only | ## The ten-second test 1. **Change the column and re-render.** If the picture is identical, that property was never carrying the column. 2. **Look for a guide.** A bound channel normally produces an axis or a legend; a fixed one produces nothing to look at. 3. **Finish the sentence.** If you cannot complete "the darker marks are the ones with more…", the colour is decoration, whatever the author intended. ## Where the legend comes from — and this genuinely differs Designs in this space do not agree about how a legend comes to exist, and that disagreement is why stale legends are a recurring bug rather than a one-off mistake: - Where a channel **has a scale and the scale owns its guide**, the legend is *derived* from the binding. Change the binding and the legend changes with it; it cannot disagree with the picture. - Where colour is a **per-call argument with no scale behind it**, there is nothing to derive a legend from. The legend is a separate object you populate with labels, usually in the order you happened to draw things — and it stays exactly as written when the drawing changes. So "the chart has a legend" does not prove that anything is bound. In the second kind of design a legend is a hand-written caption, and it is exactly as trustworthy as the last person who edited it. ## Why the distinction earns its place in an interview It is the entry point to reading any chart critically: - A chart with **four colours and no guide** is either hiding a binding or wasting one, and a reader cannot tell which. - A chart whose **colours shifted when a new category arrived** was assigning colours in the order values appeared rather than through a stable mapping, so every mark now has a different colour from last month's version and nothing on the page says so. - A chart using **six colours purely for looks** has spent a channel and bought nothing; the next person who needs an encoding has one fewer to reach for. The working rule that follows is cheap to apply and hard to argue with: **anything that varies across marks should be a stated binding, and anything that does not vary should be obviously fixed.** A reviewer can check it without running anything. ## Saying it out loud Name the three parts — the column, the channel, the mark — then say, for each appearance property, whether it is bound or fixed and what a reader is supposed to decode from it. A candidate who can do that in a sentence per chart is doing the thing this subject is about. A candidate who says "I made the bars blue" has not yet noticed there was a decision to make.
- What does a bound channel give you that a fixed appearance cannot, when the chart is regenerated on next month's data?A bound channel is re-derived: new values fall through the same scale and the guide is updated alongside them, so the picture and its legend stay in step. A fixed colour is a constant in the code; new categories inherit whatever the author last wrote, and nothing in the output announces the mismatch.
- Can one chart legitimately mix a bound channel with several fixed ones?Yes, and most charts do. Fixed appearance is the right choice for everything that carries nothing — a common marker size, a reference line's colour, the weight of a border. The rule is directional: whatever varies across marks must be a stated binding, while things that are constant should look constant.
- A colleague says a second colour was added "so the chart is less boring". Why is that worth pushing back on?Because a reader cannot tell decoration from encoding by looking. A second colour reads as a second variable, so the reader hunts for the meaning and either invents one or distrusts the chart. It also consumes a channel that a real binding may need later.
saying these in an interview costs you the question
- Thinks any colour on a chart must be encoding something.
- Says the presence of a legend proves colour is bound to a column.
- Believes one colour used for every mark still carries information.
- Cannot say what a reader is supposed to decode from a colour.
- Assumes a new category always appears in the legend by itself.