Convert B to kB

A byte is eight bits, though that was not always fixed — early machines used six, seven or nine. The eight-bit byte won because it holds one character of text and divides neatly in half.

A kilobyte is 1,000 bytes when a standards body defines it and 1,024 when an operating system does. This page uses 1,000; the kibibyte is the unit that means 1,024.

Going from bytes to kilobytes moves the decimal point 3 places to the left and changes nothing else. There is no factor to remember and no rounding to decide: the digits stay in the same order and only their position changes.

1,234 B is 1.234 kB — same digits, moved along. That is worth knowing because it is the one kind of conversion you can check at a glance: if the digits of the answer are not the digits you started with, something other than the conversion has happened to them.

Below about a megabyte the raw byte count is not yet unreadable, which changes what the conversion is for. Nobody converts 27,504 bytes to kilobytes because they cannot understand the original; they do it because everything they want to compare it against — an asset budget, a limit on a form, another file someone quoted at 30 kB — is written in kilobytes. The conversion is a translation into the unit the conversation is happening in.

That also means precision is cheap here. One decimal place carries a hundred bytes and two carry ten, so 27.5 kB and 27.50 kB are both honest and the second is not padding. Higher up the ladder the extra digits stop meaning anything, but at this scale a tenth of a kilobyte is a real quantity somebody could remove from a file.

The kilobyte here is 1,000 bytes, so the conversion is three digits moved off the right-hand end. That is the SI meaning of the prefix, the meaning used by drive manufacturers, by network equipment and by every browser performance tool, and it is the only one under which a kilobyte relates to a megabyte and a gigabyte by the same factor each time.

The 1,024-byte unit is the kibibyte, and at this rung of the ladder the gap between them is at its smallest: 2.4%. That is small enough to ignore for a single icon and large enough to matter across a bundle, where a 900 kB budget and a 900 KiB budget differ by about 22 kB — roughly a whole web font. The gap widens at every step above this one, reaching 7.4% at the gigabyte, but here it is the difference between two numbers that look the same.

Filesystems do not hand out storage a byte at a time. They allocate blocks, typically 4,096 bytes, and a file occupies whole blocks whether or not it fills them. A 300-byte file has a length of 0.3 kB and takes 4 kB of disk; a 4,100-byte file takes 8 kB. The wasted remainder has a name, slack space, and on a directory of small assets it can easily double the total.

This is the one place where the byte count and the space are genuinely different quantities rather than two renderings of one number. Which one you convert depends on the question: an upload limit and a transfer budget care about the length, while free space on a device cares about the allocation. Tools split along the same line — a file listing reports length, a disk-usage tool reports blocks — and the two will not agree at this scale.

Front-end budgets are written in kilobytes because that is the unit at which the numbers stay comparable. A stylesheet at 18 kB, a script at 140 kB and a font at 24 kB add up to a figure somebody can argue about; the same three quoted as 18,432, 143,360 and 24,576 bytes do not invite the same conversation. The conversion is what turns three build outputs into a decision.

The number that belongs in the budget is usually the compressed one. Text assets are served under gzip or brotli, and a 140 kB script may cross the network as 45 kB, so a build tool reporting only the raw byte count is reporting the wrong figure by a factor of three. When a size is quoted without saying which it is, the safe reading is that a raw byte count was divided by 1,000 and nothing was compressed.

Data crosses an ordinary network in packets whose payload is capped near 1,500 bytes, so anything measured in kilobytes is already a small pile of packets: 27.5 kB is about nineteen of them. The practical consequence is that saving bytes has a step function in it. Taking a file from 1,600 bytes to 1,400 removes a packet and a possible round trip; taking 40 kB to 39 kB removes almost nothing a user could perceive.

It also explains why a browser reports a transferred size slightly above the file size even when nothing was compressed. Each packet carries protocol headers, and a connection carries its own setup before any of the file arrives. Converting the file to kilobytes gives the content; the wire cost is a few per cent above it, and for very small files it can be most of the total.

A kilobyte of plain English is about a thousand characters, or a paragraph and a half — one byte per character in ASCII and, in practice, in UTF-8 as well for Latin text. Outside that alphabet UTF-8 spends two bytes on Greek and Cyrillic, three on most Chinese, Japanese and Devanagari characters, and four on emoji, so the same visible page of text can be three times the size in one language and not another.

Structured data is heavier per unit of meaning than prose. A JSON response with descriptive keys spends a large share of its bytes repeating those keys on every record, which is why a few hundred rows so easily reaches tens of kilobytes and why compression works so well on it. Binary formats invert this: a small PNG is mostly compressed pixel data and will not shrink again in transit.

Dividing by 1,000 and cutting to one decimal understates a file slightly more often than it overstates it, because sizes quoted in public are usually rounded down to look better. Where the figure is going into a budget or a limit, round up: a form that accepts 100 kB and a file of 99,987 bytes leave no margin at all if anything is added later, and the byte count is the number the check will actually use.

Where the figure is going into prose, one decimal is enough and two is usually false comfort, because the file will change with the next build. The exception is anything near a hard boundary — a limit, a cache threshold, a single packet — where the exact byte count is the only number that answers the question and the kilobyte figure is a summary of it.