What dpi does to a receipt scan

Dots per inch is the most quoted number on a scanner and the least useful one taken alone. It describes how finely the device samples the page, which matters until you pass the point where the print itself has no more detail to give, after which you are only making bigger files. For receipts the useful range is narrower than the range on offer, and knowing why lets you stop worrying about the number.

Optical resolution and the other kind

There are two dpi figures and only one of them is real. The optical resolution is what the sensor and the transport can actually resolve: how many separate samples the hardware takes across the width of the page, and how finely it steps along its length. The interpolated figure is what the software will produce by inventing intermediate values between real samples.

Interpolation does not add information. It makes a larger image out of the same measurements, with smooth-looking edges that were guessed rather than observed. It is not fraud exactly — a bigger image is sometimes genuinely more convenient — but a figure reached by interpolation should be read as a file-size setting, not a quality setting. When a specification lists two numbers, the smaller one is the honest one.

The transport imposes its own limit, particularly on a feeder. Resolution along the direction of travel depends on the paper moving at a controlled speed past the sensor, so the mechanical quality discussed in how a document feeder works sets a ceiling that no setting can raise.

What the print on a receipt actually needs

Till print is not fine engraving. It is produced by a thermal head or a low-resolution impact mechanism, in a typeface chosen to be legible at arm’s length, and the characters are made of visible dots. Sample it finely enough to distinguish those dots reliably and you have everything the paper contains. Sample it more finely and you are recording the texture of the paper fibres.

This has a practical consequence people find counter-intuitive: going higher can make a faint receipt look worse. At a coarse setting, several paper fibres and several print dots average into one sample, and the averaging tends to smooth out noise. At a very fine setting, you resolve the speckle in the coating, the fibre texture, and every crease shadow individually — and a stretched, low-contrast character sitting in a field of resolved noise is harder to read, not easier, both for you and for anything trying to interpret it automatically.

The setting that helps a faint slip is almost never more resolution. It is the colour mode and the exposure, which is a different control entirely: see scanning faded thermal paper.

Resolution multiplies everything downstream

Doubling the dpi roughly quadruples the number of samples, because it applies in both directions. Everything downstream feels that.

File size grows accordingly, and then gets partly reclaimed by compression — which behaves differently at different resolutions, as compression and scan quality explains.
Scan speed falls, because the transport has to move more slowly to sample more finely along the page.
Long items get shorter, because the device’s buffer fills faster. This is the trade behind most long-paper restrictions, covered in scanning very long receipts.
Text recognition does not improve past the point where characters are cleanly resolved, and can get worse when noise is resolved along with them.

So the cost of a high setting is paid on every scan, in throughput and storage, and the benefit stops arriving well before the maximum.

Finding the right setting for your paper

The only reliable method is to test on your own worst receipt rather than reason about it. Take the faintest, most crumpled slip you have. Scan it three times: at the device’s lowest sensible setting, at whatever the default is, and at something noticeably higher. Then look at all three at the same on-screen size and ask a specific question — can you read the total, the date, and the vendor name in each?

Usually the default and the high setting are indistinguishable for that purpose, and the lowest is where something starts to go. That tells you the default is fine and you can stop thinking about it. Occasionally the high setting genuinely helps, which is worth knowing too. Either way you have measured it on the paper you actually own instead of adopting somebody else’s recommendation for paper you have never seen. The broader version of this method is in testing a scanner with your own paper.

The one case for scanning higher than you need

There is a reasonable argument for capturing above your working requirement when the item is irreplaceable and about to be destroyed or is actively deteriorating. You cannot go back and rescan a slip that has faded to blank or gone in the bin, and storage is cheap relative to the value of an unrepeatable record. In that specific case, capture generously.

For the recurring stream of ordinary slips, that logic does not apply, because you can always rescan tomorrow’s coffee receipt if it matters, and it will not. Reserve the expensive setting for the items that are genuinely one-shot.

What the number does not tell you

Two devices at the same dpi can produce visibly different images. The sensor’s ability to distinguish adjacent tones, the quality and evenness of the illumination, the accuracy of the transport, and the image processing applied before the file is written all vary independently of the resolution figure. This is why a specification comparison between two scanners on dpi alone tells you close to nothing, and why the only comparison worth trusting is your own paper through both devices.

And resolution says nothing at all about whether the resulting file is an acceptable record. Whether a scan can stand in for a receipt, at what fidelity, and for how long the original must survive alongside it, are questions set by rules that vary by country and by entity type. Take them to your own tax authority or an accountant in your jurisdiction rather than to a specification sheet.