Lossy and lossless PDF compression

Lossless compression preserves the encoded information; lossy compression removes some information to save space. A PDF workflow can use both, so the label on one image format does not describe every change to the document.

The distinction that matters

The key question is which information the operation changes. Repacking a stream with a lossless encoding is different from reducing image resolution, lowering JPEG quality, or substituting a font. The first can preserve the represented data, while the others can change what a reader sees or how text behaves. An interface that groups all of them under a single compression button should still explain the tradeoffs.

JPEG is often useful for photographic content and smooth changes in tone. Sharp letter edges and thin diagrams can show artifacts at aggressive settings. A scan mixes those categories: its paper texture behaves like a photograph, but its small printed text is often the content you most need to preserve. The right quality choice depends on that content, not on a universal preset name.

Pagestays leaves unsupported images alone and re-encodes eligible JPEG streams only when the new stream is smaller. It keeps the page's text instructions rather than converting every page into a photograph. This narrows the possible savings but also narrows the kind of change you need to inspect.

A worked example

Take an illustrative scan with a 6 MB embedded JPEG. A trial export produces a 4 MB JPEG, a saving of 2 MB. That size comparison tells you nothing about a faint decimal point in the scan. Inspect the same region in both versions. If the decimal point becomes unclear, the smaller output fails the task even though the arithmetic is favorable. Retry from the original at a gentler setting, rather than recompressing the damaged output.

Inspect the detail that carries meaning

Image quality should be judged at the place where information is easiest to lose. A large heading can remain readable even after small account references, decimal points, map labels, or faint handwritten marks have become ambiguous. Pick a difficult region before exporting and return to that same region in the result. This gives you a consistent comparison as you change settings.

Look at both a normal reading view and a closer view. The normal view tells you whether the document is usable at its intended size. The closer view helps distinguish blur, block artifacts, missing strokes, and color changes. Neither view alone is sufficient: extreme zoom can exaggerate harmless pixel structure, while fit-to-page can hide damaged fine print.

Compare against the original, not against a remembered impression. If possible, put both copies side by side and examine the same content. Do not repeatedly recompress the last output while searching for a setting; go back to the original for each trial. Otherwise you are changing both the setting and the source quality, which makes the result harder to interpret.

A good delivery copy is not always the smallest possible file. Choose the smallest version that preserves the required information and works in the receiving system. If no setting achieves both goals, change the workflow: remove genuinely unnecessary pages, use an approved link, or ask the recipient for a different submission route. Destroying detail to satisfy a size field can make the document unusable.

Keep a source and a delivery copy

The original and the deliverable have different jobs. The original is a reference you can return to when a setting turns out to be wrong. The deliverable is the copy prepared for a particular recipient, format, or size constraint. Giving them different names prevents a successful-looking conversion from becoming an accidental overwrite of your only useful source.

Start by opening the original and identifying the features that matter. A plain scan may need only readable pixels, while an application form may need editable fields and a report may need working links. Write the required properties down before choosing a transformation. This makes the final inspection a comparison against a requirement rather than a general impression that the document looks fine.

After saving, close the preview of the edited result and open the downloaded file. That small separation matters: a tool can display an intermediate state that is not identical to the file it exported. Check the filename and destination so you know which copy you are inspecting. If the result is intended for someone else, use the viewer or delivery channel they are likely to use when practical.

Keep useful intermediate copies when a job involves several operations. If a merge is correct but a later image conversion is too soft, restart from the merged PDF rather than reconstructing the whole sequence. Do not treat a smaller byte count, a completed progress bar, or a successful download as proof that all document features survived.

Make the calculation reproducible

A useful measurement has an input, a unit, and a method. Record those before comparing outputs. For a physical page, distinguish millimeters, inches, PDF points, and pixels. For a file, distinguish the stored byte count from an application label that may round or use a different unit convention. Two displays can show different-looking numbers for the same underlying object.

Carry the unrounded value through intermediate calculations and round only the figure you display. Repeated rounding can accumulate an avoidable difference, particularly when dimensions are converted from millimeters to inches and then to page units. The size pages on Pagestays use a sourced base dimension and compute the other units, rather than treating a rounded inch description as exact.

Write assumptions next to the example. If an example uses a chosen image width, a hypothetical attachment size, or a user-selected margin, say that it is an input to the calculation. It is not a measured industry average or a guaranteed result. The reader should be able to substitute their own input and understand what changes.

Finally, verify the artifact itself. A correct formula in a settings panel does not prove that the saved file has the intended page dimensions. Reopen the PDF and inspect its page properties, or check exported image dimensions with an image viewer. When a recipient imposes a limit, use their current rule and leave room for any additional packaging or message content the rule includes.

Preserve context, not only a file extension

An archive needs more than a document that opens today. The file should retain the information necessary to interpret it, and the organization should know where it came from, which version it represents, and what transformations were applied. A readable PDF is useful, but its extension alone does not establish a preservation policy or standards conformance.

Keep the source and record the purpose of any derivative. A small reading copy can be appropriate for sharing while a higher-quality master remains the preservation reference. Do not discard an original scan simply because a compressed output looks similar at normal zoom. Fine annotations, marginal notes, and tonal differences can carry information that becomes important later.

If a standard is required, identify the exact family and conformance target. PDF/A concerns long-term preservation, PDF/UA concerns accessibility, and PDF/X concerns print exchange. They are not interchangeable badges. A conversion should be followed by a validator appropriate to the target, and a later edit can require validation again.

Plan for retrieval as well as storage. Use meaningful filenames, stable descriptions, and controlled locations so someone can find the right document later. Metadata can help, but it should agree with the visible content and organizational records. Pagestays can change ordinary document information and create useful derivatives; it does not certify an archival collection or replace a records-management process.

Put the distinction to work

Open the related local tool when its supported operation matches your task. Read the limits beside the controls, make a separate output, and check it against the requirements described above. For a sequence of PDF operations, the workspace lets you carry a PDF result into the next edit.