When a camera writes a JPEG to the card, it has already made several irreversible decisions on your behalf. It applied a tone curve, performed chroma subsampling on the color channels, ran a DCT-based compression pass, and discarded the remainder. What lands on your storage is the result — a finished interpretation, not the sensor data that preceded it. A RAW file is the opposite: the sensor’s unprocessed photon counts, stored with enough bit depth and metadata that the interpretation happens later, under your control.
That difference is not a minor technicality. In a professional workflow where images move through multiple hands, multiple applications, and multiple export stages, the compression cost compounds in ways that a single-image comparison won’t reveal.
What the Two Formats Actually Store
A typical modern JPEG uses 8 bits per channel, which gives 256 discrete tonal steps per channel and roughly 16.7 million possible colors in total. The DCT compression process divides each channel into 8×8 pixel blocks, converts them to frequency coefficients, and then quantizes those coefficients — rounding them to coarser values — according to a quality table. The rounding is the lossy step. At quality 80, coefficients that carried subtle tonal gradients get binned together, and that binning cannot be reversed. The visible result is the familiar blocking in flat skies and ringing around high-contrast edges.
RAW files, by contrast, typically store 12 or 14 bits of linear data per channel — meaning somewhere between 4,096 and 16,384 discrete tonal steps before any nonlinear tone mapping is applied. The compression applied to RAW files, when any is used, is almost always lossless or near-lossless: it exploits spatial redundancy without discarding frequency information. Sony’s lossless compressed ARW, Nikon’s lossless compressed NEF, and the lossless mode in Adobe’s DNG format all work this way. The sensor data is faithfully preserved; only the file size changes.
The practical consequence is headroom. Recovering two stops of shadow detail in a RAW file means lifting values that were faithfully recorded at 14-bit precision. Doing the same to a JPEG means expanding values that were already quantized, tone-mapped, and compressed — the recovery amplifies whatever compression artifacts survived that process.
Where the Workflow Compounds the Cost
A single JPEG, opened and re-exported at high quality, loses relatively little. The problem in professional workflows is repetition and handoff.
Consider a commercial shoot delivered to a retoucher. If the images arrive as high-quality JPEGs, the retoucher opens each file, makes edits in pixel-editing software (adjustments to exposure, color, cloning), and saves. Each full save that overwrites the JPEG applies the DCT quantization step again. Even at quality 95, the second quantization pass does not reproduce the first pass’s rounding errors identically — it compounds them. The file hasn’t been “re-compressed to quality 95 of the original”; it’s been compressed a second time from a slightly degraded source. Do that across a round-trip of feedback and revision, and the accumulated artifact pattern in smooth gradients becomes measurable.
RAW-based workflows sidestep this by keeping the source data intact and writing edits as metadata or sidecar files. Lightroom’s catalog system, Camera Raw’s XMP sidecar approach, and Capture One’s session structure all operate on the same principle: the RAW file is never modified. The edit instructions are stored separately, and the pixel-level render happens only at export. You can revise white balance three times across three different deliveries and the underlying sensor data has not been touched once.
For workflows involving color grading handoffs — say, a photographer delivering to a video team for a mixed photo/video project — the bit-depth advantage of RAW also matters when images get pulled into a log-color or wide-gamut pipeline. An 8-bit JPEG mapped into a 32-bit floating-point color space produces visible posterization in gradients that a 14-bit source would handle cleanly.
File Size and Storage Economics
The practical objection to RAW is straightforward: the files are substantially larger. A full-frame camera shooting RAW will produce files that are considerably bigger than equivalent JPEGs from the same sensor, and across a multi-day commercial shoot the storage and backup costs are real. Buffer depth on continuous shooting is also affected — most cameras clear a smaller burst of RAWs before the write speed becomes a bottleneck.
None of that is a reason to avoid RAW, but it does mean the workflow has to be designed around it. The standard professional approach is to shoot RAW, cull in-camera or in a fast culling tool using embedded JPEG previews (which the RAW file carries anyway), and then process only selects at full quality. The discarded frames are simply deleted. The net storage cost across a job shrinks considerably when culling happens before archiving rather than after.
DNG is worth mentioning here as a format that addresses some of the storage concerns while preserving the non-destructive advantage — the announcement that DNG became an official RAW standard has implications for long-term archival workflows specifically, since it reduces the risk of a proprietary format becoming unreadable as camera manufacturers revise their RAW specifications.
When JPEG Is the Right Answer
RAW is not always the correct choice, and stating otherwise would be reflexive rather than accurate.
Photojournalism workflows where edited images need to move from camera to wire service in under five minutes are frequently JPEG-based by necessity — the processing pipeline doesn’t accommodate a RAW decode at that speed. Sports photography where buffer depth determines whether you catch the peak moment may similarly favor JPEG, at least at higher frame rates. Event photography with a strong previewing requirement — where a client views images on a tablet during the shoot — often uses JPEG because the embedded previews in RAW files can lag or render inconsistently across software.
The calculus looks like this:
- Shoot RAW when: post-processing latitude matters (complex lighting, uncertain white balance), the image will go through multiple edit passes, the delivery chain includes color-managed handoffs, or long-term archival integrity is a priority
- Shoot JPEG when: speed of delivery is the primary constraint, the in-camera processing is well-calibrated to your specific output need, or buffer depth is a hard requirement at the frame rates you need
- Shoot RAW+JPEG when: you need immediate JPEG previews for client-facing display but want the RAW as insurance — at the cost of roughly doubling the write load and storage requirements
The Export Stage Is Where JPEG Re-Enters
Even a pure RAW workflow ends in a JPEG — or a TIFF, or a WebP — at delivery. The compression cost that RAW avoids during editing is applied at the final export, but only once, from a full-quality source. That single-pass compression, applied deliberately at the output stage, is a different situation from a JPEG that has accumulated artifacts across revisions.
For delivery exports specifically, the compression decisions at that stage matter. Our article on what compression tradeoffs happen when you export for social media covers the second-pass compression that platforms apply after upload — relevant because even a well-prepared final JPEG gets re-encoded on most platforms, and understanding that process helps set appropriate export quality before upload.
The actionable step from all of this: if your current workflow involves delivering edited JPEGs back to a retoucher or through a revision cycle, reconsider the handoff format. Delivering 16-bit TIFFs or DNG derivatives from your RAW source for the revision pass, then exporting to JPEG only at final delivery, eliminates the accumulated quantization problem without requiring the client-facing team to work with proprietary RAW files they may not have software to open.