By Jetformat
PDF preview sampling is the selection of some rendered pages for visual review instead of rendering every page. It changes the number of preview operations; it does not provide a statistical assurance that unseen pages are correct. State the selection policy and its limits before using the estimate.
State the assumptions first
This example batch contains 100 PDFs with 20 pages each. Conversion has already happened. The calculations below cover only PDF rendering at one credit per rendered page; they are not measured throughput or quality results.
| Preview policy | Calculation | Images | Render credits |
|---|---|---|---|
| First page only | 100 × 1 | 100 | 100 |
| First, middle, last | 100 × 3 | 300 | 300 |
| Every page | 100 × 20 | 2,000 | 2,000 |
Choose pages by risk, not only position
A title page may be the least demanding page in a document. Add pages containing dense tables, unusual fonts, charts, or section transitions. A fixed “first, middle, last” policy is a starting example, not a guarantee of coverage.
For a 20-page document, a three-page selection can be rendered with:
jetformat pdf render report.pdf --pages 1,10,last --width 1600 -o preview
Check the page count first. A page selector that fits this example may not fit a shorter document in the same batch.
Expand review when a defect appears
If a sample reveals clipping or font substitution, inspect other pages sharing the affected structure. After a layout fix, rerender changed pages. Review all pages when the cost of an unseen defect requires full visual coverage.
The render guide explains selectors, and PDF inspection supplies the page counts for your own estimate. Use the credit estimate article when the workflow also includes conversion and editing.
Apply the policy to actual documents
- Inspect each PDF’s page count instead of assuming every file matches the sample workload.
- Identify pages containing the content most likely to expose layout problems.
- Build a valid selector for that document and count the pages it will render.
- Open the resulting images and record the pages actually reviewed.
- Expand the selection when a defect suggests that other pages share the problem.
A one-page document cannot have three distinct first, middle, and last pages. Deduplicate the review positions in your planning rather than counting the same page three times. When page counts vary, calculate usage from each file’s actual selection instead of multiplying a fixed count by the number of documents.
Compare sampling with full review
The three-page policy above renders 300 images rather than 2,000, a difference of 1,700 rendered pages and therefore 1,700 rendering credits under the stated assumptions. This is arithmetic, not evidence that the smaller review finds the same defects. No detection rate or error probability is claimed.
First-page-only review can identify title or branding issues while missing later tables and appendices. Full-page review provides visual coverage of every page, but it still requires someone or a suitable review process to inspect the images. Generating them alone is not verification.
Record what the estimate excludes
This article counts rendering after conversion. Add conversion, extraction, redaction, or Office edits separately if the actual workflow performs them. Also keep image dimensions distinct from page counts: the example width of 1,600 pixels describes a preview, while rendering credits follow pages. Use document credit estimates for a complete workflow calculation.