Skip to content

LearnOperate

Pages per minute

An ingestion speed measure: pages processed divided by the minutes it took, always scoped to a stage, document, run, worker or project.

In DocSlurpGlossary entry

On this page

Pages per minute is ingestion speed: pages processed divided by the minutes it took.

pages per minute = pages_processed / runtime_minutes
time to ingest 100 pages = 100 / pages per minute   (minutes)

A run that processes 240 pages in 6 minutes runs at 40 pages per minute, so 100 pages take 2.5 minutes. “Time to ingest 100 pages” says the same thing in a form people can picture when planning a backlog.

Always name the scope

The same run produces very different numbers depending on what you divide by. The text-extraction stage might handle those 240 pages in 30 seconds (480 pages per minute), while the whole pipeline, with OCR on the scanned pages, embedding and indexing, takes the full 6 minutes. Both figures are correct. Quoting the first as “our speed” misleads anyone planning a backlog.

State the scope (stage, document, run, worker or project), the concurrency (one worker processing four documents at once has roughly four times the pages per minute of the same worker on one), and whether the timing was measured on completed work or estimated from historical averages. Separate cold starts from warm runs: model loading on a fresh machine can dominate a short batch.

Pair it with pages per $1. On a worker billed by the hour and kept busy, pages per $1 = pages per minute × 60 / hourly cost, so speed and cost are two views of the same measurement.

In DocSlurp

In DocSlurp

The worker emits pages-per-minute and time-to-ingest-100-pages per stage, tagged with worker class and warm or cold start (core/worker/src/pipeline/local-compute-cost.ts). Project telemetry computes it from completed runs only (core/server/src/services/project-board-telemetry.ts).

Sources

Jump to