Enormous faxes are inevitable from time to time, but they don’t have to clog your OCR pipeline and keep other critical files waiting. Systems that support parallel processing can handle large faxes alongside routine traffic, provided enough capacity is available.
Below, we’ll briefly cover how today’s fax automation tools manage bottlenecks, then discuss how to plan and test peak capacity for your own implementation.
Virtual OCR “workers” run in parallel to process multiple large faxes at the same time. Short faxes can be captured and routed while a much larger packet is still undergoing OCR, subject to available capacity.
Of course, constraints still exist. SaaS solutions also have finite processing capacity and may impose concurrency limits or usage quotas. Available adjustments depend on the provider and service plan.
Ask your vendor: What capacity constraints should we expect? How quickly can you scale up/down? Are there surcharges for doing so?
In a typical fax-over-IP scenario, each incoming fax occupies a logical channel. More pages generally mean longer use of that channel, although transmission speed and retries also affect call duration. Cloud or hybrid fax solutions offer more flexibility, so they’re extremely helpful when document volume is spiky.
Once faxes are received, OCR converts page images into machine-readable text. Field extraction turns the relevant content into structured data, using techniques that may include predefined zones or AI models. Page count and image quality affect OCR processing time; document layout and extraction requirements add to the downstream workload.
OpenText Fax Aviator supports several tasks—like extracting fields and summarizing documents—each with its own processing demands. Completion time depends on the requested task and the document it’s running against, so it’s worth a specific discussion early in the planning process.
Ask your vendor: Which OCR and AI features run concurrently? Do your processing benchmarks include transmission time?
One 500-page fax might be processed in parallel without issues, but what if two or four arrive?
Locally hosted OCR can be CPU-intensive. A server that comfortably handles peak fax transmission may not suffice for OCR processing. If you’re planning on-prem resources, we strongly recommend sizing against the entire document workload inclusive of downstream processing.
Let’s say a department receives 250 faxes/day at an average of 15 pages/fax, making 3,750 pages altogether. An additional 500-page fax would increase daily page volume by about 13%.
Before opening a conversation with an OCR vendor, start by establishing:
Ask your vendor: How exactly do you calculate capacity needs? How will you incorporate large or challenging documents into the demo/POC?
There’s a huge but subjective variable that we haven’t discussed yet: what exactly do users need out of these large faxes?
One team might require a paragraph-length summary; another might need dozens of extracted fields sent to a downstream system. The latter requires a supported integration, and the overhead, cost, and capacity requirements depend on the workflow.
Juggling these trade-offs with user needs is not a simple task, but that’s where our decades of experience can help. The Paperless Productivity® team has set up large fax processing and OCR solutions for major healthcare organizations, financial firms, public agencies, and more.
To discuss the good, the bad, and the ugly of your current document mix, reach out and speak with a solutions engineer today.
Not necessarily, but it can. Parallel processing lets other faxes proceed as long as some processing capacity remains open. This is less of a concern with cloud fax platforms because they can usually scale far more.
OCR converts fax images into machine-readable text. Field extraction turns relevant content into structured data.
Timing depends on page count, image quality, the requested AI task, and available capacity. Confirm whether quoted times include transmission.
Use peak page volume, document mix, processing tasks, and turnaround targets. Include oversized packets in capacity testing.