OCR for Records Retrieval Companies: A Buyer's Checklist
By Schema LawPublished Updated
OCR for records retrieval companies is worth adding when your law firm clients would otherwise convert every set themselves. Deliver a searchable PDF beside the untouched original, run conversion after your page-count check and before delivery, and choose a tool whose written terms your contracts with those firms allow you to use.
A retrieval company sells completeness and speed: the right records, all of them, from the right custodian, by the date promised. Searchable text is a third thing to sell. This post covers the decision to offer it, where it sits in your process, and a checklist for choosing the tool.
Your buyer is usually an attorney or a paralegal who needs medical records before a deadline. Ask how they want the set labeled for the matter, and confirm the duty of confidentiality in your service agreement covers the vendor you use to add text. Keep the custodian's pages in the order you received them, and note any records retention date the firm sets for your copies.
What do law firms expect from a records retrieval company's scans?
Law firms expect a records retrieval company's scans to be complete, in the custodian's order, legible, and labeled so the set can be matched to the request. A firm that works from the records also needs to search them. The first four are your core service. The fifth is where OCR comes in.
A delivered set can fail the firm in five ways:
- Missing pages. The custodian's count and the delivered count differ and nobody said so.
- Pages out of order or rotated. A chart that reads sideways or jumps between years.
- Unreadable copies. A fax of a photocopy, or a scan made at a resolution too low to read small print.
- One giant file with no map. Billing, chart notes and imaging reports run together with no bookmarks or index.
- No text. The firm opens an 800-page set, presses Ctrl+F and finds nothing.
What "searchable" is worth to the firm
The person who opens your delivery is building a chronology or looking for one provider's name. With image-only pages, that means reading every page. The guide to OCR on medical records for attorneys describes what the firm does with a set once it has text, and it is a fair picture of what your client is trying to get done.
What firms do not expect
Firms do not expect you to vouch for the text. They do not expect corrected or retyped records, and they should not receive them. Your deliverable is the custodian's record as produced. A text layer is an aid to finding things in it, and your delivery note should say exactly that.
Should a records retrieval company deliver searchable PDFs by default?
Yes, for most retrieval companies a searchable PDF should be the default deliverable, sent together with the original file as received. A searchable PDF keeps each page image as the custodian produced it and adds hidden text behind it, so the firm gains search without losing the record.
Why the searchable PDF and not Word
Records are evidence of what a custodian holds. A Word file rebuilt from a scan is a new document with new line breaks, and it invites edits. A searchable PDF still looks like the record. If a firm wants a billing ledger as a spreadsheet, treat that as a separate request for that document. The site's export formats section shows what each format contains.
Why keep the original as well
Send both, clearly named. The original is what the custodian certified or produced. If a question ever arises about what was received, the untouched file answers it. A naming pair such as Ruiz - Lakeside Ortho - records (as received).pdf and Ruiz - Lakeside Ortho - records (searchable).pdf keeps them together.
When not to make it the default
| Situation | Better default |
|---|---|
| The firm's engagement terms bar sending records to any outside service | Original only, unless the firm approves the tool in writing |
| The set is mostly handwriting or film printouts | Original only, with a note that text conversion would add little |
| The firm loads records into a review platform that needs its own format | Ask what it needs before converting |
| The custodian already supplied a text-based PDF | Deliver as received; there is nothing to add |
Check that last row on every order. A custodian's electronic export can arrive with text already in it. Try to select a word on a page from the middle of the file before you spend pages on it.
Say what the text layer is
Put two sentences in every delivery note: the searchable copy was machine-converted and has not been proofread, and the original scan controls. That protects the firm from over-relying on a search, and it protects you from being asked why "Hendricks" was not found on a page where the fax blurred the name.
Fitting OCR between records receipt and delivery to the firm
OCR fits after you have checked the received records for completeness and before you assemble the delivery. Converting earlier wastes pages on sets you will have to re-request. Converting later means the delivery goes out before anyone has looked at the result.
A sequence that works:
- Receive and log. Record the custodian, date, method and page count as received.
- Check completeness. Compare against the request: date range, record types, the custodian's own count or certification. Re-request now if something is missing.
- Fix what a person can fix. Rotate sideways pages and remove blank separator sheets, if your client agreements allow it. Never reorder a custodian's pages without noting it.
- Split if needed. Tools cap file size and pages per document. Schema Law, for example, takes PDF, PNG or JPEG files up to 20 MB, and its plans cap pages per document at 250 on Solo and 1,000 on Team. Split a large chart at natural breaks such as record type or year.
- Convert. Upload, wait for the pages to be read, and export the searchable PDF.
- Spot-check. Confirm the page count matches, then search for the patient's surname and one provider name. Open three pages from different parts of the set and select a line of text on each.
- Assemble and deliver. Original, searchable copy, index or cover sheet, and the delivery note.
- Remove the upload. Once the delivery is confirmed, delete the set from the conversion tool according to your retention schedule.
Where the time goes
Conversion itself is waiting. The labor is in steps 4 and 6. Budget a few minutes per set for the spot-check and make it a logged step, with the checker's initials, so that it survives a busy week.
Faxed records
Custodians still fax, and fax output is poor input for any OCR tool. Thin strokes break up and headers stamp across the first line. If you control the request, ask for electronic delivery or mailed copies. The post on how to OCR a faxed document explains what to expect when you cannot.
Page stamping
If you apply Bates numbers or your own page stamps, decide the order once and keep it. The post on how to OCR Bates numbered documents covers whether to stamp before or after conversion.
Whose confidentiality terms govern records a retrieval company uploads?
Several sets of terms govern records a retrieval company uploads, and they stack. The custodian released the records under an authorization or legal process. The firm holds them under its professional duties and your service agreement. Any OCR vendor you add sits at the end of that chain and must be allowed by every link.
The chain, link by link
| Link | Where its terms live | What to look for |
|---|---|---|
| Patient or record subject | The signed authorization, subpoena or court order | Who may receive the records and for what purpose |
| Custodian | Its release conditions and any agreement you signed with it | Limits on redisclosure |
| Law firm | Your service agreement and any vendor addendum | Whether subcontractors or outside tools need approval |
| Your company | Your own policies and insurance conditions | What you have promised about where records go |
| OCR vendor | Its published terms and commitments | Training, storage, access and deletion |
Why the firm will ask
Lawyers answer for how their vendors handle client material. The District of Columbia's Rule 5.3 on nonlawyer assistants, in its comments, gives sending client documents to a third party for printing and scanning as an example of outside help that the lawyer must make reasonable efforts to keep compatible with the lawyer's own obligations. Other states word the duty differently. The practical result is the same everywhere: expect the firm to ask what tools you use, and have a written answer.
What to do before the first upload
- Read your agreements with your three largest clients for subcontractor and tool clauses.
- Ask each firm, in writing, whether you may use a named online conversion service on its records. Keep the reply.
- Read the vendor's terms yourself. Do not rely on a sales summary.
- Record the answers by client, so staff know which firms have approved conversion and which have not.
What this post will not tell you
Health records and other regulated records carry legal requirements that depend on who you are, who your client is and where you operate. Nothing here says what those require of you, and Schema Law makes no claim of compliance with any health privacy or security standard. Take the question to your own counsel and your contracts. The attorney-side view is in the section on confidentiality when records leave the firm.
What Schema Law does publish is short. The lines it lists for every plan include "Your documents are never used to train an AI model", "Documents stay inside your own account", "Encrypted in transit and at rest" and "Delete any document, at any time". Those commitments are on the homepage. Compare them, as written, with what your contracts demand.
Estimating page volume for a records retrieval business
Estimate page volume for a records retrieval business from your own order history: orders per month, times average pages per order, times the share you will convert. Then look at the peak month, not the average, because plan limits are monthly and litigation work arrives in bursts.
A worked estimate
Take a fictional company, Tri-County Record Services, with these figures from last year's log:
| Input | Figure |
|---|---|
| Orders delivered in an average month | 140 |
| Average pages per order | 60 |
| Average monthly pages | 8,400 |
| Share already text-based on arrival | 20% |
| Pages needing conversion, average month | 6,720 |
| Busiest month last year, pages needing conversion | 11,500 |
The average month fits inside 10,000 pages. The busiest does not. Tri-County has three choices: hold some conversions into the next month, convert only for clients who ask, or buy the larger tier. Which one is right depends on whether late delivery of the searchable copy is acceptable to its clients.
Numbers to pull from your own log
- Pages per order by record type. A hospital chart and a billing record are very different sizes.
- The size of your largest single set, to compare against a per-document page cap.
- How many pages were re-requests. Converting a set twice doubles its cost.
- Seasonal swings, if your clients' filing deadlines cluster.
Sets that break the average
One catastrophic-injury chart can run to thousands of pages and use half a month's allowance in an afternoon. Flag any order over a threshold you choose, say 1,500 pages, for a decision before conversion: split it, schedule it, or quote it separately.
The arithmetic for turning a plan price into a per-page figure is in the post on OCR cost per page, and current plan limits are on the pricing page. If your operation looks less like order-by-order retrieval and more like standing teams processing documents for a few large clients, the guide to OCR for legal process outsourcing fits better.
Charging law firm clients for OCR on retrieved records
Charge law firm clients for OCR in one of three ways: fold it into the base fee, add a per-page line, or add a flat fee per order. Folding it in is simplest to sell. A per-page line tracks your cost most closely. Pick one and state it on the rate sheet.
| Model | How it reads on an invoice | Suits | Watch for |
|---|---|---|---|
| Included in base fee | No separate line | Companies using searchable delivery as a selling point | Very large sets eating the margin |
| Per-page add-on | "Searchable text conversion, 412 pages" | Mixed order sizes | Clients declining it to save money, then asking later |
| Flat fee per order | "Searchable delivery" | Orders of similar size | Needs a cap or a surcharge above a page threshold |
Working out your own cost
Your cost per page has two parts. The first is the tool: plan price divided by pages you actually convert, not pages allowed. The second is labor: the minutes spent splitting, uploading, spot-checking and assembling, at a loaded hourly rate. On small orders the labor is the larger part. Price from both.
Can the firm pass it on?
A firm may want to bill your invoice to the client matter as a cost. Whether a conversion line is recoverable, and how it must be described, depends on the firm's fee agreement and its jurisdiction's rules. That is the firm's question to answer. Help by itemizing clearly so the firm can decide.
A rate-sheet line you can adapt
Searchable delivery: each order includes the records as received and a
text-searchable copy. The searchable copy is machine-generated, is not
proofread, and does not replace the original. Orders over 1,500 pages
are quoted separately.Court reporting firms face the same pricing choice when they offer searchable exhibits, and the guide to OCR for court reporters looks at it from that side.
The buyer's checklist
Use this when comparing tools.
- Its written terms on training, storage and deletion satisfy your strictest client agreement.
- The monthly page allowance covers your peak month, or you have a plan for the overflow.
- The per-document page cap and file-size cap fit your largest routine set.
- It exports a searchable PDF that keeps the original page images.
- You know how many people one plan lets sign in, and what it costs if each person who converts needs an account of their own.
- You can delete a set yourself when the order closes.
- You have run ten of your own worst sets through it and read the results.
- You know what kind of support the plan includes and how to reach it.