# OCR With Data Retention Control: Who Sets the Clock

- Author: Schema Law (https://www.schemalaw.com/)
- Published: 2023-01-18T17:43:16Z
- Updated: 2026-10-11T19:22:20Z
- Category: Confidentiality
- Canonical: https://www.schemalaw.com/blog/ocr-with-data-retention-control

> OCR with data retention control means your firm sets how long uploads live. How to compare tools on periods, holds, per-matter rules and proof of deletion.

OCR with data retention control means your firm, and not the vendor, decides how long each uploaded scan and its converted text exist. To compare tools, check four things: who sets the period, whether it can differ by matter, whether deletion can be paused for a hold, and what evidence of deletion you get.

A tool that simply deletes everything after an hour gives you a short clock, which is useful, but it is still the vendor's clock. This post is about the difference, and about what a firm needs once it starts setting the clock itself. It is practical guidance on handling documents and is not legal advice. Preservation and file-retention duties vary by court and by state, so check your own.

## What does data retention control mean in an OCR tool?

Data retention control in an OCR tool is the ability to set, change and suspend the period after which your uploads and their converted output are removed. A fixed deletion window that you cannot alter is a retention policy. Control begins when the setting is yours.

People searching for OCR that doesn't keep your files usually want one of two things, and it helps to know which.

- **A tool that keeps nothing.** The file is converted and discarded. You have no decision to make and no way to come back for a second export.
- **A tool that keeps what you tell it to.** The file stays for a period you chose, and goes when that period ends or when you delete it.

The first is a property of the vendor. The second is a control in your hands. Neither is better in the abstract. A one-page letter you will never touch again suits the first. A production you will export three ways over a month needs the second.

### The five levers

A retention claim can be broken into separate levers. A vendor may give you some and hold the rest.

| Lever | What it lets the firm do | What it looks like when missing |
| --- | --- | --- |
| Period | Choose the number of days before automatic removal | One window for every customer |
| Scope | Apply different periods to different matters or folders | A single account-wide setting |
| Manual delete | Remove one document now, ahead of the period | Deletion only by support request |
| Suspension | Stop automatic removal for named documents | The timer runs no matter what |
| Evidence | Obtain a record of what was removed and when | The document just disappears from the list |

When a sales page says "you control retention", ask which of the five it means.

## Vendor default retention versus a firm retention schedule

A vendor default is the period a service applies to every customer unless told otherwise. A firm retention schedule is the firm's own written rule for how long each class of record is kept. Control means the tool can be made to follow the schedule, so that the vendor's default stops mattering.

The two were written for different purposes. A vendor's default is one number chosen for its whole customer base. Your schedule exists because of duties to clients, the rules of your jurisdiction on client files, and the firm's own risk decisions. There is no reason the two numbers should match.

Give records retention a line for working copies that live in a converter, shorter than the matter file. Personal data in those copies can come out once the export is saved and checked. Client confidentiality is a practical reason for that shorter line. Name the person who runs the deletion calendar so the schedule does not depend on whoever uploaded last.

ABA Formal Opinion 483 ties a schedule to breach exposure directly. In discussing former clients, [ABA Formal Opinion 483](https://www.americanbar.org/content/dam/aba/administrative/professional_responsibility/ethics-opinions/aba-formal-op-483.pdf) says lawyers "are encouraged to adopt and follow a paper and electronic document retention schedule, which meets all applicable laws and rules, to reduce the amount of information relating to the representation of former clients that the lawyers retain." An OCR account is one of the places that information accumulates.

### Where the converter sits in the schedule

A schedule written around the matter file may have no line for a converter account, which holds working copies. Give it one. A workable entry has three parts:

1. **The class.** "Working copies in outside conversion tools."
2. **The period.** Shorter than the matter file. The end of the task, or the end of the matter's active phase, are two natural choices.
3. **The trigger for early removal.** Export saved to the matter file and checked against the scan.

### When the default is shorter than you need

A short vendor default is not always a gift. If the service removes uploads after a week and you are midway through a privilege review, you will upload the same set again, which creates a second trail of copies on every laptop involved. If the tool cannot lengthen the period, plan the work so exports are finished inside the window.

### When the default is longer than you want

"Kept until you delete" is the opposite trap. Nothing forces a decision, so nothing gets deleted. If the tool has no automatic period you can set, your schedule has to be carried out by a person with a calendar entry. Write down who.

## How does a litigation hold change what you may delete?

A litigation hold suspends ordinary deletion for information that may be relevant to a dispute. Once a hold applies to documents sitting in an OCR account, automatic removal has to stop for those documents, and a retention setting that cannot be paused becomes a problem.

The federal rule shows what is at stake. [Federal Rule of Civil Procedure 37(e)](https://www.law.cornell.edu/rules/frcp/rule_37) applies when "electronically stored information that should have been preserved in the anticipation or conduct of litigation is lost because a party failed to take reasonable steps to preserve it, and it cannot be restored or replaced through additional discovery." The committee notes to the 2015 amendment add that "Many court decisions hold that potential litigants have a duty to preserve relevant information when litigation is reasonably foreseeable." State courts have their own rules, and the scope of any hold is a legal question for the lawyers on the matter.

### Is the converter copy the thing to preserve?

The converter may hold only a duplicate. The scan came from the client's records or from a production, and the original is preserved elsewhere. The rule quoted above is concerned with information that "cannot be restored or replaced". Whether a duplicate in a working tool falls within a hold is for the responsible lawyer to decide, and the answer should be written down before anyone presses delete.

Three situations deserve a closer look before you treat the tool's copy as disposable:

- **The upload is the only copy.** A paralegal photographed a document at a client's office, uploaded the photo and [deleted it from the phone](https://www.schemalaw.com/blog/delete-scanned-documents-from-phone).
- **The converted output has been edited.** Someone corrected the Word export by hand, and those corrections exist nowhere else.
- **The hold concerns the firm itself.** A fee dispute or a malpractice claim puts the firm's own handling of documents in issue, including what it sent to vendors and when.

### What to do in the tool when a hold lands

1. Identify the matters under hold and find their documents in the OCR account.
2. Turn off or extend any automatic deletion that would reach them. If the tool has no pause, export the documents and their output to the firm's own preserved storage, and record that you did.
3. Tell everyone with access not to delete manually in those matters.
4. Note the date the hold reached the tool and who made the change.
5. When the hold is released in writing, restore the normal period. Do not leave the pause in place forever.

A free tier with a fixed short window deserves particular care here. The timer will run whether or not a hold exists, so do not let such an account become the only home of anything.

## Retention rules by matter, client and outside counsel guideline

Retention periods often need to differ within one firm. One client's guidelines may demand deletion within days of use, a court order may fix an end date for a single case, and a third matter may be under hold. A tool with one account-wide period cannot serve all three at once.

### A worked example

Take a fictional six-lawyer firm, Hollis & Marrow, with three active matters in the same OCR account.

| Matter | Source of the rule | What the rule requires | Setting the firm wants |
| --- | --- | --- | --- |
| Insurance defense for Calder Mutual | The insurer's outside counsel guidelines | Client data removed from third-party tools within 30 days of last use | 30-day automatic period |
| Trade secret case, Pruett v. Dane Tooling | A protective order | Designated material returned or destroyed after final disposition | Keep through the case, then delete everything on a fixed date |
| Estate dispute, Estate of Okafor | A litigation hold issued by the firm | Nothing deleted until release | Automatic deletion off |

With per-matter settings, this is three entries. With a single global period, the firm has to choose the longest (keep everything) and then delete the insurer's files by hand every month. That works only while someone remembers to do it.

The clause in the second row is the kind that binds the firm whatever the vendor's default says. The post on protective orders explains [what a destruction clause requires of converted copies when the case ends](https://www.schemalaw.com/blog/ocr-documents-under-a-protective-order#returning-or-destroying-converted-copies-when-the-case-ends).

### Reading an outside counsel guideline for retention terms

A corporate client may attach guidelines to the engagement. Look for four things and copy them into the matter-opening memo:

- Whether third-party tools need the client's approval before any upload.
- A maximum period for client data in such tools.
- Whether the client wants written confirmation of deletion at the end.
- Whether the client may audit or ask where its data has been.

If a guideline is stricter than the tool can deliver, the fix is procedural: a shorter manual cycle for that client, or a different route for its documents.

### What Schema Law's plans say

Schema Law is an example of a tool where the firm carries the schedule itself. No plan has a retention setting, nothing is deleted automatically, and a document stays until you delete it. Every plan lists "Delete any document, at any time", which is the manual delete lever from the table of five. The product has no matters or shared workspaces either. Each account is one person's Google sign-in, and that person's uploads sit in it. A firm like Hollis & Marrow would therefore carry out all three rules by hand: a recurring calendar entry for the insurer's 30 days, a dated task for the protective order, and a written instruction that nobody deletes the estate documents until the hold is released. The plan-by-plan reading is in the post on [retention settings by plan in Schema Law](https://www.schemalaw.com/blog/does-ocr-software-store-my-documents#retention-settings-by-plan-in-schema-law), and the [pricing page](https://www.schemalaw.com/pricing) shows the plans.

## How can a firm prove an uploaded document was deleted?

A firm proves deletion with records, because it cannot inspect a vendor's servers. The proof is a chain: the firm's own log of the delete action, the vendor's written statement of what deletion removes and when backups expire, and, where the stakes justify it, a written confirmation for the specific matter.

### What each kind of evidence is worth

| Evidence | Who creates it | What it shows | Its limit |
| --- | --- | --- | --- |
| Screenshot or note of the delete action | The firm | A named person removed a named document on a date | Says nothing about backups |
| Activity history in the tool, if it has one | The vendor's system | The system registered the deletion | Not every tool offers one |
| The vendor's retention terms | The vendor | The stated purge period for live systems and backups | A general statement, not specific to your file |
| Written confirmation for a matter | The vendor, on request | That the named documents are gone from all systems as of a date | Depends on the vendor agreeing to give it |
| A contract clause requiring certification | Both | A duty to confirm deletion on request | Must be negotiated before you need it |

### Borrowing the form from media sanitization

Deletion records have a well-developed model in hardware disposal. [NIST Special Publication 800-88, Revision 1, Guidelines for Media Sanitization](https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-88r1.pdf), says that following sanitization "a certificate of media disposition should be completed for each piece of electronic media that has been sanitized", and that it "may be a piece of paper or an electronic record of the action taken." Its Appendix G offers a sample form. The guideline is written for physical media and is not a rule about hosted services. It is cited here only for the idea that deletion is something you document at the time.

Adapted to an OCR account, a deletion record for a matter holds:

- Matter name or number, and the client.
- The documents removed, by file name or count, and the date range of the uploads.
- The date and time of deletion and the person who did it.
- The vendor's stated period for backups to expire, and the resulting "last copy gone" date.
- Local copies cleared: downloads, email attachments, shared drives.
- Any vendor confirmation, attached.

File the record with the closing memo. If a client or a court later asks what became of the converted copies, the answer is one page.

### What a record cannot do

No record from the firm proves a negative about someone else's infrastructure. Be plain about that when a client asks. What you can show is that you issued the instruction, that the vendor's terms commit it to a purge period, and that the period has passed.

## Questions that test a vendor's retention control claim

Test a retention control claim by asking questions that have checkable answers, then trying the setting yourself with a harmless file. Vendors that offer real control can say where the setting is, what it covers and what happens when it fires.

Put these to the vendor in writing:

1. Where in the product is the retention period set, and who in our firm is able to change it?
2. What are the shortest and longest periods we can choose?
3. Can the period differ by matter, folder or workspace, or is it one setting for the account?
4. From what moment does the period run: upload, last access or last export?
5. When the period ends, what is removed? The original, the parsed text, every export, the file name in logs?
6. How long after that do backups still hold the document?
7. Can we suspend automatic deletion for named documents, and is the suspension recorded?
8. Can we delete a single document immediately, and is that deletion different in effect from automatic removal?
9. Is there a history showing who deleted what and when?
10. Will you confirm deletion in writing for a specific matter if a client or court requires it?
11. If we downgrade or cancel, which retention rule applies to documents already in the account?
12. If you are legally required to keep something we have deleted, will you tell us?

### Then test it

Upload a blank form or a public court notice. Set the shortest period available. Note the time. Come back after it should have fired and check whether the document, its parsed pages and its exports are all gone. Then delete a second test file by hand and compare. A test you ran yourself tells you more than a page of marketing.

Retention is one of several separate promises. A file that is deleted on schedule may still have been visible to the wrong people while it existed, which is why the post on [OCR tools that do not share data between customers](https://www.schemalaw.com/blog/ocr-tool-that-does-not-share-data-between-customers) treats separation as its own question. Schema Law's own wording on deletion sits under [its security commitments](https://www.schemalaw.com/#security): "Delete any document whenever you choose. The original file and everything parsed from it are removed." Put the twelve questions to it as you would to anyone else.

## Try it on a real document

Upload a scanned PDF and export it as Word, a searchable PDF or plain text on the free plan. Your documents are never used to train an AI model. [Start free](https://www.schemalaw.com/api/auth/google)