Public records · FOIA · Buyer's guide
The demo always looks clean.
These questions find the differences.
Buying redaction software is really buying a review process. Every demo shows a tidy document and a confident pass of automation. The differences show up later, on a messy response with a deadline attached: what the automation actually reads, who decides each mark, what the record shows when a withholding is challenged, and what the busy months cost. Here are eight questions to ask every vendor you evaluate. Including us.
Question 01
Does the automation read the content, or apply patterns configured at setup?
Pattern rules catch formatted strings. A Social Security number or a phone number has a shape, and a rule configured at setup will find that shape reliably. Narrative exemption content has no shape: a victim's identity inside a sentence, medical detail in an officer's summary, privileged advice partway down an email thread. That content has to be read, and if the automation cannot read it, a person still reads every page and the tool has only moved the highlighter.
A good answer names both layers and is plain about which kinds of content each one can catch.
Question 02
Does it propose a mark with the exemption attached, or flag a document as risky?
Some tools score a whole document and hand back a severity flag. That still leaves a person to open the file, find the line and decide the legal reason for withholding it. A proposed mark that already carries its exemption is a decision teed up, not a search assignment. Most of the labor in a response lives in that difference.
A good answer shows a specific mark on a specific line with the exemption already attached, waiting on a person.
Question 03
Who approves each mark, and is the decision recorded by name?
When a withholding is challenged, the defensible record is who decided what, mark by mark. A tool that burns redactions in bulk without named per-mark decisions leaves the office arguing from memory about choices nobody can reconstruct. The audit trail is not a checkbox to skim past. It is what you will stand on.
A good answer pulls up any single mark and shows the content, the exemption, the reviewer's name and the timestamp.
Question 04
Can anything leave before review is finished, and is the output verified?
Two failure modes live in this question. A draft goes out with marks still undecided, or a release ships with content drawn over instead of removed, ready to be copied back out. Ask whether release is gated on every mark being decided. Then ask whether the finished file is verified to confirm the redacted content is actually gone rather than hidden.
A good answer is a gate you can watch refuse to run, plus a verification pass on the output file itself.
Question 05
Is there one log for the whole response?
A response is rarely one PDF. It is documents, email exports, attachments and spreadsheets, and the exemption decisions across all of them belong in one place. If each format keeps its own trail, someone stitches the trails together by hand before the response can be defended as a whole.
A good answer shows one exemption log covering every record in the response, whatever format it arrived in.
Question 06
What happens to a whole email export?
Email is where responses get big. Ask whether attachments become their own records, whether duplicates across overlapping exports are caught once instead of reviewed twice, and whether one giant export file has to be split by hand before the tool will accept it. A tool that treats an export as one enormous document has handed the hardest part back to your staff.
A good answer takes the export whole and hands back individual records, deduplicated, with attachments intact.
Question 07
What are the limits behind the headline?
Ask for the meter. Page credits per year, per-document overflow charges, seat minimums, annual-only billing: every pricing model has a shape, and the shape decides what a bad month costs. Then ask the sharper version. What happens in the busiest month of a lawsuit year, when one request triples your normal volume?
A good answer states the meter in one sentence and can tell you what your worst month looks like without a follow-up call.
Question 08
Where does the content go, and does it train anything?
These are government records, and they often contain exactly the information the redaction exists to protect. Ask where processing happens, what is retained and for how long, and whether your documents are used to train AI models. The answer should be written down before anything uploads, because your counsel and your IT office will ask for it anyway.
A good answer is a written document, not a reassurance: processing location, retention terms and a plain no on training.
The checklist
Take the questions with you.
The same eight, in a form you can copy.
Eight questions for any redaction vendor
- Does the automation read the content, or apply patterns configured at setup?
- Does it propose a mark with the exemption attached, or flag a document as risky?
- Who approves each mark, and is the decision recorded by name?
- Can anything leave before review is finished, and is the output verified?
- Is there one log for the whole response?
- What happens to a whole email export?
- What are the limits behind the headline?
- Where does the content go, and does it train anything?
Paste this into an RFI, or bring it to any demo and work down the list.
Ask us these same questions.
The walkthrough exists so you can. We published this list because we win on these questions when they are asked, and that is exactly the kind of claim you should not take from the vendor making it. Bring the checklist and a real response set, and check every answer against the product on your own records.
Evaluating for a law practice or general document work? This guide has a version written for that world.
Questions about the list
What should a small office without procurement staff do with this list?
Use it without the ceremony. Copy the checklist into an email and ask any vendor for written answers, or keep it open beside a free trial and check each item yourself on a real response set. None of the eight questions requires a procurement office. They only require being asked.
How many of these does typical request-portal software answer?
It varies too much for a general claim. Portals differ widely, many include real redaction tooling, and what your contract includes can differ from what the product can do. Ask the same eight questions of whatever you run today and let the answers tell you what is covered, not the product category.
Why does RecordBeacon publish its own evaluation criteria?
Because we win on these questions when they are asked, and we would rather they were asked. That is a self-serving claim from a vendor, which is why the honest version ends the same way every time: put this list in front of us and verify it yourself.