Every agency with a paper backlog eventually asks a vendor what it would cost to scan all of it. The number comes back, somebody says it out loud in a budget meeting, and the project dies for another three years.
The projects that survive are the ones that never proposed scanning everything. They scanned the right 15% first, produced something staff and the public could feel within a fiscal year, and used that to fund the rest.
Score before you sequence
Four factors decide whether a box of records deserves to go early. Score each series from 1 to 5 and add them up.
| Factor | Score 5 when… | Score 1 when… |
|---|---|---|
| Remaining retention life | Permanent or decades left | Eligible for disposition within a year |
| Public and staff demand | Pulled weekly; frequent records requests | Not requested in living memory |
| Risk if lost or unavailable | Vital record; title, safety, or entitlement evidence | Convenience copy of something held elsewhere |
| Physical deterioration | Thermal paper, brittle, water-damaged, fading | Stable, stored well, recent |
A series scoring 16–20 goes in phase one. Twelve to fifteen goes in phase two. Below twelve, ask whether it should be scanned at all.
The series that almost always win
- Building permits and approved plans. Permanent or near-permanent retention, constant demand from owners, contractors, and realtors, and enormous risk if the only copy is lost. This is the single most common phase-one choice, and the one that generates visible staff time savings fastest.
- Recorded instruments affecting property. Deeds, easements, rights-of-way, abandonments. Permanent, high-risk, frequently pulled.
- Governing body records. Minutes, ordinances, resolutions. Permanent by nature, requested constantly, and usually the first thing the public looks for online.
- Records in visible decay. Thermal fax paper and 1970s–80s copier output are actively disappearing. Condition can outrank demand.
- Whatever your CPRA log says. Pull the last two years of public records requests and count which series were touched most. That log is the closest thing you have to measured demand.
Use the request log as data. Most agencies already have a list of what the public asks for and never analyze it. Sorting two years of requests by record series takes an afternoon and will change your phase-one list. See answering CPRA requests faster.
What not to scan
Scanning is not free, and a scanned record you did not need is worse than useless — it now costs storage, indexing, and possibly accessibility remediation forever.
- Records past their retention period. Run retention first. Every box that can be lawfully destroyed is a box you never pay to scan. This is the largest single cost reduction available in most programs.
- Duplicates and convenience copies. Departments keep copies of the official record. Scan the official one.
- Records already digital elsewhere. County-recorded documents, state-filed reports, and vendor-hosted submittals may already exist in authoritative digital form.
- Reference material that is not a record. Vendor catalogs, code books, published standards, and manuals are usually not agency records at all.
- Anything under an unresolved legal hold. Not never — but not on a production line, and not without counsel involved.
Sequencing against real capacity
The constraint is rarely the scanner. It is staff time for preparation and indexing, and departmental capacity to answer questions about their own records.
A workable phase-one shape for a mid-size agency:
- Months 1–2: Inventory and retention pass. Identify what can be destroyed instead of scanned.
- Months 2–3: Pick one department and one series. Write the specification, run a pilot of a few thousand images, and check the output honestly.
- Months 4–9: Production on that series. Fix the indexing scheme now, while the volume is small enough to change your mind.
- Months 10–12: Publish what belongs in public. Measure staff time saved and request volume shifted. That measurement is your budget request for phase two.
Plan for what happens after the scan
A scanned image is not a finished product. Three obligations attach to it the moment it exists:
- It needs a text layer. An image-only PDF cannot be searched, cannot be read aloud, and cannot satisfy accessibility requirements. Verify the OCR rather than trusting it — see testing OCR quality.
- It may contain personal information. Older permit files, claims, and personnel records are full of material you cannot publish as-is. See finding PII before you publish.
- If you publish it, it has to be accessible. Scanning a backlog onto a public website converts a storage problem into an ADA Title II problem. Decide deliberately what goes public.
The cheapest phase one is the one preceded by a retention pass and a duplicate check. Agencies routinely find that a quarter to a third of the boxes queued for scanning either should be destroyed or already exist in digital form. That work costs a few weeks of staff time and removes months of scanning.