Document projects fail at the budget stage more often than at the technical stage. The usual sequence: somebody asks a vendor for a rough number, the vendor quotes the whole corpus at a retail rate, the total lands somewhere between uncomfortable and impossible, and the item never makes it into the proposed budget.
A defensible request looks different. It starts from a measured page count, applies the reductions before quoting, phases the work across fiscal years, and states plainly what happens if the answer is no.
Know the denominator first
Every unit price is meaningless without a page count, and agency estimates of their own corpus are consistently low — in our audits, by a factor of five to eight. Building a budget on a staff estimate guarantees either an embarrassing overrun or a request that quietly omits most of the work.
Measure the corpus before you price anything. See estimating your corpus size and building the inventory.
Apply the reductions before you quote
Three levers cut scope before a single dollar is spent, and all three are staff-time work rather than purchased services:
- Retention disposition. Records past their retention period, properly approved for destruction, leave the scope entirely.
- Deduplication. Content-hash matching typically removes a large share of a web corpus — commonly 30–40% — because the same document is linked from many places.
- Archive classification. Content that genuinely qualifies under the ADA Title II archive exception may be handled differently from active content.
Run these first and your budget number can drop by half before you have talked to a vendor. Run them afterward and you will have paid to process material you did not need.
What drives unit cost
Do not build a budget on a single blended rate. The variance between categories is larger than the variance between vendors.
| Cost driver | Cheap end | Expensive end |
|---|---|---|
| Preparation | Clean, loose, uniform paper | Stapled, bound, fragile, mixed sizes |
| Capture | Letter-size bitonal at 300 dpi | Large-format color, microfilm, photographs |
| Indexing | One field per document | Multiple keyed fields per page |
| OCR | Clean modern typescript | Faded, skewed, handwritten annotation |
| Accessibility remediation | Born-digital text with real styles | Scanned images, complex tables, forms, maps |
| Review | No personal information present | Page-by-page redaction review |
Two structural points matter more than any individual rate:
- Scanning and remediation are different purchases. A scanning bureau produces images and a text layer. Making a document accessible is separate work with a separate price. Budgets that assume "scanning includes accessibility" are short by the entire cost of the second line.
- Automation changes the shape of the curve. Manual per-page remediation scales linearly forever. Automated processing carries a setup cost and then a much lower marginal cost, which is what makes six-figure page counts survivable. Ask any vendor what share of your corpus their process handles without human touch, and what happens to the rest.
Phase it across fiscal years
A three-year shape that survives budget review:
- Year 1 — measure and prove. Corpus audit, retention pass, deduplication, and a scoped pilot on the highest-priority series. Small dollar figure, concrete deliverable, and it produces the numbers that justify years 2 and 3.
- Year 2 — the bulk. The largest and most-used series, with automation carrying the volume. This is the year with the real money in it.
- Year 3 — the tail and the day-forward process. Remaining backlog plus the recurring capacity to keep new documents compliant as they are created. If year 3 does not fund the ongoing process, the backlog rebuilds immediately.
Fund the recurring line separately from the one-time line, and name it that way in the budget. Agencies that treat this as a one-time capital-style project are back where they started within two years, because agenda packets alone add thousands of pages a year.
What belongs in the staff report
- The measured number. Documents, pages, and how the count was produced.
- The reductions already applied. Show the corpus before and after retention, deduplication, and archive classification. This is the single most persuasive element, because it demonstrates the request has already been minimized.
- The obligation. What law or rule requires the work, and by when. Link to the Title II deadline for the accessibility component.
- The exposure if deferred. Complaint and litigation risk, staff time consumed by manual retrieval, and the cost of doing the same work later under a deadline.
- The non-compliance alternative, stated honestly. Doing nothing is a decision the body is entitled to make, and pretending otherwise damages your credibility.
- Phasing and the recurring cost, so nobody believes this ends after one year.
Funding and procurement routes
- Cost allocation. Enterprise funds — water, sewer, refuse — can often carry their proportional share of a records program that serves them.
- Existing technology or records line items. Frequently large enough to fund year 1 without a new appropriation.
- Grants. Records preservation and accessibility programs appear in various state and federal programs; availability shifts year to year, so check current cycles rather than assuming.
- Cooperative purchasing. Riding an existing cooperative or another agency's competitively awarded contract can remove months from the schedule. See our procurement page.
- Sole source, where genuinely justified by unique capability and documented properly. See sole source documentation.
The audit is the cheapest line in the budget and the one that moves the total most. A measured page count, a duplicate rate, and a retention pass routinely cut a project's scope by half before procurement starts. Bring that to the budget meeting instead of a vendor's estimate of a corpus nobody has counted.