The Records Request Process, and Why the Backlog Compounds

More people, a longer line

The records office got the new positions it asked for. A year later the list of open requests is longer than it was before they were hired, the city attorney is handling two petitions over late responses, and the council wants to know what the money bought.

Almost everything written about records backlogs comes from the requester's side of the counter. It treats the backlog as a transparency failure: an agency that is slow because it would rather not answer. Sometimes that is true. More often, what you are looking at is a queueing failure, and queues of this kind have a property most budget conversations miss. They compound. Every day a request ages, it creates new work that is not fulfilment.

The federal numbers show the shape

The federal FOIA system is the only records operation in the country that publishes a full year of its own data, and the most recent year is a clean illustration.

In fiscal 2024, federal agencies received a record 1,501,432 FOIA requests, 25 percent more than the year before. They also processed 34 percent more requests than the year before, and reported the first increase in full-time FOIA staff since 2020, to roughly 5,600 people. By any operating measure, that is a system working harder with more resources.

The backlog still grew 33 percent, from 200,843 requests to 267,056. The average time to close a simple request rose from 39.4 days to 44. The total cost of running the system reached $723 million, and FOIA fees recovered 0.34 percent of it.

Then look at the lines that are not fulfilment. Administrative appeals rose 39 percent, from 14,443 to 20,115. Agencies reported $54 million in litigation-related costs, up 10 percent. GAO, reviewing why backlogs persist, lists responding to lawsuits from requesters among the causes, alongside larger and more complex requests and staff turnover. Read that plainly: the lawsuits caused by the backlog are consuming the capacity that would have reduced the backlog.

Local governments do not publish numbers this complete, but the demand side is moving the same way. San Antonio's open government manager reported more than 86,000 requests in 2025, against roughly 4,000 a year a decade earlier, with the team growing from two people to twelve over five years. That figure comes via a records software vendor's blog and should be read as such, but no plausible correction to it changes the direction.

Where the compounding comes from

A request that is closed in its first week costs the agency one unit of work: find the records, review them, redact, release. A request that sits for four months costs that same unit, plus everything its age generated along the way.

Status contact. The requester calls, emails, or submits a follow-up. Each contact has to be read, matched to the open request, and answered, usually by the same person who would otherwise be working the queue.

Extension and acknowledgement letters. Many statutes require notice when a deadline will be missed. Every aging request produces correspondence, and often more than one round of it.

Scope drift. A requester who has waited three months frequently broadens or restates the request, or files a second one to cover what they think the first will miss. Now there are two items in the queue for one need.

Appeals and petitions. Late or partial responses invite appeals, and appeals are handled by more senior staff whose time is scarcer.

Litigation. Counsel has to assemble a history of the request, a declaration, and sometimes a rushed production under a court schedule. That work jumps the line, which pushes everyone else further back.

None of this appears on the dashboard as backlog. It shows up as a records team that is visibly busy all day and closing fewer requests than its headcount suggests it should. The staff are not slow. A growing share of their week is spent servicing the age of the queue rather than emptying it.

Why hiring buys less than the arithmetic says

This is the part that makes budget hearings go badly.

The business case for new records staff is usually built on simple division: requests per year, hours per request, hours per employee. It assumes every new hour goes into fulfilment. In a compounding queue, a share of each new hour is immediately absorbed by the correspondence, appeals, and legal support that the existing backlog is already generating. Capacity arrives, and part of it is spoken for before the new hire finishes onboarding.

That is why the federal system could process a third more requests and still end the year a third further behind. Headcount is not wasted in a backlog like this. It is just a weaker lever than the spreadsheet promised, and it gets weaker the older the queue is when the money arrives.

We made a related argument about procurement, where the delay nobody measures sits in the document routing before the solicitation posts. Records has the same trait: the visible queue is not where most of the cost lives.

The mixing problem

There is a second mechanism, and it is the one an agency can actually act on.

Records requests are not one kind of work. Some are a single document the clerk can find by name. Some are a year of email across six custodians, requiring keyword searches, legal review, and redaction of every page. The difference in effort is not a matter of degrees. It can be minutes against weeks.

When both kinds sit in one first-in, first-out list, the five-minute request waits behind the five-week one. It ages for no reason, and as it ages it starts generating its own status calls and extension letters. A request that should have cost almost nothing becomes one more source of compounding. Multiply that by every simple request stuck behind a hard one and a large part of the backlog is made of work that was never difficult.

Federal law already recognises this. FOIA permits agencies to run separate processing tracks, and the annual summary reports simple and complex requests separately. Many local offices we look at still run a single list, often because the intake form and the tracking tool were set up that way and nobody revisited it.

Triage at intake breaks the loop

The intervention that actually changes the curve is not a faster search tool or another position. It is sorting requests by expected effort at the moment they arrive, and never letting the quick ones wait behind the slow ones.

Done well, it does three things at once. The simple requests close inside the statutory window, so they never generate the correspondence and appeals that feed the compounding. The complex requests get an honest timeline on day one, which removes a large share of the "where is my request" contact. And the records officer gets an early conversation with the requester about scope, while that conversation is still cheap.

The same principle shows up in a very different setting. An insurer's submission triage matters less for speed than for deciding, early, what deserves scarce expert time. A records office is running the same problem with a statutory clock attached.

What triage cannot be is a form field. The categories, the effort estimates, who makes the call at intake, how scope negotiation is handled, and which requests go to counsel early are all specific to your statute, your records systems, and your mix of requesters. A newsroom, a law firm, and a resident asking about their own property generate very different work, and the rules that separate them depend on what your last year of requests actually looked like.

What not to buy first

Records management vendors will offer portals, automated acknowledgements, and AI-assisted redaction. Some of that is genuinely useful. A portal that tells a requester where their request stands can remove a real share of status contact. Redaction tools can shorten the hardest requests.

But a portal on top of a single untriaged queue mostly gives requesters a clearer view of how long they are waiting. And faster redaction helps the complex requests while doing nothing for the simple ones stuck behind them. As with automating permit intake before simplifying the permit types, the tool arrives before the structure it depends on. Decide how requests are sorted first. Then buy the software that fits that decision, rather than the software that makes one for you.

The cheapest question to ask before any purchase: of the requests you closed last year, how many took under an hour of staff time, and how many of those waited longer than the statutory deadline anyway? If the answer is a large number, you do not have a capacity problem yet. You have a sorting problem, and the backlog is only on file because nobody filed it by effort.

Where to start

If your records backlog keeps growing after you add staff, the compounding is probably already underway, and the useful first step is finding out how much of your team's week is going to the age of the queue rather than to fulfilment. We can help you pull a year of request history, separate the quick requests from the hard ones, measure what the waiting has been generating, and tell you honestly whether you need people, triage, or a tool. Book a consult and bring last year's request log.