How to Route a Request to the Right Custodian

Score it on intake for subject, likely custodians, andvolume, then send it to a named owner rather than a shared mailbox. Requeststouching several departments need one accountable owner and separate tasks, orevery department assumes another one is handling it.

Routing sounds like the most administrative part of thisjob, which is why it rarely gets designed. It is also where a surprising shareof late responses begin, and almost none of them begin with anyone refusing tohelp.

A request arrives, lands in a shared inbox, and sits forthree days while people work out whose it is. Nobody ignored it. Everyone whosaw it reasonably assumed it belonged to someone else, because nothing about itsaid otherwise.

That pattern repeats at every internal boundary the requestcrosses. It is not a motivation problem and it does not respond to remindersabout how important records compliance is. It responds to two structuraldecisions, made once, in advance.

Scoring means spending two or three minutes, while therequest is still in front of you, deciding three things: what it is about, whois likely to hold the material, and roughly how big it is.

Subject is usually obvious from the text. Likely custodianstake a moment of thought and are worth writing down even when you are unsure,because a wrong guess that is recorded is easier to correct than an assumptionnobody stated. Volume matters because it changes what happens next: a requestfor one contract is a different piece of work from a request for eighteenmonths of correspondence, and treating them the same way means the small onewaits behind the large one.

Scoring also tells you early whether the request should besplit. A single submission covering two unrelated bodies of records is reallytwo requests. Different departments hold them, they move at different rates,and bundling them means whichever half is ready waits for the half that is not.

Every request gets one person's name against it. Not adepartment, not a shared address, not a rota.

The reason is simple: a shared mailbox has no memory and noaccountability. When the deadline approaches, somebody has to be able to answerwhat the current state is, and that answer has to come from a person who hasbeen carrying it rather than whoever opens the inbox that morning.

Name a backup at the same time. It reads as bureaucraticdetail right up until the owner takes two weeks of leave partway through acomplex request. The backup is not there to do the work. They are there tonotice that nothing is happening.

This is where routing earns its keep. A request touchingthree departments and belonging to none of them will drift, because eachdepartment can see its own piece and assume the whole is somebody else'sconcern.

Two rules prevent most of it. The request keeps oneaccountable owner, who does not change as the work moves. And each departmentgets its own task, with its own due date, under that single owner.

Separate tasks matter more than they appear to. With oneaggregate status, partial progress is invisible: the request reads as openwhether two departments have delivered or none have. With separate tasks youcan see on any given day that procurement returned material a week ago andlegal has not started, which is the information that lets you act while actingis still cheap.

Most agencies route the same categories over and over.Police records go to the same place every time. Personnel matters go to HR andto counsel. Contracts go to procurement. Writing those down as standing rulesremoves a decision from every future request rather than making it fresh eachtime.

Standing rules are especially valuable in a small office,where the whole records function may be one person doing this alongside severalother duties. Decisions made in advance are the main thing that makes thatworkload survivable.

The rules do need an exception path. The request that doesnot fit any category is exactly the one that will sit longest if the onlyoptions are the categories, so there should be a default owner for anythingunclassified, and that default should be a person.

Govflo scores everyrequest on arrival for both priority and complexity, grading complexity low,medium, or high against the agency's own history: whether the documents arealready public, whether a previous request has produced them, how muchspecialized judgment they need. High-complexity requests route to a person withexpertise in that subject instead of waiting in a general queue for someone tonotice how hard they are. Where one submission is really two jobs, it can besplit into separate cases and each scored on its own. An outside expert can beinvited onto a single case and sees only that case.

The scoring learns the agency's own norms rather than avendor's. An office that answers a category in six days gets six-dayrecommendations; one that takes ten gets ten. What it does not do is decide whoin your agency should own a category. That mapping reflects how theorganization is actually arranged, and it belongs to the agency to define.

Routing gets the request to the right desk. It does not makethe person at that desk respond.

A department that has gone quiet is a different problem, andit is a political one wearing an operational costume. No routing rule solvesit. What helps is escalating on a schedule rather than on frustration, inwriting, with every attempt recorded, which is covered in when adepartment will not respond.

The reason the distinction matters is that agencies oftentry to fix a non-response problem by rebuilding their routing, which changesnothing and costs a month. Getting the request to the right place and gettingan answer back are two different failures with two different remedies.

‍