Summary

  • FedStats opened on 22 May 1997 to help readers find official statistics without first learning which agency produced them.
  • A usability study published in 2000 found misleading expectations around familiar labels; an OMB report also distinguished roughly 100 linked agencies from 40 indexed websites.

A button called Fast Facts has made a small promise before anyone clicks it. At FedStats, readers expected quick numbers and found compilations of numbers instead. The proposed replacement was admirably literal: Statistical Reference Shelf. The government had advertised a snack and delivered the cupboard.

That mismatch appears in a 2000 paper hosted by BLS, whose team examined a site built to make a sprawling statistical system easier to use. Eleven people attended the test, drawn from a recruited group of sixteen professionals in fields such as journalism and policy research. Across two working days, their encounters exposed a second problem: many took A to Z to mean agencies; almost nobody expected a topic index. Renaming the links gave readers a better idea of the destination. The paper presented a redesigned homepage and planned further testing, with a September 2000 public rollout envisaged.

The original assignment had been entirely sensible. On 22 May 1997, the Interagency Council on Statistical Policy launched FedStats, following the Federal Statistics Briefing Rooms introduced the previous spring. The FY1998 OMB account described a central collection of links to the agencies' own websites. Someone seeking a statistic could start with the question that mattered to them. Memorizing the federal organization chart was supposed to become somebody else's hobby.

Putting those doors together made their wording unusually consequential. A reader chose a route before seeing the underlying table, its definitions or its date. An agency name helped someone who already knew the producer; a topic label helped someone who knew only the subject. Both belonged in a useful gateway. Their labels had to tell the two readers which kind of help was on offer.

There was another boundary behind the welcome mat. The FY2001 statistical-programs report described access to approximately 100 agencies, while search indexed 40 of their websites. Those figures describe that report's service, years after launch. A link collection could introduce a reader to more institutions than its search engine could inspect. The same report recorded more than 3.5 million user sessions since inception: plenty of visits to the doorway, with the separate question of what each visitor managed to find.

FedStats therefore had two kinds of editorial work. It selected routes through distributed public records, and it explained those routes in a few words. The statistical tables could remain with their makers while a badly named button sent the reader down the wrong corridor. On the early public web, even a government full of numbers needed someone to edit the nouns.