custom package builder

A custom package builder sits between a live API and pre-packaged bulk files: you choose series interactively and the system renders a download containing only those columns. The cataloged example is the Federal Reserve Data Download Program (DDP), whose record describes it as a "bulk-download builder for statistical releases ... delivering custom CSV and XML packages of selected series" across releases including H.8, Z.1, G.19 and H.15.

Two documented behaviors define the format. Packages "default to full history unless from/to dates are supplied", so omitting dates buys you the deepest series - H.8 weekly data begin 1973, Z.1 quarterly back to 1945. And Output.aspx parameters control filetype (CSV or XML), label inclusion and series-per-column versus series-per-row layout, which decides how much reshaping your pipeline does downstream.

Fixed archives force you to download everything to use anything; APIs move shaping cost onto your infrastructure. A builder avoids both when your series list is known and stable.

• Bandwidth follows selection. Z.1 alone ships 286 CSV tables per quarter (about 8 MB zipped); if you need six banking series, a custom package is smaller than any full-release archive.
• Date defaults can surprise you. Full-history-by-default means an unparameterized request returns decades of rows - set from/to deliberately rather than by accident.
• Layout choices are schema decisions. Choosing series-per-column over series-per-row here determines whether your loader pivots later, so decide once and document it.

1. Confirm selection happens at series level. The DDP granularity field reads "Individual series within each release" - a builder that only filters at release level is a plain archive.

1. Test the date-window behavior. Packages default to full history unless from/to dates are supplied; verify both modes return what the labels claim before automating.

1. Check output formats against your stack. CSV and XML are the documented DDP outputs; confirm your candidate source offers formats your loaders accept without conversion.

1. Match frequency expectations per release. Observations arrive "at weekly, monthly, quarterly, or annual frequency depending on the release" - H.8 weekly versus Z.1 quarterly is a real difference in joins.

1. Price in refresh cadence. The cataloged DDP record updates weekly; align package regeneration with release schedules instead of polling blindly.

Entries that sit next to custom package builder in this glossary:

• bulk-download
• federal-reserve-data-download-program