Scope. Grant Platform is a Back Road Creative product: one application, one database and one shared record for work that used to run as two separate systems — award administration and compliance on one side, discovery and applications on the other. The source is not public, and it is available through a studio engagement. Contact is where to start.
One record, because the handoff is where grants go wrong
A grant has two halves. One party designs the program, makes the award and monitors it. The other finds the opportunity, applies, and then owes reports for years. The moments that cost money — application to award, award to reporting, reporting to closeout — are exactly the moments that used to sit between two systems that never spoke.
Grant Platform is one application over one database, with a profile deciding how much of it an install turns on: the recipient half, the award half, or both plus the explicit application-to-award handoff. A profile is configuration, not a separate build. Turning one off removes its routes, its navigation and its report keys; it never leaves a half-wired screen behind.
Discovery through application
- A catalogue with one stated meaning of “match”. An organization records what it works on and where; a catalogue entry records what it funds, where, and who it is open to. One module compares those two vocabularies, so what counts as a match is written once and can be argued with in one file. A closed or past-deadline opportunity is not a match, and an organization that has recorded no geography is treated as unknown rather than everywhere.
- A rank that is a count, never a percentage. Matches are ranked by how many narrowing signals hold — nought, one or two — because a weighted percentage would have to invent weights nobody agreed, and “87% match” reads as a measurement when it is an opinion. Beyond the count the order is fact again: soonest close date first, then name. Org-type eligibility narrows that ranking and never removes a row, because both sides of the comparison are free text from an upstream feed and a gate made of two uncontrolled vocabularies fails silently.
- Applications, checklists and documents. Shortlisting creates the application with its checklist already built, required items gate submission, and generated application documents are built from facts already on file.
Award administration and compliance
- Seven stages, held by the database. Pre-Award, Application/Review, Award, Implementation, Monitoring & Reporting, Closeout and Post-Closeout. The order is a constraint on the table, and the test suite reads that constraint back out of the schema and compares — so the list in the code and the list in the database cannot drift apart.
- The reporting calendar exists before anyone builds it. Confirming an award’s terms generates its reporting obligations in the same transaction, walking periods forward through the period of performance so every funded day is covered exactly once and no period reaches past the money. An uneven term leaves the last period short, deliberately.
- Closeout is one structured result, not a screen. Whether an award is ready to close is computed in one place and used by the pages, the reports and the transition gate alike: an award cannot enter closeout while a reporting obligation is still open, and an award whose outcome is already final cannot be advanced at all.
- A shared schedule that merges nothing. Upcoming deadlines are read across both halves by one core query, precisely because the tables it reads belong to neither.
What it refuses
No account gets a password the platform chose: the command that creates the first user has no password option, prompts for one, and the function under it takes the password as a keyword argument with no default. The destructive demo reset needs two independent yeses — an explicit opt-in and a marker written into the database’s own header — so an opted-in laptop pointed at a copy of production still refuses. The application ships no JavaScript and serves none, and every rendered page is checked against axe-core by the test suite, which fails rather than skips when its tooling is missing. It runs offline: network-capable connectors are opt-in and explicit, the database and its files are local, and backing it up is copying a directory.
Where it stands
Pre-release. The application runs, the suite exercises it on every change, and each wave of the consolidation carries a written record naming the command that proves each claim and the part of the claim the evidence does not reach. No customer data has been migrated.
There is no dated release to take a month from, so the stanza on the homepage roster reads available and stops there rather than inventing one. To use it for grant work you are running, get in touch.
Documentation
Grant Platform’s manual ships with the platform rather than on this site: the operator guide for standing an install up, the developer guide for how a module declares itself and where row visibility is decided, the security architecture with its own “what is not done” section, and the architecture decision records behind the choices above. The repository is private, so those documents come with an engagement — ask and they are yours to read before you commit to anything.