How Forum Forecast works

Forum Forecast predicts which agencies are likely to release RFPs in the next 90 days, with confidence scores grounded in observed signals. It is the platform's predictive layer — distinct from the synthesized Brief, which is descriptive.

What it predicts

A Forecast prediction is a cluster of related signals pointing at a probable forthcoming RFP. Each prediction carries:

• The target agency (e.g., HHS, CMS, NIH)
• The signal types feeding the prediction (pre-RFP notice, sources-sought, RFI, budget allocation, market research)
• A horizon estimate (30 / 60 / 90 days)
• A confidence score (0.0 — 1.0)
• A rationale paragraph explaining the cluster

It does not predict the dollar value, vehicle, or specific scope of the RFP — only that one is probable.

Signal types

Five signal types feed the forecaster:

• Pre-RFP notices. A pre-solicitation notice on SAM.gov.
• Sources-sought. A sources-sought notice asking for capability statements.
• RFI. A Request for Information.
• Budget allocation. A line item appearing in the OMB budget or a Congressional appropriation that maps to a workstream.
• Market research. A vendor outreach or industry day announcement.

The classifier (a logistic regression over historical signal-to-RFP outcomes) weights signals by historical materialization rate per agency.

Materialization

A prediction "materializes" when an RFP matching the predicted cluster appears on SAM.gov within the horizon window. Materialization rates feed back into the classifier — agencies that consistently materialize at high rates get higher base confidence; agencies that consistently fail to materialize get penalized.

The materialization log is visible on the Forecast view's "Materializations" tab. It is the platform's honesty mechanism — if the forecaster is wrong, the dashboard shows the misses.

When it runs

The forecaster recompiles nightly at 02:00 ET. Predictions older than 90 days are retired automatically; new signals trigger new clusters or update confidence on existing clusters.

You can re-trigger an ad-hoc recompile from the Forecast view's settings menu — Enterprise+ only.

How it's used

Three primary use cases:

1. Pipeline-fed-by-signal. Capture leads pin high-confidence predictions to a case at /cases and start accreting evidence weeks before the RFP drops.
2. Watch-driven alerting. Watches (see Watches) trigger when a new prediction matches an agency + keyword filter.
3. Digest. The twice-monthly Forecast digest emails the workspace's top-10 predictions ranked by confidence (Team+ tier).