Auditable grid forecast infrastructure for U.S. power markets.
Gramm pairs forecast delivery with an ISO-reference evaluation harness. The current production model is under active replacement; historical low-MAPE backtests are preserved as research archive, not presented as live product performance.
Production h24 MAPE is currently behind ISO references on every measured grid. Candidate models must improve this table in shadow mode before promotion.
Evaluation path
Inspect the evidence, then test the path.
The homepage starts with the live scorecard because production quality has to be visible before anyone evaluates an API feed. From there, the next step is a scoped benchmark or a free key for payload testing.
- 01
Review live scorecard
Start with the current h24 table, including gaps to ISO references and pending reference rows.
Open scorecard - 02
Request scoped benchmark
Declare grid, horizon, audit period, and decision context before any comparison is scored.
Request benchmark - 03
Test API payloads
Create a free key or inspect the docs, then test responses against your integration shape.
View API docs - 04
Move to procurement or paid plan
Use the trust, SLA, pricing, and enterprise pages when the benchmark is ready for review.
Open procurement
Evidence first, then promotion.
The old website claimed ISO-beating performance. The current evidence does not support that claim. Gramm now separates live production scores, historical research results, and shadow candidates so customers can see exactly what is real.
Live production scorecard
h24 MAPE is computed against paired actuals and ISO references where available. The scorecard is the public truth for production quality.
Research archive
Prior 1-3% MAPE backtests remain documented, but they are not marketed as current product performance until reproduced live.
Shadow promotion gate
Replacement models must pass leak-free validation and live shadow checks before they can replace production checkpoints.
Forecast API, measured by the same evidence standard.
Bearer token auth, JSON response, and forecast delivery are useful only if the scorecard is honest. The API surface stays separate from research claims so downstream teams can decide when the model is ready for operational use.
View API docsHistorical research archive.
These were the older backtest figures that originally motivated the product direction. They remain useful for research triage, but the live scorecard above controls public production claims.
Run the same scorecard on your market.
Bring your operating period, grid, and horizon. Gramm will benchmark against the relevant ISO reference and show the paired rows, not just a blended percentage.