Platform
Four things, on a loop, until the gate closes.
Forecast the distribution. Optimise the position against it. Trade it. Prove what it earned. Then do it again, because the volatility just moved the answer.
The loop
Each pass narrows the range you're trading into.
Forecast
Probability, not a single number
Price and volume forecasts arrive as distributions with explicit confidence bands, refreshed as new market and process data lands. A point forecast tells you what to expect; a distribution tells you what to risk.
Forecast
Rolling, not one-shot
The optimiser revisits your position every refresh against physical limits — state of charge, ramp rates, heat demand, process minimums — and decides whether the better move is to trade or to sit.
Forecast
Execution into thin books
Orders are worked into continuous intraday liquidity with the spread and remaining time to gate closure in view, so a good idea doesn't become a bad fill.
Forecast
Attribution against doing nothing
Every delivery hour is measured against a baseline where you simply held the day-ahead schedule. That's the number worth arguing about internally.
Why Distributions
A point forecast can't tell you how much to bet.
If the model says €78/MWh, you still don't know whether that's €78 give or take three, or give or take ninety. Those are opposite trades.
Sizing
The band sets the position
A wide distribution argues for holding volume back and trading later; a tight one argues for committing now. The same expected price supports both, depending on the spread around it.
Risk
Tail exposure is explicit
Imbalance costs live in the tails, not the mean. Quantifying the tail is what lets you decide how much unhedged volume you're comfortable carrying into delivery.
Updating
New information changes the odds, not just the number
As gate closure approaches, the range collapses. The trades worth making at Gate Closure 240 are rarely the trades worth making at Gate Closure 20.
Two Ways to Run It
Your desk, or ours.
Same platform underneath. The difference is who watches the screen at 02:00 on a Sunday.
Software
Your traders, our tooling
You keep the market access, the BRP relationship and the decisions. We supply forecasts, optimisation and execution tooling that plug into how your desk already works.
Managed
We operate the market function
For teams who don't want to build a 24/7 desk, we run it. You offload the technical and market complexity and keep the upside — with the same attribution reporting either way.
The Engine
AI-native, because the job is impossible by hand.
This isn't a spreadsheet workflow with a model bolted on the side. The decision volume alone rules out a human process — and that's for a single asset.
96
DECISIONS PER ASSET PER DELIVERY DAY, ON 15-MINUTE PRODUCTS
1,000s
DECISIONS PER ASSET PER DELIVERY DAY, ON 15-MINUTE PRODUCTS
7
MARKET VENUES IN SCOPE: DA, ID, FCR-D, FCR-N, FFR, AFRR, MFRR
24/7
UNATTENDED OPERATION THROUGH GATE CLOSURE, EVERY HOUR OF THE YEAR
Core
Stochastic optimisation, not rules
Every bidding strategy is tested against thousands of price and production scenarios, and the decision chosen is the one with the best expected value at controlled risk — not the one that only works if the forecast happens to land.
Architecture
Cloud-native, one engine, many assets
New assets onboard as configuration — physical constraints, market access, risk limits — against the same forecasting and optimisation core. We're not maintaining a fork per customer, which is what makes the marginal customer cheap to serve.
Portability
The method isn't Nordic-specific
The probabilistic core has already been ported to a second, structurally different market. Market rules live at the edge of the system; the decision engine underneath doesn't change.
Integration
No role changes required
We slot in above existing market access. Customers keep their balance responsibility and their existing counterparties — which removes the collateral, contractual and organisational objections that usually stall a pilot.