Where every number on this site comes from, and what it does not prove.
This page lists every source the site reads, and what each proves and does not prove. It states how a value is classed before it is shown, how the reference asset is defined, how the model dispatches it, and how a reader can reproduce any published figure. It is written so that a lender's technical adviser can check the site without asking.
Twelve sources, proven and limited
| Source | What it proves | What it does not prove | Cadence |
|---|---|---|---|
| ENTSO-E Transparency | day-ahead prices (A44), cross-border flows, installed capacity (A68) | balancing clearing prices, activation data | 4 h |
| Baltic Transparency Dashboard (BTD) | reserve procurement volumes, some clearing data | full auction results, all products | daily |
| Litgrid | Lithuanian grid capacity, connection queue, storage capacity and reservations | construction progress, COD timing | 4 h / manual |
| AST | Latvian installed storage, procurement contracts | merchant project data | manual, quarterly |
| Elering | Estonian storage, frequency-reserve costs, indicative connection capacity | firm connection rights | manual |
| VERT (ArcGIS) | Lithuanian permits, grid capacity | Latvian or Estonian permits, connection costs | 4 h |
| Nord Pool | day-ahead and intraday prices and volumes | balancing outcomes | 4 h |
| NVE | Nordic hydro reservoir levels | Baltic-specific hydro effect | weekly |
| energy-charts.info (Fraunhofer ISE) | TTF gas, EU ETS | Baltic-specific effect | 4 h |
| ECB SDW | Euribor, inflation | Baltic financing terms | daily |
| NREL Annual Technology Baseline | battery efficiency, cost trajectory, degradation calibration | Baltic costs | annual |
| Clean Horizon Storage Index | a European revenue benchmark for framing | observed Baltic clearing | semi-annual |
Transmission system operator documents, then regulator filings, then ENTSO-E, then developer releases naming an asset, then industry press. A source is never cited when one two rungs above it exists.
Observed, derived, proxy, modelled
- observed
- a value read directly from an authoritative source
- derived
- computed from observed values by a stated formula
- proxy
- taken from a related but not identical source, and flagged as such on the card
- modelled
- the output of a model, the lowest confidence for anything forward-looking.
A metric is published on the public pages only when its confidence grade, its distance from raw data and its coverage pass the gate; a metric that fails is shown only in a detail drawer with its class, or not at all. Reserve clearing-price estimates and project returns are model outputs several steps from raw data and are shown under scenario framing, never as observed values. A metric whose source is stale relative to its own cadence carries a stale flag.
The reference asset
Every economic figure on Signals and ROI refers to the same reference asset: 50 MW, at two-hour or four-hour duration, connecting in Lithuania, Latvia or Estonia. Its parameters are listed here with their sources and are the same parameters the report engine uses. The asset is not configurable on the public cards; the calculator on the ROI page runs the same model on the reader's own inputs.
Dispatch, period by period
The model dispatches the battery in 15-minute settlement periods against real historical price shapes, rather than applying an average spread to an assumed number of cycles. Each revenue stream has its own price formation: reserve capacity clears against a demand trajectory and a competing supply trajectory; activation is modelled per direction, because a battery charging on downward activation is earning, not spending; day-ahead arbitrage is limited by what the asset can physically do after its reserve commitments. Three decisions are worth stating, because they are where models of this kind usually go wrong. The same megawatt cannot be sold twice. Power committed to reserve is not available to trade, and the model enforces this rather than assuming it. Trading performance is measured, not assumed. The model applies a realisation factor derived from a multi-year backtest of achieved capture against a perfect-foresight benchmark, rather than an optimistic constant. Degradation is driven by throughput. Cycling is an output of the dispatch, and the wear curve consumes that output — so a case that trades harder ages faster, without anyone adjusting an assumption. Every parameter in the model carries a source, listed in the appendix beside the value. Where a parameter also carries a review cycle, it is recorded in the parameter set rather than reproduced here; not every parameter carries one.
Capacity fade is modelled from throughput: the dispatch produces a cycle count, and the wear curve consumes it. The curves available to us are characterised at one, one-and-a-half and two full cycles per day. The model does not extrapolate below the slowest of those. Where an asset is dispatched below that range — as reserve-led assets often are — the cycle count reported is an accounting output rather than a wear input, and the fade shown is the floor of the characterised range rather than a measured value for that duty.
The Index
For each country and duration the engine dispatches the reference asset against that month's observed prices and reserve clearing and reports gross revenue in €/kW/yr. A cell is published only when its inputs pass the gate in §3; until then it is a dash. The series is versioned: each point carries the engine version that produced it, and a recomputation under a later version is shown as such. The framing is comparable to Clean Horizon's Storage Index for Baltic markets; the values diverge, and the differences in assumptions are listed here.
Corrections, with their effect
Each correction is dated, with the change and its effect on the reference configuration; the list below is the same one that is a chapter in every report.
Reproducibility
Every figure in this report appears in the appendix with the model field it came from, the parameter set and version in force, and the identifier of the run that produced it. Each parameter carries its source; where that source names an edition or a year the parameter's vintage can be read from it, and where it does not the appendix says so rather than implying a currency it cannot demonstrate. The same inputs reproduce this report exactly. If a figure looks wrong, one lookup will show what produced it.
On the public cards the same rule appears as the footer under each number: the source, the time it was read and its class.