The work, and the standard
Solar Analytica does one thing, and this page governs it: it publishes data research. Reports that read an environmental or market variable against the record, indices that track a market quarter by quarter, and projections built on both. The discipline underneath all of it is a single principle. No forecast without data, and every figure named to its source. Public report figures use named, publicly accessible records unless a licensed input is explicitly identified. Each figure states its source class, observation window and limits rather than leaving them for the reader to discover.
The substrate
The reports and indices draw on one live data layer, assembled primarily from public records and refreshed as each pillar is verified. The source registry also retains licensed commercial references where they are useful for validation or bankable work; their presence does not make a public figure independently reproducible, so any figure that uses one must say so:
- NASA. MERRA-2 reanalysis (temperature, wind, surface irradiance), CERES satellite radiometry (irradiance), GISTEMP v4 (global surface temperature), and POWER (wind and beam-diffuse climatology).
- NOAA and partner agencies. The Oceanic Niño Index for ENSO, the Australian Bureau of Meteorology's Dipole Mode Index and Southern Annular Mode, ocean heat content, sea-ice extent, and the GPCP global precipitation record.
- Energy and markets. Ember's Global Electricity Review for the generation mix, AEMO's market and dispatch data for the National Electricity Market, and residential retail electricity tariffs across 92 cities.
These feed a 40-market yield baseline (modelled specific yield for PERC, TOPCon and HJT modules at each location, joined to the climate and price records market by market) and the market datasets behind our indices. Each report is a reading of that substrate, and each pillar it rests on is published alongside the report that uses it. The method is visible end to end in the heat penalty and the TBx Capture Index.
Source status and vintage
Observation year, publication status and retrieval date are separate fields. The current 2025 Grid Mix cohort is conservatively labelled provisional: live Ember rows are identified as a provisional provider annual series, while the fixed Our World in Data fallback is identified as a provisional published snapshot. A row persisted in the versioned ledger identifies its data version separately. The Grid Mix Monitor exposes these fields in the interface and export. A retrieval date proves when Solar Analytica checked a source, not when the publisher finalised the underlying observations.
The Market Opportunity Index
The Market Opportunity Scanner is a relative research screen, not a forecast. Its 0-100 scores combine four observed dimensions: storage need, solar momentum, flexible-load value and grid-transition pressure. The overall score weights those dimensions at 30%, 27%, 20% and 23%, then applies a modest evidence-confidence adjustment. National markets use Ember annual generation, a 2016-2025 transition record whose current-year source status is disclosed in the Grid Mix Monitor, and Global Energy Monitor's operating and prospective solar capacity. Operating regions add AEMO or Elexon daily demand, price, ramp and generation signals. Each dossier exposes the inputs, transformations, source window and contribution behind its score.
Scores are comparable signals within the published cohort. They do not estimate revenue, project viability, policy durability or investment return. A high score means the public record supports deeper investigation; it does not mean an opportunity is investable.
The Revenue Opportunity Lab
The Revenue Opportunity Lab is a scenario screen, not a dispatch forecast or valuation. Its NEM battery model calculates annual TB1, TB2 and TB4 perfect-foresight spread indices from complete calendar years in AEMO's five-minute regional price record. P10, P50 and P90 are linearly interpolated across the complete annual observations, then combined with the lower quartile, median or upper quartile duration-matched capture ratio from full-quarter operating batteries. Availability and round-trip efficiency are explicit user inputs. Flexible-load scenarios use the same historical regimes with a declared capture assumption. Solar and behind-the-meter scenarios are assumption-led: output, capture price, program value, capex and fixed operating cost are entered or visibly defaulted rather than inferred from an unavailable contract.
The output is a gross annual opportunity range and a simple capex-to-operating-margin payback. It excludes FCAS, network services, congestion, marginal loss factors, degradation, financing, tax and contract-specific obligations. A three-year distribution is materially stronger than annualising one quarter but remains a short history, so each output retains the contributing years and the current year is shown as YTD rather than admitted to the annual range.
The Revenue Shape Monitor
The NEM merchant-risk score combines three observed properties of the latest complete-year and multi-year spread record: 40% top-10-day concentration, 35% P10-to-P90 annual dispersion relative to P50, and 25% monthly coefficient of variation. Each component is capped at 100 before weighting. The score describes concentration and volatility in the perfect-foresight spread index; it does not describe counterparty, construction, financing or total project risk.
The SWIS flexibility-pressure score is deliberately non-monetary. Each month is ranked within the observed SWIS record for lower minimum operational demand, higher rooftop-PV share and higher evening ramp, weighted 35%, 30% and 35%. The global transition-pressure score is also relative: 50% current variable-renewable rank, 30% five-year variable-renewable growth rank and 20% carbon-intensity rank across the monitored Ember systems. Neither proxy is presented as battery revenue.
How a report is built
- Pull. Figures come from the named source at its own resolution. Nothing is smoothed and no points are invented; a plotted value exists in the underlying record or it does not appear.
- Verify. Every dataset is reconciled against an independent check before a word is written: a regional-average table, a second instrument, a second implementation of the same calculation, or the source's own published totals. A load that does not match the check is not published.
- Chart and cite. Each figure is drawn to one grammar and sits with the data table it summarises, and every report closes with its sources named and linked.
How the record disciplines projections
Solar Analytica publishes projections, but never predictions of the confident single-number kind. The record governs the forward view three ways.
- Scenarios are aggregated, and the disagreement is kept. A projection is built from published roadmaps rather than invented: the 2030 monocrystalline-silicon model aggregates ITRPV, CPIA, Exawatt and NREL into low, base and high paths. Where the roadmaps disagree, the disagreement is preserved as the spread. A range with each endpoint traced to the source that argues for it is harder to write than one confident number, and considerably more useful to plan against.
- Benchmarks are preferred to predictions. Where the record allows it, we publish the measuring stick rather than the guess. The TBx Capture Index does not forecast battery revenue; it measures what every NEM battery actually earned against a perfect-foresight benchmark computed from the same public price record, so anyone's forecast, or contract, can be marked against what happened.
- Projections stay on the record to be marked. A forecast nobody returns to is an opinion with a date on it. Our published projections remain in place as the data arrives to settle them: the monosilicon model names the roadmap projections observed data has already overtaken, including figures we would otherwise have relied on. When a method changes it is versioned, logged in the changelog, and every figure it touched is re-run, so a moved number is attributable to a data change or a method change, never to a quiet revision.
The register of these claims and markings — roadmap projections, program targets, and our own scenario endpoints, each dated, sourced and marked as the data settles it — is public at the Forecast Ledger.
The claim underneath is deliberately modest. We do not claim to see the future more clearly. We keep the record that makes claims about the future accountable, ours included.
Data standards
These are the rules that make the research reproducible and hard to spin. They are applied to every report and index.
- Source class named per figure. Public figures default to publicly accessible records. Any licensed or commercially restricted input is labelled, and reproducibility claims are limited to the records a reader can actually retrieve.
- Blanks over guesses. Where no credible figure exists, the cell is left blank and the reason recorded. A blank is a finding; a fabricated number is a liability.
- Ranges at midpoints, flagged. A source that gives a range is recorded at its midpoint with the range disclosed, and a bound is labelled as a bound rather than dressed up as a central estimate.
- Proxies labelled. A figure transferred from a nearby station is marked as a regional proxy, never presented as if it were measured on site.
- No marketing figures. Vendor and sales-blog numbers are excluded in favour of national agencies, grid operators, peer-reviewed work and reinsurance data.
- Quantities kept apart. Deposition is not mixed with atmospheric attenuation; an irradiance-basis figure is never averaged with an energy-basis one; an episodic event ledger is never annualised into a steady rate.
- Modelled is labelled. Where a figure is literature-derived or the output of a model rather than measured in our own data, it is flagged as such in the caption, the table and the method note, so a reader never mistakes a representative band for a site-specific result.
- No stake in the result. Solar Analytica takes no payment for coverage, placement or conclusions, and holds no position in any asset, market or company it writes about. Where a report names a real asset or counterparty, it states facts from the public record and draws no conclusion about any party's private position.
How the figures are drawn
Every chart follows one grammar: the context in grey, the single reading the story is about in one accent colour, and the data table kept with the figure as the auditable record and the accessibility fallback. The chart draws the eye; the table keeps the receipts. We never remove a table to make room for its chart.
Corrections
When a figure is found to be wrong, it is corrected in place and the change is dated, not quietly overwritten. Indices are republished each period with the structure held constant and the numbers moved, so the record, not a single edition, does the talking.
Product scoring lives on review.solar
Solar Analytica reads markets and the climate; it does not score consumer products. The panel, battery and inverter scorecards, and the fixed rubric behind them, are published on our sister property review.solar, which maintains its own methodology on its own domain. The two hold to the same reproducibility standard, applied to different sources: public records here, manufacturer datasheets there.