Ormeval provides decision-support intelligence derived from public sources. It is not investment, medical, legal or regulatory advice, and states nothing about the value of any security. Forward figures are deterministic model estimates, not guidance.
Ormeval · Pharma intelligence · 28 SEP 2026
Methodology
Ormeval is built for people who have to defend a number in a meeting. So every figure on the platform is one of two things, a cited public fact or a flagged model estimate, and you can trace either back to its origin. Here is exactly how that works.
Not everything on this platform moves at the same speed, and saying “updated daily” across the board would be false. Trials and filings refresh from the public registries on a scheduled ingest, and each clock below shows when its data actually last landed. Reported revenue moves when companies report. The forecast models move when a person curates them, because our own rule forbids a machine inventing any epidemiology, price, share or revenue figure.
| Layer | Cadence | Current to | Age |
|---|---|---|---|
| Trials & approvalsClinicalTrials.gov records, openFDA approvals and label changes | Scheduled ingest | 2026-09-28 | 17d · behind schedule |
| SEC filingsSEC EDGAR 10-Q/10-K/8-K/20-F/6-K and the financials review queue | Scheduled ingest | 2026-09-28 | 17d · behind schedule |
| IR newscompany investor-relations announcements (link-only for non-SEC filers) | Scheduled ingest | 2026-08-30 | 46d · behind schedule |
| Market-moves queuedifferences between the freshly-ingested facts and our curated state | Scheduled ingest | 2026-08-30 | 46d · behind schedule |
| Forecast modelsper-country funnels, market dynamics, competitor shares, HTA and IP | When curated | 2026-04-29 | 169d |
These dates are read from the data itself, not maintained by hand. If a layer falls behind, this table shows it rather than hiding it, a freshness badge somebody has to remember to update is a freshness badge that goes stale silently.
Every automated update runs on the fixed schedule below, read directly from the scheduler configuration in the repository, so this list cannot drift from what runs. Nothing on this platform publishes or updates a figure in reaction to a market event: an approval or a readout enters the data when the next scheduled pass picks it up from the public record, and curated model changes are made by a person and dated. The one page-level timer is the market wire, which re-renders hourly from already-ingested data.
Every catalyst carries its source, and a person checks it against that source. That costs a median of 12 days from the public event, 33 at the 90th percentile, across the 503 catalysts recorded in the last 60 days. Hard constraint 9 is why: no unattended process writes a curated file, so nothing reaches the record on a machine’s say-so alone. Where an operator-directed pass makes the entry, the check is queued in an open curation queue rather than skipped.
Our track record page computes, at build, the error of every modelled asset-year against the company-reported actual, splits it by regime, and shows where the engine is the better estimator (young, fast-moving, discontinuous franchises) and where a trailing extrapolation currently is (mature, steady ones). The table is the claim.
Cited actuals. Reported revenue, approvals, indications and milestones are transcribed from company investor-relations decks, SEC/EDGAR filings and regulator notices (FDA, and EMA records a person read and cited by link, never machine-extracted), each carrying a link back to its source. These are facts, shown as reported.
Model estimates. Forward-looking figures are produced by a deterministic forecast engine from the drug's public development stage and editable epidemiology, share and price assumptions. Anything beyond company guidance is labelled a model estimate and shown visually distinct from a cited fact, never blurred into one.
Company IR & SEC/EDGAR
Reported product revenue, guidance and milestone disclosures, transcribed from filings, each linked to source.
FDA
Approvals, indications and the current product label, used to establish what a drug is actually cleared to treat.
EMA
EMA and NICE records are read by a person and cited by link; they are never machine-extracted - approvals and label detail transcribed from a record a curator opened, not fetched by any script.
ClinicalTrials.gov (U.S. NLM, API v2)
Live trial records (phase, status, enrolment, primary-completion readouts) resolved across development code, INN and brand.
openFDA
Label-derived mechanism and indication text. Not a substitute for the official FDA label.
ClinicalTrials.gov. Trial data are courtesy of ClinicalTrials.gov, a service of the U.S. National Library of Medicine. Our copy is refreshed on the schedule shown above, and the date it was processed by ClinicalTrials.gov is displayed in the footer of every page and in the freshness table above. The authoritative and current record is ClinicalTrials.gov itself. Ormeval asserts no proprietary right in the database and does not represent it as anything other than a United States Government database. The NLM does not endorse Ormeval.
What we change about a registry record. We do modify the content, and their terms ask us to describe how. We resolve each record to a drug across development code, INN and brand, so one asset gathers the trials that several naming conventions scattered. We group related studies into programmes by indication and line of therapy. We map the registry's phase and status fields onto our own stage vocabulary. We select a subset - lead-sponsored, interventional, active Phase 2 and 3 - rather than mirroring the registry whole. We read dates from the record and use them as milestone dates only where the registry marks them actual. We add nothing to a record that the registry does not state, and we never edit a registry field's value; every derived judgement is our own layer beside the record, not a change to it.
openFDA and FDA data. Data provided by the U.S. Food and Drug Administration (https://open.fda.gov), published under a Creative Commons CC0 1.0 Universal dedication, which permits commercial use and redistribution. Attribution is requested rather than required; we give it. The FDA does not endorse Ormeval, and label-derived fields are not a substitute for the official FDA label.
SEC EDGAR. Filings are works of the United States government and carry no copyright in the United States. Our ingest declares a real, monitored contact address in its User-Agent and stays within SEC's stated fair-access rate guidance.
European and UK registers are read, never harvested. EMA, NICE and equivalent European and UK publications are cited by link from records a person read. No automated process extracts from them, by standing policy, because the EU database right protects registers made in the Union in a way it does not protect US federal ones.
Same inputs, same outputs: no randomness, no black box. The published Ormeval forecast is our own independent estimate from public data - read-only, like a consensus view you read rather than author. To pressure-test it, open any asset and change a single assumption (epidemiology, treatment rate, net price, peak share, erosion): every downstream figure recomputes in front of you as a private what-if, shown beside the unchanged Ormeval forecast, never blended into it. The engine (not a language model, and not us) owns every computed number. That is the rule the whole product is built on.
Some assets carry both an anchored outlook and a bottom-up build-up, and where the two diverge materially the page says so in a flagged card rather than letting you catch the platform contradicting itself. The anchored curve is pinned to cited public anchors - company-reported product revenue, which is worldwide - and is the figure of record for the asset. The build-up sums the modelled markets only (the US, the EU5, Japan and China), so it is expected to land somewhat below a worldwide total even when every model is right; across the comparable corpus the median build-up is 0.78× of the anchored peak, which is about what that scope difference alone would produce.
Beyond scope, the build-up is bottom-up from each indication's sourced epidemiology, funnel and price, multiplied by penetration and rival-share assumptions that carry no citation. So read a gap as scope first and curation backlog second - and where it runs to 2× or more, or in the other direction, scope cannot explain it and the build-up is the weaker of the two. Neither figure is ever rescaled to match the other: forcing the build-up onto the headline would manufacture per-indication numbers, and would delete the only signal that says a model needs curating.
A language model writes the plain-English weekly digest that summarises what moved. It never computes a figure, and it may only cite trial identifiers (NCT IDs) that already appear in the underlying source data. The numbers come from the deterministic engine; the model only narrates them.
Ormeval is independent and is not affiliated with, endorsed by, or sponsored by any company whose products it tracks; brand and company names are used for identification only. Everything here is decision support, not financial, investment, or medical advice. See the full disclaimer.
Method is a claim until it is checked. Every catalyst we publish is fixed in an append-only ledger the moment we publish it, and every modelled figure is measured against what companies later reported, including where we were furthest out.
Ormeval is independent. The operator holds no positions, long, short or derivative, in any company covered on this site, and does not trade the names it models.
We receive no compensation from any pharmaceutical company. No issuer pays for coverage, reviews a page before publication, or has any influence over a forecast, a citation or a competitive read. We provide no advisory, consulting or investor-relations services to the companies we cover, and the site carries no advertising and no sponsored placement.
Ormeval models drug franchises, not securities. We publish no price targets, no fair-value estimates, and no view on whether any share is cheap or expensive. That is a deliberate boundary, not an omission, the moment a revenue model is restated as a claim about a share price it becomes a different kind of document with a different set of rules. Nothing here is investment advice or a recommendation to buy, sell or hold anything.
If any of this changes, this section changes with it.