How the data is built.

The Deal Intelligence database is a structured record of roughly 1,284 pharma and biotech business-development transactions — acquisitions, licensing deals, collaborations, and option agreements — spanning the early 1990s through 2026.

Each deal carries three things: an Announcement Sentiment score that captures how the market reacted when the deal was struck, an Outcome Score that grades how the deal actually aged once results were in, and a body of sourced facts — deal value, structure, assets, revenue, and milestones — each tied to a traceable source URL.

Announcement Sentiment is a 0–100 reading of how the street received a deal at the moment it was announced. It is computed only when there are at least four sentiment-tagged sources — analyst notes, trade-press coverage, and primary press releases — each individually tagged Bullish, Neutral, or Bearish. Until a deal clears that four-source threshold, its sentiment is shown as pending rather than guessed.

This metric measures the market's reaction, not our opinion. It is a mirror held up to the analyst and trade community as the deal landed. We surface the underlying reviews so you can read the consensus yourself.

One honest caveat: announcement-day coverage skews optimistic. Press releases are written by the dealmakers, and sell-side notes near a signing tend to extend the benefit of the doubt. A high Announcement Sentiment means the deal sounded good on day one — not that it was good. That is precisely what the Outcome Score exists to test.

Scores are banded into verbal tiers — Exceptional (90+), Strong (75–89), Adequate (60–74), Weak (40–59), and Failed (below 40).

The Outcome Score grades how a deal actually aged. It is assessed at 5-, 10-, and 15-year anniversaries of the deal, each window backed by sourced facts — what was reported, approved, filed, or written down. The score unlocks five years after close; before then there is not yet enough verified history to grade fairly, so the deal shows no outcome.

The score is a weighted blend of four dimensions, then adjusted by a deal-difficulty factor:

Those base weights shift by the deal's lifecycle stage — pre-clinical, Phase II/III, launched product, platform, or licensing — so a pre-clinical acquisition is judged mostly on whether the science worked, while a launched-product deal is judged mostly on the money. The weighted total is then multiplied by a deal-difficulty factor (0.8–1.2): pulling off a hard, complex integration counts for more than coasting through an easy one.

Outcome tiers read as plain verdicts — Outperformed, Met Thesis, Tracking, Underperformed, and Failed.

Facts are researched primary-source-first. Higher tiers must be exhausted before falling to lower ones, and when two sources conflict, the higher tier wins. Different facts demand different minimum tiers — deal value and close date must come from an SEC filing; pipeline assets must come from a company's own pipeline page or its archived equivalent.

Field-to-tier minimums are enforced: revenue figures must come from a dated filing — never “analysts estimate $X” — and asset statuses are verified against Tier 1 or Tier 4 within six months of insertion.

Every record carries its own confidence level, derived from how many and how strong its sources are. This is what lets you decide how much weight to put on a given number.

Each value is also marked primary-sourced versus model-inferred, and timeline events carry a “Last verified” stamp. That stamp is the date a primary source was last successfully checked — pipeline data is re-verified on a 90-day cycle, clinical-trial status every 30 days, financials every 90, and sentiment every 180. If a refresh can't reach a primary source, the stamp is not advanced; the system never invents a timestamp to look fresher than it is.

We believe you can only trust a number if you know how it was made. So, plainly: