Biotech out-licensing

A partner search now starts
with one paragraph.

Business development teams screen with the same engines everyone else uses. If your asset is described three different ways in public, the paragraph names someone whose record agrees with itself.

The short answer

The screen happens before the first email.

Typed by someone on a BD team Who has a clinical-stage oral candidate in this indication, and what stage is each one at?

Three or four companies get named. Being the fifth-best documented programme is worth nothing here, and documentation is not the same as data: a strong asset with a contradictory public record loses to a weaker one described the same way everywhere. That is the part you can fix without generating a single new data point.

It is also why out-licensing is the first programme we built. One partner conversation is worth more than a year of traffic, which makes the usual marketing arithmetic irrelevant and the accuracy of one paragraph decisive.

Where it reads you

Six records, and you control two.

Roughly in the order of how much weight they carry. The last one is the reason a stale fact outlives its correction.

Trial registries A registry entry on ClinicalTrials.gov or its equivalents is public, dated and structured, which makes it the record an engine is most comfortable repeating — including the parts you last edited two years ago.
Your pipeline and asset pages The one source you fully control, and the one engines corroborate rather than trust. Necessary, never sufficient.
Designation and financing announcements Dated, quotable and permanent. A release describing an earlier stage is still being cited long after the programme moved on.
Conference abstracts Often the most precise description of the mechanism that exists in public, and frequently the one that disagrees with the website.
Partner and investor communications Written by someone else, about your asset, with their naming conventions. A common source of a second name for one molecule.
Third-party pipeline databases They copy from the five above, then get copied from in turn. This is the propagation layer: correct the primary source and these lag, sometimes by quarters.
What this means Most of the available gain is off your own domain, in records that take a human request and weeks of waiting. There is no API for it, which is exactly why it stays undone.

What costs the mention

Four disagreements, in the order we find them.

None of these is a mistake anyone made carelessly. They are what happens when a programme runs for four years and six functions describe it.

01 Two names for one asset

An internal code, a renamed code after a milestone, and a generic name appearing early. Each spelling splits the evidence into a weaker pile, and no pile is strong enough to be named.

02 The mechanism, worded three ways

The abstract, the website and the registry each describe the target and modality differently. An engine asked about a mechanism matches the words, not the biology.

03 A stage that moved

The registry says one phase, the deck says another, the last release predates both. Stage is the field a BD screen filters on first, so this one costs you the whole question.

04 An indication that drifted

Broad in the release, narrow in the protocol, differently worded in the database. The buyer's question uses one of those wordings and you are matched on it or not.

What we do

In this order, and not a different one.

One approved set of facts

01
The work One wording for the mechanism, one name per asset, one stage per programme, one indication phrasing — approved by you fact by fact, versioned, and reviewed by whoever owns disclosure. Nothing downstream can use an unapproved fact.
Why first Every correction after this points at the same target. Doing it second means filing the same request twice.

The diagnostic: who disagrees with whom

02
The work Collect every public description of the company and the lead asset, line them up against the approved set, and rank the disagreements by how much each one costs across the buyer questions that matter.
What you get A list where every line links to the page it was read from, with a date. This part is free, and most of the value is in the ranking rather than the list.

Correction, then pages

03
The work Correction requests to the records that carry weight, each one approved before it is sent. Then pages for the buyer questions that have no good answer anywhere — written to be cited, in sentences that survive being lifted out.
The order matters A new page arguing against a registry entry loses to the registry. Fix the record first and the page is corroborated rather than contradicted.

Measurement with a fifth held back

04
The work A panel of BD-shaped questions, run weekly against the engines, with twenty percent excluded from every pipeline so the lift has something to be measured against.
Why it belongs here A competitor's readout can lift your numbers on its own. Without a control group you would bill that to us. The method, and its limits →

Where we stop

The three your legal and IR teams will ask about.

Disclosure in this industry is not a marketing decision, and we are not structured to make one.

Never

Say anything you have not approved

Every claim resolves to a fact you signed or a public URL. No forward-looking characterisation, no inferred stage, no rounded-up efficacy language. A sentence without a source is blocked before anyone reads it.

Never

Publish or send without a named approval

Pages, registry corrections and emails all require an approval record naming a person. It is enforced in the code, so there is no configuration that turns it off.

Never

Trade on a partner's brand

Material written for a specific company carries your brand only, never theirs, and is never left open to search. Every claim we make about them links to a public page, with the date we read it.

And the obvious one: we do not invent data, trial results, interest from a named partner, or a quote from anybody. There are no customer logos on this site for the same reason.

Questions

What BD teams ask next.

Which public sources do answer engines read about a drug pipeline?

In rough order of how much weight they carry: trial registry entries, your own pipeline and asset pages, regulatory designation and financing announcements, conference abstracts, partner and investor press releases, and third-party pipeline databases that copy from all of the above. The copies are why one stale primary source propagates for years.

We cannot disclose the target. Can this still work?

Yes, and it is common. An undisclosed target is a disclosure decision, not a gap in the record, and engines handle it fine as long as everything around it is consistent: indication, modality, route, stage and the precise scope of what is undisclosed. What costs you the mention is one source naming a target that another source omits.

Is this different from writing a better pipeline page?

A better pipeline page is one source. An engine is weighing yours against a registry entry, a two-year-old release and a database row that copied an older version of your page. If those disagree, a rewrite of your own page does not resolve it. The work is making the other records agree.

Does this apply to preclinical assets?

It applies, with less to work with. A preclinical asset often has no registry entry, so the public record is your pages, any abstract, and whatever a database inferred. That is a thinner record to correct but also a less crowded answer, and the questions worth competing on are mechanism and modality rather than stage.

Related reading

The rest of it.