Matching works reliably only when it is done in two stages: AI filters the possible counterparties out of a pile of registrations and explains why they match; people talk through "why it does not match well enough yet" and put the authorisation scope, purpose of use and delivery method down on paper.
Matching is not "recommend because the keywords are the same". It looks at business dimensions and compliance dimensions at the same time — a pairing that matches perfectly on business but cannot be implemented on compliance wastes time for both sides.
A list of candidate counterparties ranked by fit score, each with matching reasons and a gap explanation, making it easy to judge quickly which are worth pursuing.
Point out exactly where the gap is: insufficient field granularity, an inadequate time span, a mismatched authorisation scope, or a missing type of compliance material.
For feasible pairings, it gives possible delivery methods and points to note, for consultants to confirm further during human follow-up.
Every dimension has three verdicts — "match / partial match / no match" — and a partial match comes with suggestions for alternatives.
Two things are most feared in matching: one is "recommended but unable to explain why", the other is "negotiations fail without knowing why". The former goes to AI, the latter to people.
The workflow is designed so that "you always know what the next step is": the input, output and owner of every stage are clear.
Turn the free text registered by both sides into structured dimension descriptions and fill in missing key items (such as use scenario and field list). Output: a standardised registration file.
Check how well defined the scenario is and how clear its boundaries are to judge whether the preconditions for progress exist; data types requiring separate assessment are flagged in advance. Output: a screening conclusion.
Calculate and rank across six dimensions, generating a candidate list with matching reasons and flagging the specific gaps of a "partial match". Output: a match list.
We speak with each side separately to verify that the need is genuine, that the resource is genuinely available and that the compliance conditions hold, adjusting the plan or widening the candidate pool where necessary. Deliverable: a follow-up plan.
Choose the delivery method according to the data tier and complete delivery in a secure, trusted environment, producing a record of the transaction and delivery process. Output: delivery records and supporting documents.
Choose the delivery method according to data sensitivity and use scenario. Most medical data scenarios prefer "usable but not visible" — the demand side gets the results while the data holder retains control.
The demand side completes analysis or modelling in a controlled sandbox environment: it can use the data and obtain results, but cannot export or take away the raw data. Suitable for scenarios of higher sensitivity with a clearly defined purpose of use.
This is the most common delivery form in medical data scenarios: the demand side gains the analytical value while the data holder retains control over the data.
Providing specific query, verification or computation capability as an interface, with metered call volumes, verifiable purposes of use and permissions automatically revoked when the agreement expires.
The interface approach turns "data delivery" into "capability calls" — an effective way to reduce the risk of data flowing out.
Delivering capability as a disease-specific model or a packaged agent, supporting controlled service calls or private deployment and meeting the requirements of a domestic IT environment and of data not leaving its domain.
Model-based delivery must also state the scope of the training data and the boundaries of application, so as to avoid out-of-scope use.
Delivered as a dataset after the compliance assessment, authorisation confirmation and masking have been completed. Suitable where sensitivity is manageable, the scope of use is clear and a complete compliance chain is already in place.
Dataset delivery has the highest threshold, so it is better suited to being part of a long-term collaboration rather than a one-off action.
* The delivery method must match the sensitivity of the data, the authorisation scope and regulatory requirements. Where human genetic resources or highly sensitive personal information are involved, the corresponding procedures must be completed as required before proceeding; the specific method is subject to the agreement between the parties and the relevant regulatory requirements.
Consultants first speak with each side separately to confirm that the demand is genuine and the conditions feasible, then arrange contact according to the wishes of both parties. This avoids wasted effort when "direct contact reveals that it is simply not suitable", and makes it easier to clarify the authorisation scope and compliance conditions at an early stage.
Consultants will report where the gaps lie and offer optional adjustment paths: widening or narrowing the candidate pool, adjusting field granularity requirements, switching the delivery method, or first completing a certain type of compliance material. If they judge that the market genuinely lacks a match, they will say so honestly rather than force a recommendation.
Prices are agreed between the two parties to the transaction; the platform does not set prices unilaterally. Consultants can help both sides clarify the pricing logic and the key variables that affect price (data scope, update frequency, scope and term of use, exclusivity and so on), giving the negotiation a common basis.
This covers registration and initial-screening records, the matching conclusion, the key confirmations made during matching and follow-up, records of the delivery method and data destination, and any supporting documents required. The purpose of keeping a trail is to be able to answer where the data came from, where it went, and the scope within which it was used.
Yes. At registration you can list in the notes any competitive relationships or existing cooperation constraints to be avoided; matching and follow-up will exclude the corresponding counterparties accordingly, and the information is used only for internal judgement within this process.
Register your resources or needs; AI narrows the field first, then a professional consultant makes the plan clear. We will confirm the next step with you after the initial screening.
