The system of record for your evidence plan.
Keystone replaces gap spreadsheets and slideware with one traceable plan. Every evidence need is linked to the sources that raised it, the tactics that cover it, and the people who signed it off.
Evidence gaps
23 gaps from 14 sources · last ingest 2 hours ago
| ID | Gap | Sources | Coverage | Status | Priority |
|---|---|---|---|---|---|
| G-014 | Comparative effectiveness vs regional SoC, age ≥65Mapped: retrospective RWE chart review | 3 | Partial | Split first | |
| G-021 | Payer-relevant HCRU in EU5 marketsNo completed, ongoing or planned tactic | 5 | Open | High | |
| G-008 | Treatment sequencing after progressionMapped: prospective registry (shared with G-009, G-011) | 2 | Addressed | — | |
| G-017 | Quality of life on long-term therapyMapped: prospective registry PRO module | 4 | Addressed | — |
One plan, shared across the asset team
- Medical Affairs
- HEOR
- Real-World Evidence
- Market Access
- Clinical Development
- Regulatory
- Commercial
Spreadsheets can't tell you what's still unresolved.
Every asset team has to answer the same question: what do we need to know, what evidence exists, and which open questions matter most? Static gap trackers get it wrong in four predictable ways.
Opinions become gaps
A stakeholder comment goes straight onto the tracker as a validated gap, with no test of whether it's real.
Keystone: every comment starts as a candidate need.A study existing means "closed"
A chart review with no comparator gets marked done, even though the HTA question was comparative.
Keystone: coverage is scored on ten dimensions.One study, four copies
A registry covering sequencing, HCRU and quality of life gets pasted onto three rows that drift apart.
Keystone: one record, linked to every gap it covers.Coverage confused with priority
A mostly covered HTA question can still be the most important open item. Trackers can't express that.
Keystone: priority is a separate, signed-off decision.From source documents to a signed-off plan.
Keystone drafts and computes. Your team reviews, decides and signs off. New sources flow back through the same four stages, so the plan stays current between planning cycles.
Upload
Ingest advisory board notes, KOL interviews, HTA feedback, protocols and publications.
- Extracts candidate needs
- Maps existing tactics
Gaps
Each gap gets a computed status: Open, Partially Addressed or Addressed. Reviewers validate every one.
- Partial gaps must be split
- Overrides need a reason
Prioritize
Place open gaps on a configurable matrix and set High, Medium or Low.
- Set by people, never the engine
- Separate from coverage
Tactics
Propose studies and dissemination tactics for high-priority gaps and lay them out on a timeline.
- Interactive Gantt view
- Save and export the final plan
A partial answer stays partial.
"We don't have enough evidence in elderly patients."
- Keystone records this as a candidate need and links it to a named gap.
- It maps the tactic that already exists: a retrospective real-world study in patients aged 65 and over.
- Coverage is scored by dimension. The comparator is missing, so the gap is Partially Addressed.
- The reviewer splits it. The original moves to version history and a new Open gap is created for the comparator question.
- HEOR sets that gap to High ahead of the HTA submission, with a recorded reason.
| Dimension | Covered |
|---|---|
| Population | Yes |
| Intervention | Yes |
| Comparator | No |
| Outcomes | Yes |
| Decision utility | No |
| Overall | Partial |
Built for teams that have to defend their decisions.
The engine proposes. People decide. Every decision is recorded with who made it, their function and why.
Sign-off at every gate
Nothing is closed or prioritized automatically. A named reviewer validates each gap and records their function.
Complete version history
Split or rewritten gaps are retired, never overwritten. Every child traces back to its parent.
Linked, not copied
Each need is stored once. One registry can cover several gaps without duplicate rows going out of sync.
Ten-dimension coverage
Population, comparator, outcomes, decision utility and more, each assessed, with an overall degree you can audit.
Conflicts are flagged
When new evidence disagrees with a reviewer's override, Keystone marks it stale instead of silently replacing it.
Evaluated against gold
Each pipeline stage is measured against reference cases, with run traces you can inspect.
Modular, observable, model-agnostic.
Each stage, from upload to timeline, is an independent module with its own contract, monitoring and evaluation. Swap models without rebuilding the plan.
Pipeline stages
Each one versioned, traced and runnable on its own or as a chain.
Model providers
Claude, GPT, Gemini, Grok and OpenRouter, routed per stage.
Models required
Without a connected model, deterministic rules run the whole pipeline.
Traceable chain
Source → need → gap → tactic → coverage → priority → roadmap.
See your evidence plan the way an HTA reviewer will.
We'll walk you through Keystone on a sample oncology asset, from source documents to a signed-off, exportable plan.
- A 30-minute walkthrough with the product team
- Bring a current gap list and we'll show you where the open questions are
- Demo data is fictional. No patient or payer data is used.
Request received
Thanks. We'll reply within two business days to schedule your walkthrough.