Public beta · Product guide

EvidenceLab

EvidenceLab helps people structure and compare business decisions using objectives, evidence, constraints and trade-offs. This guide is readable without JavaScript or authentication.

Open the interactive workspace · Plain-text product information

Choose your starting point

v0: guided start leads with a prepared example and a three-step outline for your own goal, possible actions and one limit. It requires no account or API key. Missing details can stay unknown; it does not invent a best action from incomplete inputs. Optional AI assistance is offered after the guided result and opens the existing visitor-key flow.

v1: detailed workspace preserves the existing five-stage investigation editor, tree map, evidence register, comparison, import/export and provider options. A v0 outline can be transferred into v1 after an explicit choice. Existing saved v1 work is protected by a replacement confirmation. No provider request is made by transferring an outline.

What the detailed workspace can do

  1. Define an objective, baseline, target, time horizon and success measure.
  2. Map influencing, controllable and affected variables. Membership can overlap; permission and controllability are independent.
  3. Record relationships, assumptions, fixed conditions, structured constraints and protected outcomes.
  4. Compare candidate actions using supplied costs, duration, effect estimates, ranges and evidence. Feasibility can be feasible, infeasible or unknown.
  5. Recommend a supported action conditionally when appropriate, or identify missing evidence when a reliable ranking is unavailable.

Try the example without a key

The synthetic SaaS scenario compares five interventions under fictional budget, headcount, satisfaction and capacity rules. Two are feasible and three are infeasible under those assumptions. All numerical estimates are fabricated demonstration inputs, not measured or validated business results.

Optional AI drafting

Manual input, evaluation and the synthetic example do not require an API key. For AI input, choose OpenAI, Gemini or Groq in Draft with AI. Visitors must supply their own provider key, approve sending the request, and are responsible for provider usage and any charges. EvidenceLab does not supply a shared AI key. The application server receives that key transiently. The app does not intentionally store or log keys; provider and infrastructure policies still apply. Generation sends the investigation to the selected provider; connection testing sends only the key for model metadata. Provider charges and quotas may apply.

Save your work

Work is session-only by default. Optional browser saving and JSON export/import preserve an investigation. Markdown exports provide a readable decision report. API keys are excluded from saved investigations and reports. There is no persistent server investigation database.

Optional v0 beta feedback

After working on your own decision in the guided beta, you can choose to send one answer about whether the comparison helped you find a next step. This is optional and does not affect the result. The sample does not ask for feedback.

EvidenceLab stores your selected answer, the submission date, a beta version and a record ID with a protected deletion code. The feedback request does not contain your investigation, name, contact details or API key. Responses are private to the project owner and are used to guide product development. Any published findings will use combined counts.

Keep your receipt if you want to remove your answer. Open the detailed workspace and use Manage past feedback at the bottom. Earlier questionnaire responses remain subject to their original notice.

Limits

The software does not independently verify sources, discover causal effects or guarantee a globally optimal decision. Natural-language resource descriptions must be entered as structured rules to be enforced. Automatic sprint scheduling, adaptive cognitive-capacity measurement and guided discovery are proposed future directions, not current features. Valid-key generation across every provider has not been independently verified in this beta.

Access and loading

The site is public and the example requires no login. The interactive workspace needs JavaScript. If it does not load, refresh the page or try another browser. A web-reading or LLM service may have its own network or JavaScript restrictions; availability to one tool does not establish availability to every tool.

Check service health · Return to EvidenceLab