PiacRadar
Identity · hackathon case study
LLM market-intelligence workflow. One question becomes hypotheses, evidence, and an action plan.
Structured analysis / Qdrant evidence / workflow trace
Evidence boundary
What is demonstrated, and where the claim stops.
- Synthetic data
- Repository fact
What this proves
- /One strategic question can be transformed into a structured decision package.
- /Hypotheses, evidence references, actions, and validation tasks can share one typed workflow.
- /A cached demo path can remain inspectable when live services are unavailable.
What this does not prove
- /Causal attribution or a validated business recommendation.
- /Production evidence quality across all public sources.
- /Measured executive time savings or commercial impact.
Data boundary
Core company and market data for the FreshCart scenario is synthetic/generated. Retrieved public evidence is kept separate and varies in quality.
Evidence type
synthetic data / repository fact
Core company and market data is synthetic/generated; public evidence quality varies. Decision support, not causal proof.
Known limitations
- /The core FreshCart scenario uses synthetic/generated data.
- /Public evidence quality and scrape success vary.
- /The deterministic path is directional and the optional LLM path may not be active.
What production would require
- /Governed client connectors, identity, access control, and data isolation.
- /Human review, evidence-quality scoring, and persisted analysis history.
- /Monitoring for the API, retrieval layer, orchestration, and model calls.
Problem
A strategic market question needs to become an auditable decision workflow rather than a one-shot answer.
System flow
5 stages · click to inspect
Strategic question
One market or strategy question enters the workflow as typed input, not as free-form chat. The question itself becomes an artifact that can be re-run and compared later.
Signature module · hackathon case study
Inspect the decision package
Structured analysis / Qdrant evidence / workflow trace
Follow one strategic question through context assembly, evidence retrieval, hypothesis ranking, actions, and validation tasks.
- INPUT
- one strategic market question
- CONTEXT
- synthetic company data + public evidence
- RETRIEVE
- Qdrant evidence when configured
- REASON
- rank hypotheses + validation tasks
- OUTPUT
- brief · actions · sources · workflow trace
- BOUND
- decision support · not causal proof
Demonstrated change
Demo scope
Before
A strategic question spread across disconnected signals and ad-hoc analysis.
->
Demonstrated build
A structured demo package with hypotheses, evidence, actions, and validation tasks.
The hackathon build produces a structured decision package with ranked hypotheses, an executive brief, actions, evidence references, and a workflow trace.
Core company and market data is synthetic/generated; public evidence quality varies. Decision support, not causal proof.
Proof signals
Classified evidence
- 8validated demo artifactsRepository fact
- 60companies in source listSynthetic data
- 1decision package per requestRepository fact
Every signal is classified as a repository fact, recorded test, configuration, or disclosed data boundary.
Why this is hard
3 engineering challenges
Challenge · 01
Evidence separation
synthetic business context must not be confused with retrieved public evidence.
Challenge · 02
Structured reasoning
hypotheses, actions, and validation tasks need a stable contract.
Challenge · 03
Demo resilience
cached analysis must remain transparent when live integrations are unavailable.
Engineering depth
3 topics · click to expand
The analysis orchestrator stages refinement, aggregation, retrieval, hypothesis generation, scoring, executive briefing, and trace assembly.
Key numbers
- 3request modes defined: demo, enriched, live
- 8cached artifacts required by validator
The synthetic FreshCart scenario and the retrieved public evidence travel through the workflow as separate, labelled inputs, so a reader can always tell which part of a conclusion rests on generated data.
Boundary
- 2input classes kept separate end to end
- 0client datasets involved
Ranking is directional. The package is designed to make a question investigable — to show what to check next — rather than to assert a validated business recommendation.
Limit
- 1decision package per request
- 0causal claims made
- Built with
- React
- Express
- TypeScript
- Qdrant
- LangSmith
Synthetic core data / hackathon case study
What this demonstrates
Decision work can be made structured and reviewable: hypotheses, evidence and validation tasks in one contract rather than an answer in a chat window.
When this pattern is useful
Recurring strategic or market questions are answered ad-hoc, and the reasoning cannot be reviewed, reused or challenged afterwards.
Relevant engagement
A Data & AI Sprint scopes this pattern against your own question and data, and ends with a decision on whether to build it.
Data & AI Sprint →Is there a question in your business shaped like this one?
Bring the question. The first conversation is about whether it is worth building anything at all.