RivalMap

From product idea to an evidence-backed competitive map—live, visual, and ready for your strategy deck.

Comment

Inspiration

Competitive analysis is usually slow, fragmented, and hard to visualize. Founders and product teams often jump between search results, product pages, notes, and slides, then still have to manually turn everything into a usable market map.

We built RivalMap to make that process faster and more intuitive: starting from a product idea, it progressively researches the market, identifies the closest competitors, and turns the results into an evidence-backed visual competitive landscape.

Our goal was not just to generate a list of competitors, but to create something a team could actually use in strategy work: a map that shows where your product sits, who is closest to you, how you are differentiated, and where opportunities may exist.

What it does

RivalMap turns a product idea into a live competitive positioning map.

A user starts with a simple idea such as “AI interview coaching platform.” RivalMap asks a few framing questions, begins research as soon as the market is sufficiently defined, and progressively builds a market view.

The product experience highlights:

  • a user-centered market map
  • a Focus Ring / closest competitive space
  • nearby and broader market context
  • concise strategy insights:
    • Closest Rivals
    • Your Differentiation
    • Opportunity Around You
  • on-demand product comparison
  • exportable presentation-ready output for strategy decks and internal discussions

How we built it

We built RivalMap as a bounded multi-agent system with Strands as the orchestration layer.

The system combines:

  • a Market Framing Agent to understand the user's product idea
  • a Strands Orchestrator Agent as the high-level decision-maker
  • a Research Agent that runs parallel web research with Exa
  • a Market Intelligence Agent that performs structured and semantic product analysis
  • a Presentation Agent that prepares human-facing summaries and presentation metadata

Under the hood, RivalMap uses a progressive pipeline:

  1. frame the market brief
  2. research candidates in parallel
  3. validate candidates
  4. analyze products with structured + semantic intelligence
  5. compute similarity to the user's idea
  6. generate a stable competitive map
  7. identify the closest competitive space
  8. progressively update the UI and exportable presentation view

We used Amazon Bedrock for model execution, FastAPI for the backend and SSE streaming, and a lightweight frontend to render the progressive map experience.

Challenges we ran into

One major challenge was balancing agentic flexibility with deterministic reliability.

We did not want a system that simply “looked agentic” but was slow, unstable, or impossible to control. So we designed RivalMap so that:

  • Strands agents decide what to do next
  • deterministic systems enforce budgets, validation, and map stability
  • useful output appears early, instead of waiting for the whole run to finish

Another challenge was making the visualization meaningful. We wanted the map to feel more like a strategic artifact than a generic dashboard or scatterplot. That led us to focus on:

  • stable incremental layout
  • a user-centered map
  • a progressive reveal of the closest competitive space
  • presentation-ready export

We also had to handle edge cases such as sparse markets, uncertain evidence, and partial degradation while preserving useful intermediate output.

What we learned

This project taught us a lot about building bounded agent systems that are both interactive and reliable.

We learned that:

  • progressive output matters more than one big final answer
  • visual structure can make market research much more intuitive
  • the difference between “an agent demo” and “a useful product” is often in the deterministic layers: validation, budgeting, and presentation
  • exportability matters — if users cannot easily use the output in a meeting or strategy deck, the value is lower

What's next for RivalMap

Next, we want to improve:

  • richer presentation and export modes
  • better refinement during an active run
  • stronger comparison views
  • more polished deck-ready outputs
  • broader support for different product categories and market densities

Our longer-term vision is for RivalMap to become a practical tool for founders, PMs, and strategy teams who need to quickly understand the competitive landscape around a new product idea.

Built With

Share this project:

Updates

Submission history