Set up market position

Use cases

• “Set up market position”

Workflow

How It Works

• 1

  Confirm the competitor set

  Collect competitor names, approaches, URLs, or the explicit discovery fallback.

• 2

  Articulate the value proposition

  Reuse or articulate the value proposition from the exact solution and customer segment already in the workspace.

• 3

  Discover candidate competitors

  Surface 3-6 credible competitor candidates for the value proposition. Outputs evidence-finding rows and a structured outputs.candidates list that drives the per-candidate fetch fanout.

• 4

  Pre-fetch each candidate's pages (deterministic)

  For each candidate from discover.outputs.candidates: scrape the homepage + standard subpages, save each as an evidence-source, and create a cites edge tying each evidence-source to a fresh solution entity. No LLM. Per-candidate manifest of solution_id + evidence ids drives enrich/assess.

• 5

  Profile each competitor solution

  For each candidate: build a picture from the pre-fetched evidence and update its solution with description/features/packages/pricing. One write per clone (reliable). Raw profile only — no coverage, no differentiation.

• 6

  Judge each competitor's coverage (returns data)

  For each enriched competitor: read its solution + the segment, RETURN per-moment coverage judgments (keyed by moment label) + differentiation as outputs. Writes NOTHING — the write-coverage fetcher persists the cells. min_success=1.

• 7

  Judge our own solution's coverage (returns data)

  Judge OUR OWN solution against every moment using the same rubric (richer internal data, no differentiation). RETURNS per-moment coverage judgments (by moment label) as outputs. Writes NOTHING — the write-coverage fetcher persists the YOU column alongside the competitors.

• 8

  Write coverage entities (deterministic)

  Deterministic fetcher: resolve each returned moment label to a real moment id and create one coverage entity per judgment — competitor (from assess.items) and own (from own-coverage.outputs) — attaching cited_source_id as a cites edge. Partial-tolerant: an unmatched label or a failed create is skipped and counted, never fails the step.

• 9

  Consolidate per-competitor differentiation

  Pure-Python fan-in. Reads assess.items and produces differentiations_by_solution (keyed by solution_id) + solution_ids for the VP writer. Coverage is NOT consolidated here — it was written as entities by write-coverage.

• 10

  Persist competing solutions to the VP

  Single LLM agent that takes the consolidated differentiation and writes it in ONE batched update to value_proposition.competing_solutions[] (the vp_competing_solutions junction). Does NOT write coverage or touch the segment.

• 11

  Build coverage matrix

  With the VP's competing_solutions populated AND every coverage entity (competitor + own) written, build the coverage matrix showing how each solution covers (or doesn't) each moment.