CONTEXT(CLONE)
Sign in

Competitor teardown from public sources with citations

For: Strategy analyst, product leader or founder preparing a market entry or board discussion

Pattern: Hierarchical decompositionNeeds a connectorDesigned for 20 to 500 agents
Get early accessSign in to use

The pain today

A teardown is stitched from a few web pages and hearsay. Claims about pricing, customers and roadmap lose their sources by the second draft, and nobody can tell fact from guess.

The ask

“Build a teardown of the competitors on my attached list from public sources only: product, pricing, packaging, target customers, hiring signals and documentation. Every statement needs a link and a quote, and I want guesses marked as guesses.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Competitor list
  • Dimensions and questions to cover
  • Public web pages, documentation, job posts and filings
  • Our own product summary for comparison

The unit of work

One worker task per one question about one competitor from one source.

Why a swarm fits

The goal decomposes: competitor, dimension, question, source. Leaves are small page reads that run in parallel, and each dimension is verified before it enters the profile.

Not for

A quick look at one rival's pricing page: open the page.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Split or keep serialYes or no, with a probability

    While planning

    Does this dimension for this competitor break into several factual questions with different likely sources, such as the pricing page, documentation and job posts?

    Sees only: One dimension's question list and source types

    Why: Keeps each leaf to one question and one source so every statement stays traceable.

    • Yes: 0.60 or higherthenSplit the work
    • Unsure: 0.30 up to 0.60thenKeep serial
    • No: below 0.30thenKeep serial

  2. Scope checkA choice among options

    Before work starts on a unit

    What kind of source is this page?

    Sees only: The page's address, publisher line, date and opening text

    Why: Provenance is judged before content, so improper material never enters a profile.

    • Published by the competitor or an official filingthenAccept
    • Independent press or review sitethenAccept
    • Forum, rumour or undated pagethenMark unresolved
    • Looks non-public or leakedthenAsk a person

  3. After workers, the judge checks

    Evidence checkA choice among options

    After a worker answers

    Does the quoted text state the fact outright, or does the statement infer it?

    Sees only: One statement and its quoted text

    Why: Guesses are labelled as guesses instead of hardening into facts by the second draft.

    • States it directlythenAccept
    • A reasonable inference from the quotethenMark unresolved
    • Promotional language with no checkable factthenSkip this unit
    • The quote does not bear on itthenReject and retry

  4. Reconciler, while merging

    Conflict checkYes or no, with a probability

    While reconciling

    Do these two sources give different answers to the same question about the same competitor for overlapping dates?

    Sees only: Two verified statements with their quotes and capture dates

    Why: Conflicting sources stay visible so the analyst can weigh them.

    • Yes: 0.70 or higherthenMark unresolved
    • Unsure: 0.40 up to 0.70thenEscalate to a strong model
    • No: below 0.40thenAccept

  5. Run control, between rounds

    Another round?Yes or no, with a probability

    Between rounds

    Is every question under this dimension either answered with a citation or listed as unknown?

    Sees only: The dimension's question status table

    Why: A dimension enters the profile only when its gaps are named.

    • Yes: 0.85 or higherthenStop
    • Unsure: 0.50 up to 0.85thenContinue
    • No: below 0.50thenContinue

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does this statement concern a competitor's customers, finances, roadmap or staff, and would it be repeated to the board or outside the company?

    Sees only: One verified statement with its source

    Why: The analyst confirms provenance and owns what the findings mean for strategy.

    Accountable: The analyst decides what the findings mean for strategy and confirms nothing came from non-public or improperly obtained material.

    • Yes: 0.30 or higherthenAsk a person
    • Unsure: 0.10 up to 0.30thenAsk a person
    • No: below 0.10thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model builds the tree of competitors, dimensions and questions, and names source types for each.

    Decisions here:1. Split or keep serial2. Scope check

  2. Workers

    Small fast workers from an open-weight family each read one source for one question and return a quoted answer.

    Designed for 20 to 500 agents, one worker task per one question about one competitor from one source. Each worker receives only its own unit.

  3. Judge, from a different model family

    A decision model from a different family checks the quote supports the statement and is not marketing puffery.

    Decisions here:3. Evidence check

  4. Reconciler

    A strong reasoning model rolls up each dimension, keeps conflicting sources visible and labels inference as inference.

    Decisions here:4. Conflict check5. Another round?

  5. Accountable person

    The analyst decides what the findings mean for strategy and confirms nothing came from non-public or improperly obtained material.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every statement carries a link, a quote and the capture date
  • Statements resting on one source are marked single-sourced
  • Inference from hiring or documentation is labelled as inference
  • A dimension rolls up only when its questions are answered or listed as unknown

What comes back

  • Profile per competitor with cited statements
  • Comparison table by dimension
  • Unknowns public sources do not reveal
  • Conflicting sources

What to measure

  • Statements that hold up in a sampled source check
  • Unknowns closed against the previous teardown
  • Analyst hours per competitor
  • Cost per cited statement

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Research and intelligence

Literature review where every claim cites a passage

For: Research lead, analyst or doctoral researcher scoping a field before committing to a direction

A review means a very long reading list.

Pattern: Map, verify, reduceNeeds scale5 decisionsDesigned for 30 to 800 agents