Competitor pricing and product page watch

The decision tree

5 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. 1

   Run control, between rounds

   Another round?A choice among options

   Between rounds

   Is this capture of the competitor page complete, a partial load, or blocked?

   Sees only: The current capture of one watched page with its fetch status

   Why: A half-loaded pricing page read as a price change would mislead every seller.

   • Complete→thenContinue
   • Partial load→thenReject and retry
   • Blocked or removed→thenMark unresolved
2. 2

   Planner, while planning

   Scope checkYes or no, with a probability

   Before work starts on a unit

   Does the difference between the two captures change a price, a plan limit, a feature's availability or a stated claim, rather than layout or imagery?

   Sees only: The changed passages from the previous and current capture

   Why: Design refreshes are logged and dropped before any battlecard line is re-opened.

   00.25 and 0.551

   • Yes: 0.55 or higher→thenAccept
   • Unsure: 0.25 up to 0.55→thenAccept
   • No: below 0.25→thenSkip this unit
3. 3

   Before workers, before a task runs

   Small worker or strong modelYes or no, with a probability

   Before a task runs

   Is the changed passage a pricing table or plan grid whose conditions and footnotes need a strong model to read correctly?

   Sees only: The changed passage

   Why: Footnotes are where plan limits hide, and small workers skip them.

   00.25 and 0.501

   • Yes: 0.50 or higher→thenEscalate to a strong model
   • Unsure: 0.25 up to 0.50→thenEscalate to a strong model
   • No: below 0.25→thenAccept
4. 4

   After workers, the judge checks

   Evidence checkA choice among options

   After a worker answers

   Set against the new page text, is this battlecard line still true?

   Sees only: One battlecard line and the before and after passages it rests on

   Why: Sellers repeat battlecards to buyers, so a line is only called wrong on quoted evidence.

   • Still true→thenSkip this unit
   • Now false→thenAccept
   • Partly true because a condition changed→thenEscalate to a strong model
   • The page no longer says either way→thenMark unresolved
5. 5

   Accountable person, before anything is settled

   Person decidesYes or no, with a probability

   Before anything is reported as settled

   Would this correction change what sellers tell buyers about a competitor's price or capability?

   Sees only: One draft correction with its before and after quotes

   Why: Product marketing confirms anything that will be said to a buyer.

   Accountable: Product marketing decides what changes in the battlecard and confirms anything sellers will repeat to buyers.

   00.10 and 0.301

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