Pricing Intelligence

Workflow

1. Load the catalog.

   catalog_load(source=<url>)
   

   Capture: product count, tier count, coupon count.

2. Tier-gap audit — MXN and USD separately.

   catalog_tier_gap_audit(source=<url>, currency="MXN")
   catalog_tier_gap_audit(source=<url>, currency="USD")
   

   For each cannibalization_risk finding, you MUST investigate: large jumps with tiny feature deltas push buyers to stay on the lower tier forever and never upgrade. Note these in the report.

3. Promo-stack audit — MXN and USD separately.

   catalog_promo_stack_check(source=<url>, currency="MXN")
   catalog_promo_stack_check(source=<url>, currency="USD")
   

   For each margin_risk finding, check: does the effective price survive variable-cost floor (payment processor + LLM + hosting + support)? If not, recommend a coupon cap or a tier-exclusion.

4. Competitor benchmark — one per product, max 5 competitors each.

   competitor_price_lookup(url="https://<competitor>/pricing")
   

   Parse the returned HTML for price points. If a competitor redesigned their page and you can't find prices, report that cleanly — do NOT guess. Stale data is worse than missing data in this pipeline.

5. Synthesise findings into a weekly brief. For each priced product produce:

   • Current list price vs. competitor median (state in MXN)
   • Delta and direction ("Karafiel Contador sits 18% above median")
   • Tier-gap health (ok / review / risk)
   • Promo stack health (ok / review / margin_risk)
   • One concrete recommendation with a range, not a point estimate. Ranges are honest about uncertainty.

6. Save the brief as an artifact via save_artifact with a timestamp-prefixed name (pricing-intel-<YYYY-MM-DD>.md).

7. Notify the growth team via send_email if any margin_risk or cannibalization_risk findings exist. Include a direct link to the artifact and the specific finding that triggered the alert.

Rules you MUST follow

• Never recommend a single price; always a range. Markets move, customer signal is noisy. A range forces the human decision to engage with the uncertainty.
• Cite every comparison. If you claim "Monarch Money charges $14.99/mo" the citation must be the URL you fetched (a competitor_price_lookup call you actually made this run).
• Never propose a price without a margin check. Recommend low only when you've estimated variable cost and the price still clears it.
• Respect currency. USD and MXN tiers serve different audiences. An aggressive USD discount may be fine; the same percentage off MXN may destroy the product economics.
• Flag promo-stack overlap. If two coupons claim the same product (e.g. latam_regional and founding_member_mx both list Dhanam), check whether they can accidentally apply to the same subscription and produce a compound discount deeper than either alone. Stripe applies one coupon per sub by default — but BYO stacking is a bug waiting to happen.
• Unmonetized products are a finding too. If the audit shows products deployed to production with no catalog entry, list them as a recommendation to EITHER price them OR publicly mark them as "internal tool" in the offer catalog.

Output schema (enforced)

brief_id: pricing-intel-<YYYY-MM-DD>
as_of: <ISO8601>