Syntalic — Data Analytics for Agentic Commerce — MPP Grade C

Find the lowest current price for a product across retailers in US and Canada. Cross-retailer comparison is entity-matched (barcode-anchored resolution links the same physical product across platforms). Returns the cheapest option plus other retailer prices for comparison. Every price row is labeled match_type: 'entity' (verified same product) or 'title' (text match — may be a variant on broad queries); pass strict=true to restrict the comparison to entity-verified rows only. Comparison rows carry observed_at so mixed-vintage prices are distinguishable.

Get historical price observations for a product within a date range. Returns current price, period low/high/avg, trend (rising/falling/stable), good-deal flag, and an observation time-series for charting.

Discover discounted in-stock products in a category above a minimum discount threshold (discount = current price below the retailer's list price). Quality-gated: unbranded listings, sub-$5 items, and discounts above 70% (the inflated-list-price spam signature) are excluded. Ranked by discount depth weighted by log(price), so meaningful discounts on real products outrank deep cuts on trinkets. Each deal carries observed_at (when its price was seen).

Check whether a product's current price is below its rolling average within a configurable lookback window. Returns current price vs. the average across the window plus the lowest-seen price and date. Response field `avg_price_last_30d` is a fixed name for backwards compatibility; the value is always computed across the configured `lookback_days` window.

View every product and its current pricing within a category across retailers. Sortable by price (rating/review sorts are accepted for compatibility and fall back to price ordering). Each item carries a condition label (new/refurbished/used) so refurb listings are distinguishable. Offset-cursor paginated.

Track a brand's average/min/max price, product count, in-stock count, and promo count day-by-day across a date range. Brand input matches case-insensitively and across normalization variants (jbl == JBL, tplink == TP-Link). Each point reports insufficient_sample when its basket is under sample_threshold products, plus stable_avg_price computed over only the products priced on every day of the series — use stable_avg_price for trend reads; the all-products average moves with basket composition.

Analyze promotional activity within a category - promo frequency, average and max discount depth - over a date range. Pivot the breakdown with `aggregate_by`: default `brand` ranks brands within the category; `retailer` ranks retailers (use together with `brand=<name>` to answer 'which retailers run the deepest promos on Brand X in Category Y'). The response key mirrors the dimension: `brands: [...]` or `retailers: [...]`.

Measure each brand's market share within a category by product count. Shows digital shelf dominance. Brand rows are merged across source variants ('JBL'/'jbl', 'TP-Link'/'tplink') with one canonical display label; placeholder brands (null, 'no', 'Generic') are excluded.

Compare a brand's average current price to the category average/median and classify positioning as premium, mid-range, or value.

Track category-level price inflation with configurable daily, weekly, or monthly granularity. Period-over-period change_pct uses matched-pairs methodology: only products priced in both consecutive periods count, aggregated as the geometric mean of price relatives (Jevons), so basket-composition shifts don't masquerade as inflation. Each period reports matched_basket_size and method ('matched', or 'naive' avg-vs-avg fallback when fewer than 10 products match); overall_change_pct chains the per-period changes. The catalog refresh cadence is monthly, so granularity=monthly gives the most stable matched baskets.

Analyze the spread of current prices within a category - mean, stddev, coefficient of variation, and percentiles (p10..p90) with IQR outlier exclusion. The population is representativeness-filtered: refurbished/used listings, accessory/parts subtrees, and rows under a category-aware price floor are excluded from headline stats, with the counts disclosed in `excluded`.

Compute a normalized price index (retailer avg / category avg) for a specific retailer day-by-day versus its category baseline. Points report insufficient_sample below sample_threshold products plus stable_avg_price over the products priced every day — read trends from the stable series.

High-level category statistics - product/brand/retailer counts, pricing (avg/min/max/median), promo rate, in-stock rate, and top brands. Headline stats are representativeness-filtered (no refurb/used, no accessory/parts subtrees, category-aware price floor) with exclusion counts disclosed in `excluded`; brand_count and top_brands merge brand variants on the canonical key and exclude placeholders.