April 2026

April was a dense month. Five distinct feature areas shipped, plus the public documentation site you're reading right now.

Listing strategy and keyword classification

The most significant listing-optimization change this month was the introduction of the listing classification engine — an automated scorer that assigns a keyword_class to every keyword in an ASIN's keyword master: conviction, proven, opportunity, or none.

Conviction is the manual anchor. You pin a small set of focus keywords as conviction terms; the engine never touches them. Proven and opportunity are computed from a combination of ranking signal, search volume, and SQP/STR data. The engine runs after every keyword harvest, after every strategy save, and on a 5-second client-side debounce after you edit a listing field.

Alongside the engine, a new strategy brief feature lets you document the ASIN's positioning, target customer questions, and audience notes. The brief feeds the AI listing optimizer as context — the AI now generates copy informed by your actual positioning, not just keyword frequency.

The AI refinement flow got two new capabilities: an "AI-refine opportunities" sub-panel that surfaces ranked improvements for opportunity-class keywords, and a "Suggest conviction" popover that uses AI to recommend which keywords deserve conviction status based on your strategy brief. Both are cached by input hash to avoid redundant API calls.

See Strategy briefs and keyword classes and Conviction / proven / opportunity.

Advertising v2 — master/detail workspace

Campaign setup was refactored out of the research diagnostics flow and into a dedicated /advertising master/detail workspace. The new workspace gives each ASIN a blueprint panel with a full campaign list on the left and an editable detail pane on the right.

The blueprint model captures not just campaign structure but also creative availability flags, product type, review count tier, portfolio name, and a product nickname for display. Blueprint items track purpose (ranking, performance, research, shielding, retargeting), ad product, match type, targeting type, bidding strategy, placement adjustments, and linked campaign name for STR matching.

See Advertising workspace.

Per-ASIN performance imports

Previously, the Performance section only stored account-level snapshots. This month added a full per-ASIN import flow backed by Amazon's "Detail Page Sales and Traffic By Child Item" business report.

Each upload creates an import record; snapshots are cascade-deleted when an import is removed, which makes it safe to re-import a period with corrected data. Re-uploading the same (seller, cadence, period_start, period_end) replaces the previous import cleanly — no orphaned rows.

The by-ASIN panel has three view modes: Trends (time-series charts per ASIN), Grid (overlaid multi-line chart for all ASINs on a shared axis with a clickable legend), and Compare (sortable table for a selected period). The "All ASINs" rollup uses weighted aggregation — CVR and Buy Box are weighted by sessions, not averaged naively.

Performance history now has a tab switcher: account snapshots (existing) and ASIN imports (new), giving you two entry points for managing imports.

See Importing Business Reports · Performance workspace.

PDF catalog export

Selecting "All ASINs" in the Performance Trends view now offers a PDF export. The export generates a branded catalog document with: a header showing the full data date range, a "Last Week" comparison table, a "Last Month" comparison table, and per-ASIN detail cards with the product image and six KPI cells.

Product images are fetched client-side using a canvas cross-origin fetch, with a graceful "No image" placeholder when unavailable. Image URLs are resolved from the primary listing baseline's image_main field.

The date range in the PDF header always reflects the full dataset span, not whatever 7d/30d/90d filter is active in the UI — so the export is always a complete picture.

Public documentation site

This documentation site shipped this month. It is built directly into the amazon-platform Next.js app at /docs and features: MDX content with a custom component library (Callout, Steps, FieldTable, Field, InRankasin), a sidebar navigation tree driven by meta.json files, a table of contents on every page, a Pagefind-powered ⌘K search modal, per-page feedback widgets (thumbs up / down), OG image generation, JSON-LD metadata, /llms.txt and /llms-full.txt for AI ingestion, and a .md mirror route for AI scrapers.

Content launched with Getting Started, Research, Listing Optimization, Advertising, Performance, Compliance, Brands, Tasks, Reference, and Changelog — roughly 60 pages total.

Minor fixes

• Catalog PDF date range now derives from the full (unfiltered) snapshot dataset, not the active time filter, so the PDF header is always accurate.
• The listing editor's Title↔Item Name mismatch warning now compares against the baseline value rather than the live edit state, preventing false positives during editing.
• Fixed a bug where removing the asin_filter cookie after deleting an ASIN could leave orphaned baseline and import data.