OrderHUBx RightPick · Early Access

Your AI Product Advisor

Shopping carts have looked the same for 25 years: a search box and a grid. RightPick replaces that with a guided buying layer that behaves like your best sales rep — recommending the right products, the right companions, and the right quantities, with evidence behind every answer.

12+ Decision Facts per SKU
100% Evidence-Backed
3 Deployment Modes
0 Dead-End Searches

The Problem & The Solution

Most stores answer "what do you sell?" — not "what should I buy?" RightPick closes that gap.

Search-and-Grid Store With RightPick
Shopper guesses which product fits their job RightPick maps the job to the right product and explains why
"How much do I need?" goes unanswered Quantity logic sizes the order to coverage, area, or run length
Required companions forgotten → failed projects, returns Required and optional companion SKUs surfaced before checkout
Marketing copy overstates what a product can do "Not suitable for" and "choose instead when" stated honestly
Recommendations are a black box you can't trust Every fact carries a source, an excerpt, and a confidence score
Carts are smaller and returns are higher Shoppers leave with the right mix and high value-for-money

Six Questions a Good Advisor Always Answers

RightPick turns shallow catalog data into a decision layer that answers the questions an experienced owner-operator would — for every product, automatically.

🎯

What is it for?

Primary and secondary use cases, plus the contexts a product is genuinely suitable for — not just the marketing headline.

🚫

When is it the wrong choice?

Explicit "not suitable for" conditions and "choose instead when" rules, so shoppers avoid the return and trust the store.

🔗

What else is required?

Required companion SKUs for the job to succeed — primer, hardener, fixings, consumables — plus optional add-ons.

🔢

How much do I need?

Quantity and configuration logic that sizes the order to coverage area, run length, pack size, or usage rate.

🔁

What's the better option?

Ranked alternatives with the trade-offs spelled out, so a shopper on a budget or a deadline still gets the right fit.

📎

Why should I believe it?

Every conclusion is backed by evidence — a source URL, an excerpt, an extraction method, and a confidence score.

From Catalog to Counsel

RightPick is a pipeline, not a chatbot bolted on top. It ingests your catalog, enriches it with public evidence, and normalizes everything into decision-grade JSON before a single recommendation is made.

1
Ingest

API-first wherever possible — WooCommerce Store API and similar — with respectful, robots-aware crawling as fallback. Identity, pricing, variations, images, and JSON-LD captured per product.

2
Enrich

Adds evidence from on-site FAQs and help content, public PDFs and technical docs, video transcript signals, and approved peer sources — exactly where merchant copy runs thin.

3
Normalize

Converts raw and enriched content into a canonical decision schema: use cases, suitability, companions, alternatives, quantity logic, and risk notes — with provenance and confidence on every fact.

4
Guide

The normalized corpus powers guided selling, comparison, and cart composition — via cloud LLMs or a private small model you control. Low-confidence or safety-sensitive facts route to human review first.

Decision-Grade, Not Marketing-Grade

RightPick never turns flowery copy into false certainty. Every product becomes a structured pack of facts — and every fact knows where it came from.

What RightPick Returns for Each Product

Primary Use Cases Suitable For Not Suitable For Choose Instead When Required Companions Optional Companions Alternatives Quantity Logic Risk Notes Value Positioning Provenance Confidence

Exported as canonical JSON per product, JSONL fact records, or knowledge packs grouped by category and use-case.

Confidence-Based Review

High confidence — eligible for export to production guided-selling.
Borderline — surfaced in the review queue for a quick human check.
Safety-sensitive — held back until a reviewer approves, amends, or rejects it.

Evidence From Where It Actually Lives

The depth that makes a recommendation trustworthy rarely sits in the product description. RightPick gathers it from the sources that do — and keeps the provenance.

On-Site FAQ & Help

Application notes, usage limits, and "does it work with…" answers that buyers actually ask.

📄 PDFs & Technical Docs

Spec sheets, datasheets, coverage tables, and safety guidance — parsed into structured facts.

🎥 Video Transcripts

Caption and transcript signals from how-to and demo videos, via multiple transcript adapters.

🌐 Approved Peer Sources

Comparison and limitation language from approved peer or manufacturer sites — per-tenant policy, never unbounded scraping.

Contradiction detection runs across sources, and provenance is never collapsed — conflicting claims are scored, not silently merged.

The Right Quantity, Not Just the Right Product

Picking the product is half the job. RightPick reasons about how much, which configuration, and what has to ship alongside it.

Shopper Intent What RightPick Recommends
"Coat a 40 m² garage floor"Quantity from coverage rate × two coats, plus required primer and a recommended top-coat companion
"This product, but for outdoor use"Choose instead — flags the indoor-only limitation and ranks a UV-stable alternative
"Cheapest option that still works"Best value-for-money pick with the trade-offs stated, not just the lowest price
"Add the kit to do the whole job"Composes the cart: base product, required companions, consumables, and quantities
"Is this safe for food-contact surfaces?"Answers only from evidence; if unverified, routes to review rather than guessing

SaaS-First, Private-Capable, On-Prem-Deliverable

One codebase, three deployment modes. Keep product intelligence and business data exactly where your compliance team needs it — and feed cloud LLMs or a private small model you train yourself.

Most teams

SaaS Cloud

Multi-tenant managed cloud. Fastest to onboard, continuously improved extractors, cloud LLMs by default. Ideal for SMB and mid-market brands.

Privacy-sensitive

Private Managed

Single-tenant VPC, dedicated VM, or managed Docker. Hybrid model strategy — cloud or private inference — for mid-market manufacturers with data concerns.

IP-protective

On-Prem Appliance

Docker Compose or Kubernetes in your environment. Private small models and local retrieval by default — cloud LLM use can be disabled entirely.

The durable advantage isn't the crawler or the chat box — it's the versioned, evidence-backed product intelligence corpus, which can feed both guided-selling apps and your own domain-specific model.

Pricing

RightPick is in early access. Ingestion, enrichment, and decision-grade exports are taking shape now — join early access for priority onboarding and pricing.

🎯 RightPick Module

Early Access
A guided-selling add-on for OrderHUBx. Pricing finalized as features reach general availability.
API-first ingestion with respectful scrape fallback
Evidence enrichment (FAQ, docs, video, peer sources)
Canonical decision-grade JSON & knowledge packs
Provenance & confidence on every fact
Human review queue for safety-sensitive facts
Cloud LLM or private small-model downstream
SaaS, private managed, or on-prem deployment
Join Early Access →

Frequently Asked Questions

Common questions about RightPick, your data, and how recommendations stay trustworthy.

Does my product data go to a cloud AI vendor?

Only if you want it to. RightPick uses provider-abstracted model adapters, so cloud and private model paths are interchangeable. On the on-prem appliance, private small models and local retrieval run by default and cloud LLM use can be disabled entirely.

How do I know a recommendation is accurate?

Every derived fact keeps its provenance — source URL, source type, extraction method, an excerpt, a timestamp, and a confidence score. Contradictions across sources are detected and scored, and safety-sensitive claims are held for human review before they ever reach a shopper.

Does it replace my store's search?

It sits on top of it. Search answers "what do you sell." RightPick adds the guided buying layer that answers "what should I buy, how much, and what goes with it" — turning browsing into a conversation with your best sales rep.

What if my product pages are thin?

That's exactly what the enrichment layer is for. RightPick pulls supporting evidence from your FAQs, PDFs and spec sheets, video transcripts, and approved peer sources — so a sparse catalog still produces decision-grade answers, with every added fact sourced.

Can I use the data to train my own model?

Yes. The same normalized corpus exports as canonical JSON, JSONL fact records, or knowledge packs — ready for retrieval, agent workflows, or fine-tuning a domain-specific small model you deploy privately. The platform is built to be a knowledge and training-data factory, not just a chat wrapper.

Is RightPick available now?

RightPick is in early access. Core ingestion, enrichment, and decision-grade exports are in active development, with deployment-flexibility and vertical knowledge packs on the roadmap. Join early access for priority onboarding and to help shape the first vertical packs.

Help Shoppers Buy Right the First Time

Schedule a demo to see RightPick turn a real catalog into decision-grade product intelligence.