Velou: enrich the catalogue before an agent touches the sale

External-source research synthesis, not an OfferSignal client result. Based on the retained October 2026 research and source-specific dates.

A catalogue can contain the right product and still fail the buyer. Descriptions may omit material, fit or use case; images may carry information absent from text; and inconsistent tags make search and recommendations less useful. Velou addresses that information problem first. It is a product-data enrichment layer, not evidence of an autonomous procurement or sales agent.

What Velou does to a catalogue

Velou describes a workflow that starts with product text and images, infers missing attributes, normalises tags and categories, adds contextual content and writes results back to commerce systems or search indexes. Its public Shopify example turns a sparse dress record into category, material or finish, neckline, occasion, seasonality, fit guidance, FAQs, image alternative text, related products and an agent-ready export. The stated use is discovery across product pages, Google and ChatGPT—not selecting suppliers, negotiating, placing orders or authorising a binding quote. 2 3

The packaged route is Shopify-centric for smaller teams. The App Store listing supports Fashion, Jewelry, Furniture and Home goods, with permissions to view and edit products, listings, collections, Online Store data and custom data or metaobjects. Velou says it supports Online Store 2.0 and headless Hydrogen or custom stacks. Algolia describes offline or batch enrichment pushed through an application programming interface, alongside a near-real-time path. The customer’s commerce, inventory, ranking, checkout and approval systems remain responsible downstream. 4 8

Status and reported results

Algolia announced on 6 October 2026 that it acquired Velou for undisclosed terms. It says the six-person team joined Algolia, Velou’s application programming interfaces, software development kits and integrations remain the same, and the company had served more than 50 retail and premium-brand enterprise customers. The Shopify app remained publicly listed with free and paid tiers and a 0.0 rating from zero reviews when retained research was accessed on 27 October 2026. 1 4

The strongest public outcomes are provider or acquirer reported. Algolia says U.K. sportswear retailer Get The Label added more than 190,000 attributes over six months and saw site-search revenue rise 60%. It says Everything5pounds increased catalogue attributes by more than 85% and search conversions by 33%. Starting catalogue size, traffic denominators, control design, seasonality, attribution method and absolute revenue are undisclosed. Velou’s Shopify page quotes Kut from the Kloth reporting a 20% return-on-ad-spend increase, without a period, spend, baseline or independent corroboration. These figures are not comparable or causal benchmarks. 1 3

Commerce-1 claims training on 4 trillion retail text and image tokens, reinforcement learning from merchandisers, double-digit accuracy gains, more-than-half hallucination reduction, nightly clickstream retraining and safety filters. Its stated May 2025 benchmark used hundreds of real stock-keeping units against three frontier models. No public test set, per-attribute confusion matrix, evaluator protocol or reproducible report is retained, so these remain vendor claims. 7

An illustrative industrial application

For a workshop, distributor or manufacturer, use Velou upstream of search and quoting: ingest approved catalogue text and images, map outputs to a controlled taxonomy, review high-risk fields and publish approved attributes to Shopify, Algolia or an equivalent index. This is illustrative, not a reported industrial deployment. A bounded pilot could cover one repeatable component family and require source-linked values for units, revision, material, compatibility, certification and authority.

Do not let an inferred image or free-text value become a technical fact. Use immutable product and revision identifiers, source and timestamp per attribute, unit normalisation, duplicate detection, confidence thresholds, approval queues, change diffs and rollback. Materials, certifications, compatibility, capacity, prices and lead times need authorised owners. RFQ, configure-price-quote and binding-offer decisions should remain in governed commercial and engineering systems.

Cost and a sensible test

Public Shopify plans are free for up to 10 products; $99 per month plus $1.50 per product up to 500; or $250 per month plus $1.25 per product up to 3,000, billed every 30 days. Enterprise pricing, integration, service levels, retention, export and post-acquisition commercial terms are undisclosed. Total cost also includes catalogue cleanup, mapping, review, exception handling, integration, testing and ownership. 4

Measure source-grounded precision and recall for high-risk attributes, unsupported-claim rate, correction time, duplicate or stale-record errors and search outcomes. Quarantine low-confidence values and compare approved publication with the source record. Velou’s positioning moved from a 2024 AI ecommerce copilot to product-data infrastructure for agentic commerce. The practical lesson is to improve the information layer first, then test discovery, while keeping authority for a sale elsewhere. 2 5 6 7

Sources and next reading

• Velou official product site
• Velou Shopify product workflow
• Velou Shopify App Store listing
• Velou Commerce-1
• Velou developer documentation
• Algolia acquisition announcement

Explore OfferSignal · Focused Business & AI Workshop · All research

References