Xumulus Logo
Agent-ready search & discovery on Next.js + Elastic/OpenSearch

E-commerce strategy consulting

Most strategy engagements produce a document. The useful ones produce a decision you can defend to a board and hand to an engineering team in the same week.

Elastic
Relevance + speed
Next.js
Modern UX + performance
RAG
Agentic discovery ready
AI Discovery Signals
Retrieval-first architecture makes agentic discovery possible.
Scroll

What we are usually asked

Three questions account for nearly all of it: what platform should we be on, why does our catalog fight us, and why does search convert worse than the category pages. They look independent and are usually the same problem seen from different seats.

We answer them against your data — catalog exports, query logs, Search Console coverage, conversion by landing type. An opinion formed without those is a preference, and you can get preferences for free.

The stack

Where the work concentrates

Three areas. Most engagements are two of them.

01
1) Platform

Selection and total cost

Including the option of staying where you are, which is frequently correct and rarely recommended by anyone being paid to replatform you.

  • Adobe Commerce, BigCommerce, Shopify Plus and headless compared against your catalog, not in general
  • Migration cost modelled honestly, including merchandising and integration rebuild
  • Metered costs projected forward — search requests and records both bill
02
2) Data

Catalog and product information

Most storefront problems are catalog problems wearing a storefront costume.

  • Attribute model reviewed against what merchandising actually needs to do
  • PIM case assessed on the evidence, Akeneo where it fits
  • Feed and sync architecture, including where copies drift out of step
03
3) Discovery

Search and browse audit

Run against your query logs, so the findings are about queries your shoppers actually type.

  • Zero-result and low-conversion queries ranked by revenue at stake
  • Facet coverage measured against catalog attributes
  • Relevance, synonyms and merchandising rules reviewed as a system

How we work

Short engagements with a named deliverable. We would rather give you a decision in three weeks than a programme in three months, and we will tell you when the answer is that nothing needs to change.

What you get back is a sequence, not a wish list: the work ordered by dependency rather than by enthusiasm, build-buy-defer called explicitly on each item, and enough detail that an engineering team can start from it without a second discovery phase.

Bring us the question you are stuck on

Platform selection, a catalog that will not behave, or search that underperforms the rest of the site. Tell us which, and what data you can share.

Search is the foundation of AI commerce