Skip to content
Branemind
All case studies
Case studyManufacturing · Ecommerce··5 min read

Search that understands a part, not just its name

Forty thousand SKUs, half of them near-identical, and buyers who search by symptom rather than part number. A reranker and a graph fixed what keyword search never could.

Search-to-cart
+41%
Zero-result queries
−78%
Support tickets
−1/3
'which part do I need'

Measured against a baseline captured before the work started. One engagement, not a forecast for another business. See the methodology note below.

A dental technician does not search for 'high-speed handpiece turbine cartridge, 4-hole'. They search for 'drill making a whining noise'. Keyword search returns nothing, so they call support, and support reads them a part number off a laminated sheet.

Three layers, in order

  1. Normalise the catalog

    Every SKU gained a structured spec: fitting, compatibility, consumable class, failure symptoms. Most of this already existed in PDFs nobody had parsed.

  2. Retrieve hybrid, rerank hard

    BM25 for the people who do know the part number, dense retrieval for the people who don't, and a cross-encoder to settle the argument between them.

  3. Walk the graph

    Compatibility and supersession live as edges. 'This part is discontinued, here is what replaced it' is a graph traversal, not a prompt.

Zero-result queries, the clearest signal that search is failing, dropped 78%. The support team stopped being a lookup service and went back to being a support team.

Keep reading

All insights
Case study
Finance & accounting·26 Aug 2026·8 min

Books that close every day, not every quarter

An agentic workflow that reconciles transactions against invoices continuously, keeps a live read on cash and working capital, and leaves the books audit-ready at any hour. It sits on top of Tally Prime and Zoho Books rather than replacing them.

A multi-entity Indian services group
Case study
Customer experience·9 Jul 2026·6 min

A voice agent that holds its own in nine Indian languages

Support in English served maybe a third of the callers. Sovereign Indic models and a hard latency budget served the rest, including the ones who switch language mid-sentence.

A consumer services company operating in nine Indian languages
Case study
Fintech · Credit·18 Jun 2026·7 min

An underwriting copilot that took decisions from days to minutes

A lending team was drowning in manual file review. We put an agent in the loop that reads every source, argues its own case, and hands the underwriter a decision worth signing.

A mid-sized Indian NBFC

Methodology and privacy

  • Clients are described rather than named. Where a quote appears it is attributed to a role and published with approval.
  • Figures come from a single engagement and are not a forecast of what a different business would see. Volume, data quality and process maturity move them more than the technology does.
  • Baselines were captured before the work started. The measurement window and sample for any figure on this page are available on request.
  • Where a number is illustrative rather than measured, it is labelled as such in the text.
If this looks like your loop

Scope one customer conversation flow

Tell us the flow you want to automate and see where it sits on readiness.

Scope one customer conversation flowOr book a 30-minute call