‹ All work
Enterprise SaaS · Global Search,
Bectran,
2026
Designing one search across 125M+ financial records
A unified way to find customers, credit requests, accounts, and related records across 59 searchable fields. Used by 2,400+ people in its first two weeks.
Company
Bectran
Role
Lead Product Designer on feature
Team
Product and engineering
Timeline
October 2026
Platform
Enterprise SaaS · Global Search


The short version
1 min
Problem
The old search made users choose a record category before entering a query. Relevant details inside records were difficult to find, and matches lacked context.
What I did
I designed a unified search panel, organized results by record type, made matches understandable, and documented 14 implementation scenarios with engineering.
Result
In the first two weeks after the October 30, 2026 launch, 2,400+ people interacted with the new search 81K+ times. About 89% of daily searchers chose it over the old search bar, a share sustained through week two.
My role
Lead Product Designer on feature
Context
Across credit and AR workflows, analysts needed to find customers, credit requests and ERP accounts using whatever information they had: a company name, person, email, phone, address or identifier. The previous category-first search forced users to know where to look before they could search.


Old search experience - users had to select a record category and search field before entering a query.
My role
I led the feature UX from exploration through engineering handoff. I designed the search panel, result grouping, field-level match explanations, and states, then partnered with engineering on field mapping and 14 implementation scenarios.
Research
I mapped common lookup journeys alongside less predictable cases: a single person appearing across multiple records, a customer matching through a related contact, several matching fields within one record, and exact-match identifiers. The goal was not simply broader search coverage; it was a result users could trust.
59
searchable fields
14
engineer-ready cases
4
record types
Search started with a decision, not a query
Users needed to select a record category before they could begin searching, even when they were unsure where a record belonged.
Results did not always explain the match
A request could match a contact or guarantor email that was not visible in its main title, making the result difficult to interpret.
More fields meant more edge cases
Searching 59 fields raised questions about duplicate records, exact identifier matches, hidden data, and keyboard behavior.
Approach
The design principles were simple: one entry point across record types; reveal the field that caused each match; offer category filters without forcing an upfront choice; and make ambiguous, sensitive and empty-result cases behave predictably. These rules were translated into 14 engineering-ready scenarios across 59 indexed fields.
Let people search first
Replace the category-first flow with one search entry point and filters that remain available after searching.
Show why each result matched
Display matching customer or request details beneath the primary record label to support confident scanning.
Make complexity predictable
Document deduplication, exact matching for sensitive identifiers, keyboard navigation, and empty states across 14 scenarios.
The complexity was not the search box itself. It was defining predictable behavior across different record types, matching fields, identifiers, filters and edge cases before engineering handoff.

Search behavior across 14 scenarios — from names and contact details to identifiers, multiple matches, filters, and empty states.
Solution
The redesigned search brings customers, credit requests, consolidated customers and ERP accounts into one panel. Results are grouped by record type, with counts, status labels and the exact field responsible for a match. From there, users can narrow the list with category filters, recognize multiple matching fields within a record, and recover from a search with no results.


01 / Explain the match — A person-name search highlights the matching contact field beneath each request, alongside its status and key details.


02 / No results — Clear feedback and guidance help users refine a query instead of reaching a dead end.


03 / Filter by record type — The Requests tab narrows the results while preserving the search query and match details.


04 / Multiple matches — A partial name can match several roles or fields within the same record, without duplicating that result.
Consistent behavior across real-world search cases
The system accounts for searching by names, email addresses, phone numbers, company names, addresses, DUNS numbers, request IDs and account identifiers. Results expose why they matched; repeated fields within the same record are summarized rather than shown as duplicate entries. Exact-match rules protect sensitive identifiers, and empty states explain how to try again.
Result
of daily searchers chose new search
2,400+
people using new search in two weeks
−48%
weekly users of old search
Global Search launched October 2026. In its first two weeks, 2,400+ people used the new experience, with 81,000+ search-box interactions. About 89% of daily searchers chose the new panel over the old bar, steady through week two. Roughly 90% of people who received it did not switch back. Weekly users of the old search fell from 613 to 318 (−48%), and median response time was 0.18 seconds. Interactions count search-box use, not confirmed completed searches; these adoption and performance metrics do not establish time savings.
What I learned
I went into this thinking the hardest part would be making so much information searchable. What changed my thinking was realizing that finding a result is only half the job—people also need to understand why it appeared. I learned to challenge seemingly small interaction details early, especially when different records and sensitive fields can behave in unexpected ways. Now I think about search less as an input and more as a system for building trust.
