Client experiences
Outcomes matter more than promises. Here's how organisations across Hong Kong describe working with Verdian.
Back to Home47+
Engagements completed
4.8
Average client rating
94%
On-time delivery rate
6
Years serving HK businesses
Direct feedback
"Our procurement team was making decisions based on data that was at least two weeks stale. After Verdian connected our ERP and warehouse systems, we had a live view for the first time. The disruption alerts alone saved us from a significant over-order situation in November. The dashboard is clean enough that operations staff actually use it."
Wong Lai-Shan
Operations Director · Electronics Distribution, Kwun Tong
January 2026
"We needed a voice interface for our kiosk product that could handle Cantonese, Mandarin, and English from real shoppers — not controlled lab speakers. Verdian's testing approach was thorough and I appreciated that they surfaced issues during the process rather than at the last moment. The final product performs well in a noisy retail environment. One minor note: timeline slipped by a week, though they communicated about it clearly."
Chan Tsz-Ho
Head of Product · Retail Technology Startup, Causeway Bay
December 2025
"We were deploying a credit-scoring model under SFC scrutiny and needed a proper compliance review before it went live. The gap analysis was specific and actionable — not a generic checklist. Three of the identified issues would have been flagged in an actual audit. The remediation roadmap helped us prioritise what to fix immediately versus what could follow."
Michael Kwok
Chief Risk Officer · Financial Services Group, Central
January 2026
"What I valued most was the lack of theatre. They came in, understood how our logistics data was actually structured (a mess), and worked with it rather than telling us to clean everything before they could start. The integration was genuinely thoughtful. Six months in and the demand forecasting is still running without us needing to touch it."
Lam Chi-Fung
Logistics Manager · Import/Export Company, Kwai Chung
December 2025
"We contracted Verdian for the AI compliance review as part of a broader PDPO readiness exercise. The review identified data retention gaps we hadn't considered and flagged a third-party data sharing arrangement that needed restructuring. The written report was detailed enough to hand directly to our legal counsel. Will engage them again for our next model deployment."
Yu Nga-Man
Data Protection Officer · Healthcare Group, Kowloon
January 2026
"We tried building the voice interface in-house and ran into accent robustness problems we couldn't solve. Verdian took three weeks to diagnose the issue and four more to fix it. Honest about what was going wrong and methodical in how they addressed it. Our app now handles our customer base without the embarrassing misrecognition errors we were seeing."
Derek Tong
CTO · Property Technology Platform, Wan Chai
December 2025
Detailed outcomes
Challenge
Data lived across three separate ERP modules and a shared spreadsheet no-one fully trusted. Procurement was making decisions on instinct, and two significant overstock situations had occurred in the preceding year. Inventory write-downs were affecting margin.
Solution
Verdian built an integration pipeline connecting all three ERP modules and the warehouse management system into a unified data layer. A category-level demand model was trained on 36 months of historical sales data and calibrated against seasonal patterns specific to the Hong Kong consumer electronics market.
Results
Within the first quarter of operation: overstock value reduced by 28%, one disruption early warning triggered that prevented a missed replenishment before the Lunar New Year period, and procurement team decision time on weekly orders dropped significantly. Dashboard is updated every 4 hours.
Engagement duration: 9 weeks · Completed: December 2025
Challenge
The internal team had built a voice search feature using a standard ASR API. In testing, recognition accuracy for users with Cantonese-accented English fell below 70% — unusable in a public-facing product. Retraining the base model was beyond the team's scope, and the product launch was delayed.
Solution
Verdian diagnosed the issue — the base model lacked sufficient training data for the specific accent range — and implemented a fine-tuning pipeline using real user recordings collected under consent. A noise robustness layer was added for open-plan office environments. Intent mapping was redesigned to handle mixed-code input.
Results
Recognition accuracy for the target user group improved from below 70% to 91% in independent testing. The voice feature launched publicly in January 2026 without the misrecognition issues previously observed. User engagement with voice search is now measurable and positive.
Engagement duration: 7 weeks · Completed: January 2026
Challenge
The group was deploying an ML-based credit scoring model under SFC oversight. Internal legal counsel identified that the model documentation did not meet the level of explainability the regulator expected. A pre-audit review was needed before the formal submission date.
Solution
Verdian conducted a four-week review covering: model documentation, data lineage, bias testing across protected attribute proxies, explainability of individual decisions, and third-party data processing agreements. Findings were organised by regulatory reference and severity tier.
Results
Three high-priority gaps were identified and remediated before the regulator submission. The written report was incorporated into the formal compliance file. The submission proceeded on schedule without requests for further information from the regulator on model governance.
Engagement duration: 5 weeks · Completed: January 2026
Get in contact
Phone
+852 5428 7163Office
42 Hennessy Road, Wan Chai, Hong Kong
Office Hours
Monday – Friday: 9:00 AM – 6:30 PM HKT
Your turn
Start with a conversation. We'll tell you honestly whether there's a fit before anything is scoped or committed.
Get in Touch