What clients say about working with Infrawise
Practical feedback from platform operators and product teams across Malaysian e-commerce, publishing, and content businesses.
Back to HomepagePlatforms served across Malaysia
Average client satisfaction rating
Projects delivered on agreed scope
Combined recommendation AI experience
Direct feedback from clients
"We came in unsure whether our data was sufficient for a proper recommendation system. The team was honest about what was workable and scoped the project sensibly. The resulting engine increased product page CTR noticeably, and the A/B test framing meant we could actually attribute that to the recommendations."
Yong Chong Wei
Product Lead · Kuala Lumpur marketplace
March 2026
"The content personalisation work took slightly longer than initially estimated because we had some data quality issues on our end — but the team was upfront about that when it came up and adjusted the plan accordingly. The final personalised article feed has meaningfully improved time-on-site for returning visitors."
Nur Izzati Mahmud
Head of Digital · Selangor media publisher
February 2026
"What I appreciated most was that they didn't oversell what recommendation AI could do for a platform at our stage. We had a candid conversation about our data volume and ended up with a hybrid approach that made sense for our catalogue size. It performs well and our team can maintain it."
Rajesh Pillai
CTO · Penang B2C e-commerce platform
January 2026
"We've been on the Analytics & Tuning retainer for four months now. The monthly reports are actually readable — they explain what changed, why, and what was adjusted. It's not a dashboard dump. Cold-start handling for new SKUs has been the most tangible improvement for our catalogue team."
Siti Balkis Razak
E-commerce Manager · KL fashion retailer
March 2026
"Our content discovery problem was specific to a bilingual feed — Malay and English articles mixed with different engagement patterns. The Infrawise team took that seriously rather than treating it as a minor data formatting issue. The profiling model handles the mixed-language catalogue well."
Lim Hwee Chin
Senior Editor · Bilingual news portal
February 2026
"We integrated the recommendation API into our product listing pages without needing much back-and-forth. The documentation was clear enough that our own developer could handle it. That's a higher bar than it sounds — previous vendor deliverables have needed a lot of interpretation."
Ahmad Faizal Hasan
Lead Developer · Johor Bahru marketplace
January 2026
Three platform stories
Multi-vendor marketplace, 8,000 SKUs, low product discovery
A Klang Valley marketplace with multiple product categories had strong search usage but low browsing discovery. Most users never went beyond their first-page search results, and cross-category exploration was minimal.
Hybrid collaborative and content-based engine
With 14 months of purchase and browse history available, a hybrid model was trained to surface items based on similar-user purchasing patterns supplemented by category affinity scoring. Cold-start handling for new vendor SKUs used catalogue embeddings.
+28% cross-category click-through over 8 weeks
A/B testing across a 50/50 user split showed a 28% lift in cross-category CTR and a 12% improvement in average session depth. New vendor SKUs reached recommendation visibility within 48 hours of listing.
Timeline: 10 weeks · Service: Product Recommendation Engine
"We had been discussing recommendation features internally for a long time but didn't have the internal ML capacity to build it ourselves. The Infrawise engagement gave us something we could actually maintain going forward."
Digital publisher with 400+ articles/month, falling return visits
A Kuala Lumpur-based digital publication was publishing high volumes of content but seeing declining return visit rates. Most readers arrived via search, read one article, and left. The homepage feed was the same for all users.
Content personalisation with automated tagging
User interest profiles were built from article engagement history. An automated content tagging pipeline kept new articles classified without editorial overhead. Real-time personalisation was served through an API integrated with the existing CMS.
+19% average session depth for returning users
Returning visitors exposed to personalised feeds averaged 19% more article views per session compared to the control group. Newsletter open rates for personalised digest emails improved by 11% in the same period.
Timeline: 7 weeks · Service: Content Personalisation System
Recommendation engine deployed but performance declining after 6 months
A fashion e-commerce platform had deployed a recommendation layer with a previous provider but noticed CTR declining month-on-month. The system wasn't keeping pace with catalogue growth and was over-recommending established best-sellers.
Analytics & Tuning engagement — diversity and cold-start focus
Monthly metric reviews identified that new SKUs weren't entering the recommendation pool adequately and that diversity scores had dropped significantly. Cold-start handling was adjusted and diversity-relevance weighting rebalanced over a 3-month period.
CTR stabilised and catalogue coverage improved by 34%
CTR recovered to previous levels within 6 weeks. Catalogue coverage — the proportion of items receiving at least one recommendation impression per week — improved from 38% to 72% over three months of tuning.
Ongoing · Service: Analytics & Tuning
Speak with us directly
If you'd like to discuss your platform's situation before committing to an engagement, a brief introductory call is always available.
Phone
+60 3-5748 2316Address
40 Jalan Imbi, 55100 Kuala Lumpur
Working Hours
Mon–Fri: 9:00 AM – 6:00 PM MYT
Add your platform to this list
If the work described here sounds relevant to what you're building, we'd be glad to have a conversation about your situation.
Get in Touch