Client stories
After the readout
These notes come from talent and product people who sat through a readout. They name screens and counts. They are not star ratings.
We had been celebrating a jump in screening volume. The review showed that most of those events fired when a graduate opened the question set, not when they answered the last item. The actual finish rate on Android was far lower, and that is where we spent the next sprint.
Talent operations lead · Graduate hiring app, Kuala Lumpur
The sourcing table was the part I used the next day. Our loudest job-board tag produced applications; the referral code produced interviews that people actually attended. I still wish the first pass had included the Malay-language apply flow instead of treating it as a follow-up.
Head of talent · Retail hiring app, Petaling Jaya
Lim sat with our live booking screen and watched the ‘interview booked’ event fire when a candidate picked a slot that the hiring manager had already cancelled. We had been nagging managers about no-shows that were, in part, ghost slots.
Product owner · In-house hiring app, Penang
The findings pack was blunt about our CV parse. PDF scans from polytechnic applicants failed silently, and those people never reached submit. We knew parse errors existed; we did not know they accounted for most of the mobile drop between attach and submit.
Campus programme manager · Apprenticeship hiring app, Shah Alam
A longer note from a warehouse hiring app
A Klang Valley operator asked us to read the apply path used by warehouse and dispatcher roles. Candidates mostly arrived from Facebook job posts and a QR code at factory gates. The weekly report showed healthy application volume and a mysterious shortage of people at Sunday interview sessions.
We reconstructed events for four weeks. Apply-started was firing on the landing screen, including people who opened the QR, saw a three-page Bahasa Malaysia consent wall, and left. Submit was honest. Interview-booked counted calendar holds that the shift supervisor later dropped without a cancellation event. Offer-sent was almost unused because offers still went out on WhatsApp.
The readout was uncomfortable. The talent lead had promised leadership a “full funnel.” What they had was a landing-page counter and a WhatsApp thread. The useful outcome was narrower and better: they stopped reporting apply-started as demand, they shortened the consent wall to a single screen with a downloadable PDF, and they asked us back only after the booking vendor could emit a cancellation.
That engagement was a standard measurement review, one path, ten working days. We did not touch the WhatsApp offers; that sat outside the app, and we said so in the pack.