Client notes
Specific feedback from language product teams who sat through audits, workshops, and advisory months with us.
“They caught that our ‘daily active’ count was inflated by passive vocabulary reviews. Once we separated intentional lesson starts, the week-two drop looked very different — and so did our roadmap.”
“The workshop forced us to pick twelve events instead of forty. Painful for a week; then our cohort charts finally matched what tutors were seeing in support tickets.”
“Nadia’s readout spent more time on what we should ignore than on vanity wins. We still argue about speaking-task completion, but at least we argue from the same definition.”
“Advisory months are useful when we ship; quieter when we do not. I wish we had paused the retainer during our content freeze — that one’s on us, not on Sitecornerhub.”
“Farid refused to treat streak length as a success metric until we proved learners returned after breaking a streak. That restraint saved us from a foolish gamification push.”
Engagement story: week-two collapse on a Thai path
A Southeast Asian language app saw strong day-one lesson starts for Thai beginners, then a sharp fall by day fourteen. Sitecornerhub’s audit showed the drop concentrated after the first tone-drill unit — not after payment walls. Passive “listen along” events had been counting as practice, so the funnel looked healthier than tutor feedback suggested.
The team reprioritised instrumentation for intentional tone exercises and moved a speaking checkpoint earlier. Three releases later they still used the same warehouse; they simply trusted a narrower set of learner-state definitions. The full engagement took just under four weeks from data access to readout.
Engagement story: framework before a rebuild
A bilingual app preparing a schema rebuild booked a metric framework session before engineers renamed everything. In one day the group agreed primary metrics for lesson mastery versus casual practice, and deferred six “nice to have” events that would have duplicated each other. The rebuild then shipped with fewer aliases — and the later retention audit took less inventory time because naming was already shared.