canvas.getoncourse.ai — Launch Report

PostHog web analytics · scoped to host canvas.getoncourse.ai only
Generated 15 Jun 2026
Live since 11 Jun 2026 · 5 days of data
Project Oncourse · PostHog #61503
Visitors (5d)
254
unique persons
Pageviews
950
~3.7 / visitor
Peak day
146
12 Jun launch spike
Install CTA clicks
7
9 clicks
Errors
2
exceptions · stable
01

Traffic since launch

Daily visitors & pageviews

Classic launch curve — a spike on 12 June (shared links), tapering over the week

Unique visitors Pageviews

How people arrived

Almost entirely direct / shared links via chat — no organic search engine yet

02

What visitors did

Reach = unique users firing each action (of 254 visitors). Many arrive directly on a shared canvas link rather than browsing the gallery, so "opened a canvas" runs ahead of "clicked a gallery tile". Deep engagement (share / like / copy) and install intent are still small — the early funnel to grow.

Engagement reach — unique users

Landing → canvas → interaction

Product events — last 30 days

Custom canvas instrumentation, by volume & unique users

03

Performance — Core Web Vitals (p75)

Interactivity and layout stability are excellent, but loading is the weak spot: largest contentful paint of 5.1s and first contentful paint of 3.9s are both in Google's "poor" band — felt most by the mobile majority.

04

Audience

Device mix — unique users

Mobile-first audience (~68%) — load speed matters even more here

Top countries — unique users

Heavily India + US — matches where the launch was shared

05

Findings & opportunities

Slow page load FIX FIRST

p75 LCP 5,105 ms and FCP 3,936 ms are both in the "poor" range (good is <2.5s / <1.8s). With ~68% of visitors on mobile, a 5-second largest paint is a real bounce risk on a first impression.

Evidence: $web_vitals = 638 samples / 203 users · INP 166ms (good) · CLS 0.004 (excellent)
Action: The gallery & canvas pages are force-dynamic (server-rendered every request, so TTFB is paid each visit) — data reads are cached but the pages aren't; consider ISR/full-page caching for the gallery. Then profile the LCP element: lazy-load below-the-fold tiles, preconnect to the thumbnail CDN, and serve right-sized previews.

Organic not kicked in yet — but the SEO stack is already built EXPECTED

Traffic is 95% direct/shared with a single Google referral — normal for a domain that's only four days old and not yet indexed. The discovery plumbing is already shipped & verified in production: a 234-URL sitemap, per-canvas Open Graph + Twitter cards + canonical, and JSON-LD LearningResource structured data all render live.

Verified live: sitemap.xml = 234 URLs · og:image 1200×800 · twitter summary_large_image · JSON-LD · robots → sitemap
Action: Nothing to build — submit/confirm the sitemap in Search Console, validate a card in a sharing debugger, then give indexing a few weeks. Optional: an "explore more" loop from a viewed canvas back to the gallery to lift on-site discovery.

Shallow engagement funnel ACTIVATION

Of 254 visitors, only 43 browsed the gallery, 26 opened a canvas, 5 interacted (share/like/copy) and 7 clicked install. Search is barely used (4 users). Most landings don't turn into a canvas open.

Evidence: canvas_gallery_viewed 43u · canvas_viewed 26u · canvas_tile_clicked 20u · install_cta_clicked 7u
Action: Make the first canvas one tap from landing; surface like/share/copy more prominently; test a clearer install prompt after a canvas view.

Rock-solid stability HEALTHY

Across 254 visitors and 2,753 events there were just 2 exceptions and 2 rage-clicks, with excellent CLS (0.004) and INP (166ms). The app itself is behaving well — the issues above are about speed and reach, not bugs.

Evidence: $exception 2 · $rageclick 2 · CLS 0.004 · INP 166ms
06

Recommended priorities

1
Cut LCP/FCP toward <2.5s.

Biggest lever on first impression for the mobile majority — lazy-load tiles, right-size canvas previews, trim blocking JS/fonts.

2
Let SEO mature — it's already wired.

Sitemap, per-canvas OG/Twitter/canonical and JSON-LD are live. Submit the sitemap in Search Console, validate cards, and give indexing a few weeks — no build work needed.

3
Tighten landing → canvas activation.

One-tap into a featured canvas; "explore more" loop back to the gallery to lift the 10% open rate.

4
Track return visits & install conversion.

5 days in, retention is unknown — instrument D1/D7 return and install-CTA → installed to learn if it sticks.