Conversion Dashboard

*Checkout Started Rate = Checkout Starts ÷ Lander Sessions • Checkout Completion Rate = Purchases ÷ Checkout Starts • Lander Conversion Rate = Purchases ÷ Lander Sessions

By lander

Top 10 by Sessions, last 7 full days. Spend = Meta + Google Ads + Everflow. Deltas: last 7 full days vs the 7 before. Sparkline: daily Checkout Started Rate; the dashed line is the average rate across all landers (Overview tab) or the product's landers (product tabs) over the last 30 days. Hover over a sparkline for each day's value.

Daily by lander

Metric

Full days only (America/Los_Angeles); the current day is never shown. Shading scales with the value within each column. Rates are hidden (—) on days with fewer than 20 sessions (checkout completion rate: fewer than 5 checkout starts).

Purchase options

Conversion by device

Landers

By device

By operating system

By device and operating system

Where the page is losing people

Sorted by sessions. Rage / dead clicks per 1,000 sessions; scroll depth in CSS px from the top; Core Web Vitals at p75. Hotspot = the 500 px band with the most rage + dead clicks; open the page's heatmap in PostHog to see which element sits there. Dead-click share = (rage + dead) ÷ all clicks in that band. Hover a column heading for its definition; all terms are defined under Methods & Glossary.

Needs attention

Every flag uses the last 7 full days. Checked per device class; Mobile = phones + tablets. Rules under Methods & Glossary.

Desktop

Mobile

Methods & Glossary

Source dashboard in PostHog

Lander names
Which landers count as live: The destination pages of ads that spend in the Meta ad account (Meta Ads API), plus lp-glp1-stack-c and lp-glp1-stack-d, which carry paid traffic from outside Meta (Google Ads / Everflow), and lp-glp1-results-a, the quiz results page. lp-magnesium-a is linked by ~160 ads but redirects to lp-magnesium-b, so its spend is folded into -b. Any lander that appears in the data but is not classified below gets an "Other" tab automatically.
Product tabs: Collagen: lp-amc-ff-a, lp-amc-ff-b · Magnesium: lp-magnesium-b · Fiber: lp-fiber-ff-a/b/c/d · GLP-1 Bundles: the quiz funnel lp-glp1-quiz-a + lp-glp1-companion-a → lp-glp1-results-a · The Elevate Stack: lp-glp1-stack-a/b · GLP-1 Solo: lp-glp1-stack-c/d · Warehouse: lp-blowout-b. Each tab shows only its landers' sessions, checkouts, purchases and spend.
Quiz funnel group: lp-glp1-quiz-a and lp-glp1-companion-a both send visitors through a quiz to lp-glp1-results-a, where the purchase happens, so the three pages are reported as one lander ("GLP-1 Quiz"): sessions are counted on the two entry pages, checkouts and purchases on all three. Which entry page drove a given purchase can't be told apart, so they are credited to the pair. Friction metrics stay per page.
Spend: "Spend" on the lander cards is the total of every ad source read: Meta, Google Ads and Everflow. Everflow: affiliate payout by offer and day from the Everflow network API (the same payout figure the Performance Dashboard uses), credited to the lander in the offer's destination URL; offers pointing at retired landing pages are ignored. Google Ads: daily campaign cost from the GA4 property's linked Google Ads account, assigned to landers by each campaign's landing-page sessions (the four non-brand campaigns each land on one lander; Branded Search, PMAX, Demand Gen retargeting and YouTube prospecting don't land on a lander and are left out of every tab). Meta: ad-level daily spend from the Meta Ads API for every ad set that spent in the window (active and since paused), assigned to a lander by each ad's destination link. Ads whose link can't be read (older creatives beyond the ad-list cap) are assigned to a product by their ad name and counted in that product's totals, but not in the per-lander cards. The small remainder Meta no longer breaks out by ad (deleted ads) is included in the Overview only.
Steps: Session: a PostHog session with a pageview on the lander, excluding crawler sessions (below). Checkout started: sessions firing checkout_started with that lander attributed. Purchase: unique checkout tokens with checkout_completed for that lander. Checkout events fire from the Shopify web pixel in a separate session, so purchases are counted per checkout, not per lander session; express checkouts can complete without a start event, which is why Checkout Completion Rate can exceed 100% on small counts.
Crawlers excluded: Since 2026-09-29, sessions from three kinds of automated visitor are left out of every session count, rate, friction and device figure on this page: headless browsers that report an 800×600 screen and a UTC clock (run from Hetzner, WorldStream, AWS, Google Cloud and Palo Alto Networks servers); Meta's ad-review crawler (Facebook's own network, 2a03:2880::/32); and scripted browsers on Hetzner's cloud networks that report a non-American clock. Network owners were checked against the RIPE and ARIN registries. None of the 6,735 matching sessions from 8/18 to 9/29 ever reached checkout, so checkout starts and purchases are unchanged. In the last 7 days they were about 6% of desktop sessions and 2% of mobile sessions across all landers. Traffic that is likely automated but not proven, such as visits reporting a UTC clock from small US internet providers or Microsoft Azure addresses reporting a Los Angeles clock, is still counted.
Purchase options: Collagen, Magnesium and Fiber tabs only. Each checkout's cart (products, quantities and the subscription flag sent by the Shopify pixel) is matched to the option the lander sells. Collagen: 1-Bag subscription = one bag on subscription; 3-Bag bundle = three bags, one-time; 1 Tub and 2-Tub bundle = one or two tubs, one-time (the bundle adds a free BlenderBottle). Magnesium: 1-, 2- and 3-pack subscription = one, two or three bags on subscription (the free coaching, bottle and stick-pack gifts are ignored). Fiber: 1-, 2- and 3-pack subscription = one, two or three canisters on subscription. Checkout starts and orders are counted as in Steps above; a checkout start can't be matched to its order, so each is counted separately. In this table Checkout % = orders ÷ checkout starts for the option (checkout completion rate), hidden under 5 starts. Source: PostHog insight Xyb2X6OD.
No add-to-cart step: The site does not emit an add-to-cart event yet. When one is instrumented it slots in between session and checkout.
Device classes: Conversion by device uses PostHog's device type and OS from the browser (Mobile / Tablet / Desktop). The Needs Attention tables fold Tablet into Mobile. Rage and dead clicks per device come from the heatmap data, which records the viewport instead of the device: 1,024 CSS px and wider counts as Desktop, narrower as Mobile.
Attention rules: Every rule looks at the last 7 full days (L7). Checked separately for Desktop and Mobile: zero checkout starts on 150+ sessions · checkout started rate down 30%+ versus the 7 days before, with 300+ sessions in both weeks · desktop starting checkout at under half the mobile rate on 200+ desktop and 200+ mobile sessions · 30+ rage clicks or 80+ dead clicks per 1,000 sessions on 500+ sessions. Lander-wide rules, listed in both tables: spend over $1,000 with fewer than 5 purchases · more than half of sessions leaving inside 10 seconds on 500+ sessions.
Windows: Days are America/Los_Angeles and only full days are shown; the current day never appears anywhere on this page. Prior Day = the last full day; L7 / L14 / L30 = the last 7 / 14 / 30 full days; "vs prior" compares with the equal-length window before. The Daily, device, friction and purchase-option tables follow their own period toggles (default L14).

* Tracking on the GLP-1 quiz funnel is a work in progress: lp-glp1-companion-a and lp-glp1-quiz-a hand off to lp-glp1-results-a, and the results page is currently recording only a handful of sessions and checkouts, so the GLP-1 Bundles numbers are low until that instrumentation is fixed.

* PostHog data begins 2026-08-18, so 30-day comparisons have no full prior period until mid-October.

* Roadmap, to be completed before this dashboard build is closed out: move the page to a hosted URL of its own (like the Performance Dashboard) so it is served without the claude.ai artifact bar.