Reading the dashboard

The analytics dashboard tracks how your published experiences perform. Data appears once visitors start opening your models — a fresh workspace shows empty states with guidance, not zeros pretending to be insight.

Views

• Daily views trend — momentum over the selected period.
• Period comparison — this period against the previous one of equal length.
• Top viewed projects and top viewed models — where attention concentrates. A project groups models a shopper never sees apart, so the two rankings answer different questions.

Sessions & engagement

Views say that people came; sessions say what they did:

• Model opens, AR launches and total sessions — the three counts everything else is read against.
• AR share of opens — how much of the attention becomes an in-room preview.
• Time on model — the average time a shopper kept the model on screen after it rendered. Only visible time counts: a hidden tab pauses the clock. Sessions shorter than a second and longer than two hours are not counted.

Traffic sources

Every model open records where the shopper came from:

• Referring sites — the site the shopper was on before your page. Direct or QR is an open with no referrer: a typed link, a QR scan, a message.
• Embedded on — which of your pages hosted the viewer.
• Campaigns — the utm_source on the link that brought the shopper. Add utm_source, utm_medium and utm_campaign to campaign links exactly as you would for any web analytics.

For an embed, the referrer and campaign are read from the page hosting it; for a direct viewer link or QR code, from the viewer URL itself.

AR launches

A launch is the last moment we can observe on native AR, and the dashboard says so plainly rather than inventing engagement numbers it cannot source.

• AR launches and launch rate — how often a viewer becomes an in-room preview.
• Launches by device — iOS (Quick Look), Android (Scene Viewer), other (desktop / WebXR).

Why there is nothing after the launch

iOS Quick Look and Android Scene Viewer are native operating-system apps. Once AR opens, your web page is no longer running, and neither Apple nor Google offers a callback to it. Photos, recordings, sharing and time spent inside native AR are therefore invisible to every web platform — not just this one. Any tool reporting “AR engagement” is counting button taps.

WebXR is the exception. It keeps the session inside the page, so for WebXR launches the dashboard reports real session duration, completion, and in-session photos — labelled as WebXR-only, never blended into the AR total.

The one thing native AR can report: the add to cart

A shopper in native AR is invisible to you. Adding to cart is the one action they can take that you get to see — a real conversion, not inferred engagement. The dashboard shows it as “Added to cart from AR”, and it counts toward the configurator funnel’s enquiries alongside the summary panel’s quote request button.

Enable it per project in WebAR settings → AR Add to Cart. The button text defaults to Add to cart; blank it and there is no button, so it stays off.

The shopper’s tap reaches your page as a single togenar:ar-add-to-cart event on the embed — the same event, the same payload, on both platforms. Write one handler and wire it to your cart; see Add to cart.

It renders on two surfaces

	Where the button appears
iOS (Quick Look)	Inside the AR session
Android (Scene Viewer)	The moment AR closes

The dashboard splits the taps by surface so you always know which one you are reading. The tap rate is over all native AR sessions, not iOS alone.

On iOS it is not free. The banner takes the slot Quick Look uses for its photo button, so turning it on means iOS shoppers can no longer screenshot or share the model from inside AR. You are trading organic reach for a measurable conversion. That’s why it is off by default — it’s a merchandising call. (Android loses nothing: its button was never inside AR.)

iOS Safari only. Chrome, Firefox and Edge on iOS ignore rel="ar"’s banner, so no button is drawn there — AR still opens, it just cannot be measured.

Configurator funnel

The part a merchandiser actually acts on. Requires the consent="analytics" embed attribute — see Attributes reference — and is off until you set it.

• Funnel: viewed → configured → summary opened → enquired. Counted in distinct sessions, not events: one shopper flipping through nine fabrics is one person considering a sofa, not nine. Enquired counts both ways of asking — the summary panel’s quote request button and the add to cart from AR — and splits them underneath, so you can see which surface is doing the work. An AR tap from a session without consent="analytics" is reported in the AR section instead of the funnel.
• Most-picked options — the share within each part (“62% of everyone who chose a fabric chose linen”). This is what tells you which options to stock, promote, or retire.
• Abandoned configurations — sessions that configured something and then left. A product gap, a price surprise, or a missing option: each one a recoverable revenue signal.
• Enquiry rate — of the shoppers who configured, how many asked for it.

Revenue

What the 3D actually earned, when your storefront reports it back — see Revenue tracking.

• Verified revenue, orders and cart adds — from orders confirmed by the Shopify webhook; orders your page reports with purchased() are listed separately as unverified. Counted for sessions that opened the model, not for your whole catalogue.
• Revenue per session — the figure that answers whether the 3D paid for itself.
• Revenue by channel — orders grouped by the campaign or referring site the shopper arrived from.

Amounts in different currencies are never added together; each is totalled on its own. The block stays empty until your storefront reports its first cart or order.

Everything here is about your shoppers

Every number on this page is filtered to events a visitor produced. Your own work in the editor — uploading models, opening projects, previewing scenes — never counts as a session, a view, or a load.

Model reliability is measured on the shopper’s device and network, not on yours. A model that a visitor waited for and never got is recorded as a failed load (and is never billed as a view), so the success rate has a denominator that includes the failures.

Practical notes

• Detailed analytics requires an active subscription — a lapsed workspace shows the gate message until renewal (Plans & billing).
• Embeds can opt out per instance with the na attribute — keep internal dashboards and staging pages out of your production numbers (Attributes reference).