1. The consent-gap problem, and what changed in 2026
When a cookie banner is up, a meaningful share of visitors decline or ignore it. Google and Shopify shipped a server-to-server purchase event in mid-2026 that captures the order itself even when the visit was never tagged — so for Shopify merchants, the revenue total in GA4 is now reasonably solid. That's a real improvement, and we're not going to pretend otherwise.
What it doesn't fix is the path to the sale. Server-side purchase capture tells GA4 that an order happened; it does not reconstruct which channel, campaign, or visit led to it. Once a session declines consent, multi-touch attribution across it stays unrecoverable at any traffic volume — behavioral modelling estimates an aggregate total, it doesn't rebuild a journey. And below roughly €0.5–1M in annual online revenue, GA4's modelling never activates at all: sub-threshold stores get a raw undercount with no fill, on the parts server-side capture doesn't reach.
You don't have to take our word for any of this. Qusto's own reconciliation report shows you, per order, how many came from your store's own order record with no matching browser session — a real, checkable count, not a model. See it in the API docs or ask for it directly over MCP.
The second cause: buyers who never visit at all
Consent decline is one way the path to a sale disappears. Since 2026 there is a second, and it is not a regulatory problem at all. When someone buys through Google AI Mode, Gemini or Copilot, an AI agent completes the purchase without the buyer ever loading your store. No pageview, no session, no tag — not because anything is blocked, but because there was no page for a tag to run on.
Give GA4 its due here too: it shipped an AI Assistant channel on 13 May 2026 that groups referred sessions from ChatGPT, Gemini and Claude. If someone clicks through from a chatbot and lands on your site, GA4 now labels it. That is a real improvement and worth having.
But it measures clicks that reach your site, and an agent-placed order produces none. You can still push the order into GA4 yourself through the Measurement Protocol — the transaction will appear. What you cannot do is attribute it: the Measurement Protocol joins a server event to a session only if you supply the browser's client_id, and an agent-placed order never had a browser session to take one from. The prerequisite isn't missing, it's impossible. So the revenue lands under (not set).
That is the honest shape of the claim: you can tell GA4 the money arrived; you cannot tell it where the money came from. On Shopify, Qusto reads the order stream, so an agent-placed order is separated in the order record. Per-order revenue from an AI referrer waits until the shopper says yes. A decliner is not shown with a channel.
Two caveats we'd rather state than have you find. First, GA4's AI Assistant channel only counts forward from 13 May 2026 — and ours only counts forward from 11 September 2026, for the same reason. Everyone's agentic number starts the day they began counting; the question is whether your tool counts at all. Second, we close the channel gap, not the campaign gap: we can tell you an agent bought and which platform it was, but not which ad persuaded the agent to recommend you. Nobody can — the model's reasoning isn't published, and the platforms don't forward it.
2. The e-commerce blind spot
GA4 can track e-commerce events, but checkout and revenue are not first-class objects in the way a store-native tool treats them. Product → cart → checkout → revenue is something you wire up, tune, and hope survives consent changes — not the default journey the dashboard is built around. For an operator who wants to know what sells without becoming a data engineer, that gap shows up every week: channels that spend still spend, but the report cannot reliably tie them to orders.
3. A low-risk parallel run — "second analytics"
You do not have to rip out GA4 on day one. Qusto is designed to run alongside whatever you already have: add the script, keep your existing setup unchanged, and compare what each tool shows. Zero changes to your current tracking during evaluation — a materially smaller ask than a full migration, especially when your team is cautious about touching the live stack. When you are ready, Qusto is built to become first-class store analytics; until then, it is a second lens on the same store.
Beyond GA4 — the same structural problem
The same pattern applies whether you are on GA4, Adobe Analytics, Mixpanel, Amplitude, or a default Matomo install — every tool, Qusto included, loses the path to the sale when the visitor declines. The difference is whether it counts the orders with no channel. On Shopify, a channel is named only when the shopper says yes. None of these other tools treat store revenue as a native object by default.
Location intelligence — where revenue actually comes from
Generic analytics maps answer "where do my visitors come from?" On Growth and above, Qusto ties revenue, orders, AOV, conversion, and cart-abandonment visibility to country, region, and city — in the dashboard and via API. Visitor geo on Traffic is table stakes; the difference is connecting geography to revenue and where carts thin out.