The questions that come up, answered plainly.
Most of these are about Google’s side of the connection. None of them require a developer.
Connecting Google
Why does one Google sign-in ask for both Analytics and Search Console?
Because they are two halves of one picture. Analytics tells you what visitors did once they arrived; Search Console tells you what they were searching for and how often Google offered your page. You sign in once, grant read-only access to both, and we bind the property for each that belongs to your domain. You can untick Search Console on Google’s consent screen and still connect; the dashboard will simply show search data as not connected until you add it.
Do I need a “Domain property” in Search Console? It wants me to edit DNS.
No. Search Console offers two kinds of property, and we accept either:
- URL-prefix property — for example
https://www.example.com/. This is the easy one. Google can usually verify it automatically through the Analytics tag already on your site, with no DNS changes. If a verified URL-prefix property exists for the address your site uses, we connect it immediately. - Domain property —
sc-domain:example.com. Cleaner, because it covers www and non-www, http and https, and every subdomain in one place. But creating one requires adding a DNS record, which many site owners cannot or would rather not do. It is optional. If you have both, we prefer the Domain property.
We will never block your setup because you lack a Domain property, and we will never ask you to touch DNS.
Search Console is connected but shows nothing. Why?
Three common reasons, in order of likelihood:
- The property is for a different address than your site actually uses. A URL-prefix property for
https://www.example.com/holds no data if visitors land onhttps://example.com(or vice versa). Google indexes the address people end up at. Your dashboard checks this and says so in plain words; the fix is to add a URL-prefix property for the address it names, then click Re-match property. - The property is new. Search Console starts collecting when the property is created and finalises data about three days behind. A brand-new property is legitimately empty for a few days.
- Your site genuinely has no search impressions yet. New or small sites often do not. The connection is working; there is nothing to show.
Why does the search data stop three days ago?
Google finalises Search Console data roughly three days after the fact. We report only the window Google has finalised, so you never see a fake dip on the most recent days. Analytics data is reported through today.
Which Analytics property gets connected?
The one that is actually measuring your domain. We look for a property whose tag is installed on your homepage, or whose configured website URL is your domain. We never offer a menu of every property your Google account can see, so you cannot accidentally attach another site’s analytics to yours.
It said “Google no longer accepts the saved authorization.” Did something break?
Not on your site. While our Google app is in its pre-launch (testing) phase, Google expires the connection every seven days by design. Click Reconnect and it resumes. Once the app completes Google’s verification, connections stop expiring.
What exactly can Outcomes By Page do inside my Google account?
Read reports. Both permissions are read-only: analytics.readonly and webmasters.readonly. We cannot change settings, create or delete properties, or write anything into your Google account. You can revoke access at any time from your Google account’s security page or by clicking Disconnect in your dashboard; either stops the connection immediately.
Installing the snippet
The installation check says the loader appears twice. I only pasted it once.
The second copy is almost always the original verification tag, still sitting in a different place than where you pasted the new snippet — typically a “footer scripts” field in your CMS or theme while the new snippet lives in the header. Only the first copy runs, so nothing is broken, but remove the older one to keep the check clean. Verification does not depend on it.
Where does the snippet go?
Inside <head>, above any Google tag. Most site builders have a “header code” or “code injection” field for this. Install exactly one of the two snippets shown: the version with the consent banner, or the analytics-only version if you have determined you have no obligation to ask visitors for consent.
My CMS stripped part of the snippet.
Some header-code fields remove <script> tags that carry attributes they do not recognise. The consent snippet is written to survive this, but if the check reports that only half of it is present, paste the whole block again and, if it happens twice, use a “custom HTML” or “code injection” feature rather than a generic header field.
Reading the results
Findings says “Not enough traffic yet.” Is that a problem?
No. It is the honest answer. A finding is only reported when a difference between two groups of pages is large enough, and measured on enough visits, that it cannot reasonably be chance. Roughly a hundred sessions or measured views per segment is the floor. Below that, we say so rather than guess. As traffic accumulates, findings appear on their own; you do not need to change anything.
Why does “Setup and measurement” only tell me about plumbing, not performance?
Deliberately. Anything of the form “these pages do better than those” has to earn its place statistically and appears under Findings. The setup notices are about whether measurement is working at all: connections, outcomes named, pages tracked, decisions pending. A hand-written rule that stated a performance claim without testing it would be an opinion dressed up as a finding.
Why are there two “position” numbers that do not average?
Search position is an average, so combining pages by taking the mean of their means would be wrong. We weight each page’s position by its impressions, which is why a group’s position can be closer to the page that Google shows most often.
What does the AI actually decide?
Less than you might expect. It reads each page and describes what the page is and what a visitor can do on it, using your site’s own vocabulary and, where available, the words people actually searched for to reach that page. It proposes segments worth testing. It writes the plain-language explanation of a finding after the statistics have already decided the finding is real. It never decides whether a difference exists, never sees a number it did not compute, and nothing it writes is executed. Every step records the model and prompt version used.
Something missing? [email protected].