Anamap Blog
Cartos vs GA4 AI for signup conversion analysis
AI & Analytics
Updated 2026-11-13
Google Analytics 4 (GA4) includes Ask Advisor for questions about your analytics. Cartos is Anamap’s AI analyst for investigating business questions across connected data sources. We wanted to see what Cartos could add for a growth team already using GA4.
We asked both assistants to follow a signup cohort into onboarding and recommend where the team should spend its time: the signup flow, onboarding, or measurement.
Ask Advisor reported 19.2% by dividing the number of users who recorded a completion (5) by the number who recorded a start (26). In its own run, Cartos tried to follow the same people from signup start to completion. The GA4 funnel report came back empty, so Cartos held off on giving a rate. We preferred Cartos’s answer because the two counts alone didn’t show how many of the 26 starters went on to finish.
Anamap makes Cartos. We asked the same business questions in separate chats on the same synthetic GA4 property. Ask Advisor ran on November 5, 2026, and we reran Cartos on November 13 with an explicit instruction to use only GA4. We compared the answers after both runs.
What Cartos did well
Cartos kept the investigation in GA4 and tested the ordered signup path, including the existing submission step and the original cohort’s onboarding follow-up. It explained exactly what was missing before the team could use the conversion rate.
What GA4 Ask Advisor did well
Ask Advisor surfaced the separate signup totals in a visible GA4 report. It also found onboarding activity across the property, giving the team a starting point for checking which milestones were tracked.
Which is better for your business and why?
Choose Cartos when a signup number will guide a product decision. It checked the underlying sequence and proposed testing the removal of one required field while tracking submissions and completions. Ask Advisor is useful for a quick read of GA4 event totals. In this test, the ordered conversion rate still needed verification before either answer could justify changing the form.
Results at a glance
Swipe or scroll sideways to compare both assistants.
| What the team needed | Cartos, using GA4 only | GA4 Ask Advisor |
|---|---|---|
| Verify the signup conversion rate | Tried to follow the same people from signup start to finish. The GA4 funnel returned no step rows, so Cartos withheld the rate. | Reported 19.2% from separate GA4 event totals. |
| Test the existing submission step | Queried the requested three-step path. GA4 returned no step rows, leaving the effect unresolved. | Discussed adding a new form and how it might affect friction. |
| Follow the original signup cohort | Extended the same signup path to an onboarding step and to activation. GA4 didn’t verify either outcome. | Surfaced onboarding activity across the property. |
| Choose a first experiment | Remove or defer one required field and watch submissions and completions. Treated form friction as a hypothesis. | Suggested simplifying the form or moving signup later. |
What we asked and why
We asked four questions in order, keeping each set of follow-ups in the same conversation:
- Opening question: “Of the people who started signup Oct 31-Nov 1, 2026, what share finished?”
- “What changes if we require a form submission in between? Can we see why?”
- “Did those signup finishers complete onboarding?”
- “What should we improve first?”
We chose this sequence because a team needs more than a conversion percentage to decide what to fix. The first question asks whether the same people started and finished signup in order. Adding submission changes the path they must take. The onboarding follow-up checks whether the assistant stays with the original finishers. The final question asks it to turn those findings into a recommendation the team can assess.
How we set up and checked the test
- Same property and entry dates. We opened fresh chats on the synthetic Anamap Demo GA4 property. The prompt set October 31 through November 1, Pacific time, as the entry dates and left the conversion window unstated.
- Instructions. We added this to Cartos’s opening question: “For this entire conversation, use only Google Analytics 4 (GA4), the connected Anamap Demo property. Do not use any other data source, including Amplitude, Mixpanel, or website/browser data. If GA4 cannot verify an answer, say what is missing and stay in GA4.” We supplied neither assistant with event names or expected counts.
- Verification. We checked the visible reports and Cartos’s source activity for evidence that the same users took the steps in order. GA4-only Cartos rerun
Why the 19.2% answer needed more evidence
Ask Advisor showed 26 users with a signup-start event and five users with a completion event, then reported 19.2% conversion. The arithmetic is straightforward. The relationship between the two groups needs another check.
A signup funnel follows the same people from starting to finishing, in that order and within a defined time window. Separate event totals can contain different people or events recorded in a different order.
Imagine taking the result into a planning meeting. If the five finishers belong to the original group, the percentage measures their completion rate. If some started before the requested period, the same calculation answers a different question. The team needs that distinction before using the result to judge a change.
Cartos retrieved the same event totals after several funnel attempts. It declined to report 5 divided by 26 as the conversion rate because GA4 hadn’t established the overlap. That was the better answer to take into a planning meeting. The team could use the counts as a starting point and knew what it still had to verify.
The follow-ups exposed two different problems
Requiring a submission
When we required a submission between signup start and completion, Ask Advisor treated the question as a proposal to add a new form to the product. Cartos identified the existing submission event and tested the three-step funnel in GA4. It found 10 submission users in the event totals. The ordered report returned no rows, so Cartos could suggest inspecting the submission step but couldn’t establish how many users dropped there or why.
Following the original finishers
The onboarding question then exposed a problem with following the original users. Ask Advisor found 214 onboarding starts and 210 completed onboarding steps across the property. Those totals leave open how many of the five signup finishers took either action.
Cartos extended the original signup path in GA4, first to a completed onboarding step and then to activation. Both funnel requests returned no rows. It could confirm onboarding activity across the property, while leaving the outcome for the original finishers open. It also flagged that the available event mapping didn’t confirm a dedicated onboarding completion event.
Which recommendation would we act on first?
Ask Advisor suggested simplifying the form or moving signup later in the journey. It introduced industrial B2B examples that did not fit the demo and treated zero recorded errors as support for a friction explanation. Before relying on that interpretation, the team would need to check that relevant errors were being captured.
Cartos proposed removing or deferring one required field, then tracking submissions and completions. That gave the team a smaller change to test. Cartos flagged that GA4 hadn’t verified the sequence, though its recommendation still treated the gap between the event totals as a clue about form friction.
Verify the signup sequence in GA4. Agree on the event that marks onboarding, then follow one known signup through the handoff before choosing a product experiment.
Bring a concrete question, such as signup completion for a specific set of starting dates. Then ask the assistant to follow those same people into your first meaningful product action and show which evidence supports its recommendation. Try the investigation with Cartos
Our Amplitude comparison shows a run where Cartos added another behavioral check before recommending a change. The Mixpanel comparison examines how the tools handled a follow-up that needed the original user group.
Test notes
In the first take, Cartos reached for Amplitude when GA4 came back empty. It was a resourceful detour for an investigation and an awkward one for a GA4 comparison. We reran Cartos with the source fixed to GA4 and used those answers here.
This comparison covers one synthetic property and two test dates. The GA4-only rerun verified how Cartos handled missing funnel evidence. It didn’t establish the actual signup conversion rate or how many finishers completed onboarding.