Journeys Explorer
Ask your AI app for the paths to purchase, a funnel or a flow. MarStack builds it from your GA4 BigQuery export and returns a link to an interactive explorer in the portal, where every filter and view is free.
Who can use it
- Starter, Pro and Agency plans.
- You need the GA4 BigQuery export (free to turn on in GA4: Admin, Product links, BigQuery links). To try it first, ask for a demo on Google's public sample store.
Ask for it
In Claude, ChatGPT, Gemini or MarStack Intelligence, ask for example "Show me the paths to purchase over the last 28 days". The assistant works in three steps and checks with you between them:
- Suggest (free): looks at your data and suggests goals, steps and page groups, then asks whether one journey is one visit or one person across visits, which goal, and which steps lead to it.
- Estimate (free): the BigQuery size and the MarStack credits, and a warning if your export is missing any days in the range.
- Build: runs one read-only query in your own Google Cloud project, then answers with the funnel, the biggest drop-off, the top paths and the explorer link.
What the explorer shows
| View | What it answers |
|---|---|
| Funnel | Any steps in order (up to 10), counted the way GA4 closed funnels count them, with the drop-off at each step, the median time and time spread between steps, and where people went instead. Compare two groups. |
| Flow | Every visit from session_start, step by step. Click a step for its audience and timing. Its menu has Start here, End here, Only sessions through here, with or without an event, Hide, Group with, Split by (landing page, device, channel and more), Set as goal and Add to funnel. Colour routes by how many reach the goal, or by which of two groups they lean to, and drag the timeline to pick days. |
| Paths | The most common complete journeys, with share, conversion, revenue and median duration. |
| Goals and exits | Which paths lead to each goal, and the last step before people left. |
| Segments | Device, channel, country, source, medium, campaign or landing page side by side. |
| Full funnel | With Google Ads or CRM data: ad cost and clicks, journeys, the website goal, leads, won deals and revenue by campaign or channel, with cost per session, lead and won deal, and return on ad spend. |
Filter every view by date, device, country, channel, source, medium, campaign, landing page, or journeys that include or skip a step. Click a step to show only the journeys through it.
Leads and won deals from your CRM
If your CRM data is in BigQuery and mapped in MarStack (Platforms, CRM, BigQuery), journeys can end in a lead, a deal or a won deal.
- Join check (no credits): MarStack counts how many CRM records have each ID (GA4 client ID, GA4 user ID, Google click ID or Meta click ID) and how many of those it finds in GA4, then recommends one. A join under 30% is flagged as weak; under 5% it suggests the Full funnel by campaign instead.
- Joined build: "Lead" and "Won deal" become funnel steps and goals, and won value is added. Each CRM record counts once, on the latest journey with its ID. You choose how many days after the range to count CRM records (30 by default, up to 90).
- No shared ID? The Full funnel still matches CRM records to journeys by their UTM campaign, from a mapped table or your MarStack CRM uploads.
Credits
- 30 credits for a build of up to 28 days, +15 for each further 28 days (up to 92 days).
- +20 credits for a cross-platform build: a CRM join, Google Ads cost, or CRM by campaign.
- Suggest, estimate, the join check, and everything in the explorer are free. A failed build is refunded.
- BigQuery is billed to your own Google Cloud project at Google's price; most builds fit inside BigQuery's free monthly allowance. The estimate shows the size first, and your workspace cap applies.
Accuracy
- Funnels count the way GA4's closed funnel exploration does: each step is the first time it happens after the step before. Checked against GA4's own funnel report and an independent query.
- Sessions and users match GA4 for the days your export holds. If the export started later or a day has not arrived, MarStack names the missing days, because GA4's reports will show more for those dates.
- Visitors who declined analytics cookies are not in the export, as in GA4's reports.
- With more than 100,000 journeys, the Flow and Funnel use a random sample of 100,000 and say so; the summary, Paths, Goals and Segments count every journey.
Privacy
- The query and any CRM join run inside your BigQuery. MarStack receives totals plus up to 100,000 anonymous step lists (steps, timing, device, country, traffic source and landing page path) so the Flow and Funnel respond instantly. IDs never leave your project.
- Anything seen by fewer than 10 journeys is grouped as "(other)" or "Smaller steps". Page paths lose their query strings.
- Builds are private to your workspace and deleted after 30 days.
Problems? See Troubleshooting, or the explore_journeys tool reference.