Analytics and conversion tracking
Overview
Your agent can read what is happening on your store and act on it. It can query your own BigQuery data warehouse, read a shopper's screen recording, read the server's render logs to find a broken template, and read or change what the storefront reports to analytics, ad and affiliate platforms.
This area covers:
- The data warehouse.
uc-storefront warehouse queryruns SQL against your UltraCart BigQuery project through your ownbqtool and Google credentials. The results go from BigQuery to your machine and nowhere else. - Screen recordings. Check whether recording is on, read a recorded session with named events such as rage clicks and script errors, and replay chosen page views locally.
- Render logs. Read the server log for a page render, the same logs the Developer Tools panel shows.
- Tracking and privacy settings. Privacy rules by region, advanced conversion settings, UTM shortcodes, Smart Attribution, tracking providers, custom tracking and conversion HTML, and affiliate networks.
Example requests
- "Which ten pages had the most sessions last week, and what was their conversion rate?"
- "Find three recordings of shoppers who abandoned the checkout on a phone, and tell me what they had trouble with."
- "The /about page shows an error. Check the render log."
- "Show me which tracking providers are on, and what each one is sent."
Commands
| Command | What it does |
|---|---|
warehouse query | Dry-run a query, check it against a cost ceiling and the allowed datasets, then run it |
sf recordings status, show, render | Check whether recording is on, read a session, replay chosen page views locally |
sf recordings enable, disable | Turn screen recording on or off for a storefront |
sf logs list, sf logs get | List recent render logs, or read one |
sf tracking get | Read the privacy rules and conversion settings |
sf tracking privacy, advanced, utm-shortcodes, smart-attribution | Change privacy rules, conversion settings, UTM shortcodes and Smart Attribution |
sf tracking providers, integration, html, affiliate-networks | Read and change tracking providers, custom HTML and affiliate networks |
sf tracking check | Plan a browser check of a page's tracking, then grade what the browser collected |
How warehouse queries are kept in bounds
- Every query is dry-run first. The real query runs only if its estimated scan is under a ceiling (20 GiB by default) and every table it reads sits inside the two warehouse datasets your account grants.
- Results are capped at 100 rows by default.
--audit-logrecords each query's SQL, estimate and row count, never the rows themselves.- You need the Google Cloud
bqtool, signed in with a Google identity that has access to your warehouse datasets.
Queries scan your own BigQuery project, so any BigQuery charges are on that project.
Costs to be aware of
Screen recording records real shoppers' sessions and is billed per 1,000 sessions once its 14-day trial ends. The agent reports whether recording is on and what it costs, and turns it on or off only when you ask. Each change emails UltraCart support.
What it cannot do
- Read tracking secrets. A provider's secret is never shown, and is written only through a handle for a value you enter in the admin. The order hash secret key is set only in the admin.
- See data from before recording was turned on. Recording collects from the moment it is turned on.
- Read orders or customers outside the warehouse.
sf orders test-orderslists test orders only, for rendering receipts; nothing changes an order or a customer.
What needs approval
Tracking changes that can affect shoppers' privacy or what runs on the storefront need approval:
- Putting HTML in a custom tracking or conversion slot, after UltraCart reviews it.
- Adding or changing an affiliate network's pixel or conversion type, or deleting a network.
- Turning a tracking provider on, loosening its consent gate, setting a script host, or turning on Google Tag Gateway.
- Clearing a provider when that removes a consent gate something else still fires.
- Changing Smart Attribution while it is active, making it active, or deleting it.
Every tracking write needs --live and publish permission. See Changes that need approval.
Related documentation
- Data warehouse (BigQuery): your warehouse and how to get access
- StoreFront Recordings: recordings and heatmaps in the admin
- Experiments: measuring a change with an A/B test